HYPOTHESIS Nouns are in their theta positions: the patient in VP, the agent in Spec,vP and then Spec,TP. A single noun names the lowest third-person participant unless its meaning rules that out (H4″): łį́į́ʼ dzaanééz yiztał (NX-029) vs dzaanééz łį́į́ʼ biztał (NX-030); ʼashkii yiyííbį́į́ʼ (NX-027). Details: results §1.
HYPOTHESIS In the inverse the patient moves above the raised agent (H3); this happens before the question clitic is placed (*NX-086). Results §2.
HYPOTHESIS Question words stand where nouns stand; Q is in C⁰ and pronounced on a host (H11, H12). Results §3.
HYPOTHESIS Sentential negation has its head in the final da, above TP, with doo adjoined lower (H23); hanii is constituent negation (H20). Results §4–6.
HYPOTHESIS Surface order does not fix quantifier scope: the inverse-scope reading of NX-267 needs one more step (H22a or H22b, OPEN). Results §7.
HYPOTHESIS Reported speech (review 12): who “I”, “you” and “he” can be before ní / jiní “said” follows from context shift and competition with dedicated forms (H41–H43; H44 a stipulation, NOT UNDERSTOOD): doogááł ní (NX-387), “Heᵢ said heⱼ will be there.” These facts do not decide whether the reported clause is a separate clause (H40a) or the object of “say” (H40b); H40a stays slightly ahead on review 11’s “when” contrast. Results §10.
HYPOTHESIS=go clauses (review 13): the second of two linked transitive clauses reuses the first clause’s people (H45, mostly a stipulation); the =go clause attaches above the main clause’s Spec,TP (H47), since a main-clause noun subject follows it: yah ʼííyáa=go ʼatʼééd dah neezdá (NX-397). The (9a)/(9b) source conflict is OPEN; a telicity-dependent =go form is contradicted. Results §11.
OPEN Whether T⁰ is final (nothing else moves) or initial (a remnant VP moves): headedness; where each element sits: clause spine.
2026-11-14 overhaul (approved by Grant for the live site, 2026-11-14). This page used to be framed by the A/B headedness changeover (“the word is the changeover point”). It is now framed by the spine (derivation): V→v→T and subject raising to Spec,TP are assumed in tensed clauses, so the head-direction question becomes which heads are final and what else has to move. The changeover appears below only as one labelled contrast. The earlier sections 1–9 are kept unchanged under History. No Navajo form, gloss, judgment or citation changed.
Navajo puts the verb last, and nouns, postpositional phrases and most clause particles before it or after it in fixed ways. Once the verb is known to raise into T in every tensed clause, there are only a few ways to get that order. Either T⁰ (and the heads above it) are final, and nothing else has to move; or T⁰ is initial, and everything left in the verb phrase has to move above it. This page sets out those hypotheses, the sourced evidence for and against each, and what would decide between them.
HYPOTHESIS Fixed by the spine: VP with the internal argument, vP with the external argument in Spec,vP, T⁰ above vP, V→v→T, the subject in Spec,TP, and, when present, Neg⁰ above TP and C⁰ above that. Specifiers are on the left. What is not fixed is whether each head comes before or after its complement. Lexical V⁰ is final on every hypothesis below, since objects precede the verb (WALS 83A, OV; §2.1). So the question is about v⁰, T⁰, Neg⁰ and C⁰, and about what has to move given each choice.
The live hypotheses
Hypothesis
Heads
What else moves to give verb-final order
Cost
Status
HS-1 head-final clause
v⁰, T⁰, Neg⁰, C⁰ final
Nothing. The verb in T⁰ is pronounced after everything in TP.
No extra movement. Initial elements (da’, doo) must be non-heads: specifiers, adjuncts or operators.
HYPOTHESIS; cheapest on present data
HS-2 T-initial with remnant VP
v⁰, T⁰ initial; Neg⁰, C⁰ final
The remnant VP (object, low PPs, adverbs) moves above T⁰, in every clause.
One extra movement per clause with no independent trigger yet; mixed directions (initial T⁰ under final Neg⁰).
HYPOTHESIS; not excluded
HS-3 antisymmetric roll-up
All heads initial (Kayne 1994)
Each complement moves to the left of its head at every level.
Most movement; every final particle needs its own complement-raising step.
HYPOTHESIS; no Navajo test run
Analysis A: Working default with T⁰ final (HS-1): nothing else has to move for verb-final order Hypothesis
Analysis B: Working default with T⁰ initial (HS-2): the remnant VP must also move above T (HYPOTHESIS) Hypothesis
(1) Schematic (placeholder leaves, no Navajo example): the whole clause spine with every optional layer present Hypothesis2026-11-14 overhaul draft (Grant, 2026-11-13: VPISH aligns syntax and semantics; V→v→T and subject raising are assumed in tensed clauses). All leaves are generic placeholders; no Navajo form. The layers drawn are the ones the first-principles reviews needed: vP with the external argument in Spec,vP, TP, NegP with final Neg⁰ (H23, review 8) and CP with a polar question feature (reviews 4 and 8). Head directions of T⁰, Neg⁰ and C⁰ are hypotheses; see /headedness/#spine-2026-11-14.Bracket notation (copy)
DP (id S-high; landing site of Move from subject): agent a
T′
AdvP (id negmark): lower negative marker
T′
vP
DP (id subject; goal of Agree φ (optional); struck copy; moved to S-high): agent a
v′
VP (id lowVP)
V′
DP (id object): theme t
V⁰ (head; id V; head-moves to v): predicate
v⁰ (head; id v; head-moves to T)
T⁰ (head; id T; probe for Agree φ (optional) with subject; hosts the “verb word” (V + v + T))
Neg⁰ (head; id Neg)
C⁰ (head; id C; [Q, polar])
Analysis B
CP
C′
NegP
Neg′
TP
DP (id S-high; landing site of Move from subject): agent a
T′
VP (id VP-high; landing site of Move from lowVP)
T′
T⁰ (head; id T; probe for Agree φ (optional) with subject; hosts the “verb word” (V + v + T))
vP
DP (id subject; goal of Agree φ (optional); struck copy; moved to S-high): agent a
v′
v⁰ (head; id v; head-moves to T)
VP (id lowVP; struck copy; moved to VP-high)
V′
DP (id object): theme t
V⁰ (head; id V; head-moves to v): predicate
Neg⁰ (head; id Neg)
C⁰ (head; id C; [Q, polar])
Figure: HS-1 (panel A) and HS-2 (panel B) on a schematic negative yes/no question. Placeholder leaves only.
Evidence, head by head
SOURCEDNegation. The position of da decides which clause is negated: Jáan Mary doo chidí neidiyoołnih da yóʼní. (NX-180; Perkins & Fernald 2014:88 (38a)), translated “John expects Mary not to buy a car.”; Jáan Mary doo chidí neidiyoołnih yóʼní da. (NX-181; Perkins & Fernald 2014:88 (38b)), translated “John doesn’t expect Mary to buy a car.” A minimizer before doo is still under the negation: Béeso łáʼí ndi doo naashʼáa da. (NX-287; Perkins & Fernald 2014:104 (20 (§5.3))), translated “I haven’t so much as one dollar.” HYPOTHESIS So Neg⁰ is the final da, and doo is a lower adjoined marker (H23, review 8), not an initial head. This fits HS-1 and does not distinguish HS-1 from HS-2, since both have Neg⁰ final.
SOURCEDClause-final particles.ya’, -go and the relative -ę́ę are final (§2.1, §4); the relativizing enclitic is at the right edge of its clause (review 9). HYPOTHESIS If they are heads (C⁰, D⁰), they are final heads, which fits HS-1 and HS-2 alike (both have final C⁰).
SOURCEDThe polar particle da’ is grammatical only clause-initially (Navajo Questions v3.2022, exx. (2), (3a–c); §2.1). OPEN If it is C⁰, C is initial and HS-1 fails for C. If it is an operator in Spec,CP, its initial position is neutral, since specifiers are on the left. No test of its category has been run.
SOURCEDEnclitics-ísh and =sh follow the word or phrase they attach to, and their host varies (§2.1; reviews 3 and 8). HYPOTHESIS Their host is not their scope position (H12), so they say nothing about head direction.
OPENT⁰ itself. No sourced example yet places low material (a VP adverb, a selected PP) relative to the verb in a way that shows whether that material stayed in VP (HS-1) or moved with a remnant VP (HS-2). The PP/doo orders in NX-162–164 are compatible with several attachment sites (§9.5).
HYPOTHESIS*NX-246 (hanii cannot follow “object + verb”) follows from V being in T on both HS-1 and HS-2, given H20, so it supports the working default but does not choose a head direction.
HYPOTHESISWorking verdict. HS-1 needs no movement beyond the working default and fits every sourced order tested, provided da’ is a specifier or operator. HS-2 fits the same orders at the cost of a remnant-VP movement in every clause. HS-1 is the cheaper hypothesis, not a result. OPEN
The verb word
OPEN Inside the verb word the stem is last and the prefixes come before it, in a fixed order (§2.2–2.3; Young & Morgan’s position classes are the record of that order). On the spine the verb word is the V+v+T complex.
GRANT'S HYPOTHESISRestated 2026-11-15 on Grant’s correction. Grant’s hypothesis: headedness is fixed locally by the domain in which a head is Merged, not by one parameter (his working draft Merge Sites and Agree Sites, Aug. 2026, §2, §§3.1–3.3), and Navajo’s sublexical domain is linearized opposite to its phrasal domain. So the order of heads inside the word is predicted to be the mirror image of their order in the clause: in the clause a higher head is further right, in the word further left, with the stem at the right edge. That predicted order is the one Baker’s Mirror Principle gives for prefixes; the sourced orders that go the other way (Rice’s wider scope nearer the stem) are misfits for it, which is why the site no longer says simply that “the mirror image” is predicted (fact-check, row 14). Because prefix order is then read as height, the word-level claim is a scope claim. This is the site’s original head-direction mismatch (“the syntax is right headed and the morpho is left headed”, §1 below), now stated as a consequence of Merge sites rather than as a single changeover point. What it predicts for the prefix-to-prefix order, how it fares on the sourced template (it fits the conjunct core; aspect, object/subject, distributive and preverb orders are OPEN misfits), the question of which element counts as the head of the word, Grant’s two-tier linking proposal (prefix pronouns on a head-to-head tier, nouns on a spec-to-spec tier, tied by a linking element) and his earlier, untested sublexical V-to-T are set out on the unified structure, §3.3–3.4.
The headedness changeover, as labelled contrast
Grant’s earlier proposal (§1 below, 2026-10): “the syntax is right headed and the morpho is left headed”, with the word as the changeover point (rule set A up to the word, rule set B above it). The site tested it in sections 3–9 below. Result, kept as history: no head-by-head boundary was found, and the changeover was never established. On the spine it is not needed as a parameter. Its clause half is HS-1. Its word half is restated (2026-11-15) as Grant’s Merge-site hypothesis, under which the mismatch follows from where heads are Merged rather than from one changeover point (the verb word; unified structure §3). Sections 1–9 below keep every datum the changeover was tested on, and their analyses (A/B rule sets, CYC, the B1/B2/B3 boundaries, the Mirror Principle and template rivals) are contrast, not the frame.
What would decide
OPEN Between HS-1 and HS-2: one matched paradigm, from a source or a speaker, with a transitive verb, an independently low VP adverb and a selected PP, in a negative clause, direct and inverse. If the low material always stays together with the object and can be separated from material that is clearly above TP, that favours a remnant VP (HS-2) only if it also moves as one unit across a fixed landmark; if it can be stranded while the verb is still last, that favours HS-1. For da’: a test of whether it selects the clause (a head) or occupies its edge (an operator). No Navajo sentence is constructed here; see §9.7 for the earlier statement of the missing test.
Headedness fact-check (2026-11-15)
2026-11-15 · requested by Grant, who wrote that he “may have the facts about headedness a little confused”. Every headedness claim on the core pages, and in Grant’s working draft Merge Sites and Agree Sites: Exploratory Notes (Aug. 2026, §2, §§3.1–3.3), was checked against sources opened for this pass. Verdicts: CORRECT; LOOSE (true, but the wording allows a false reading, usually about which element counts as the head); WRONG; UNSOURCED (no opened source either way). Following Grant (2026-11-15: “a unified syntax and semantics, not just word order”), each row also says whether the claim affects interpretation (scope, binding, theta or event structure) or is about linearization only. No Navajo form, gloss, judgment or citation was changed, and no Navajo data were added.
Plain words. The facts themselves are solid: Navajo is verb-final (SOV, OV), has postpositions and possessor–noun order, and builds its verb with prefixes before a final stem. That combination, verb-final with prefixes, is typologically rare. The confusion is in one word, “head”. If the stem is the head of the word (Williams’s Righthand Head Rule), the word is right-headed like the clause, and nothing flips; Grant’s notes §3.3 say this, and it is also Rice’s analysis. If each prefix is the head of its own layer (Hale’s analysis), the word is head-initial and the clause head-final, and that is the flip Grant stated on 2026-11-15. Both describe the same string; only the second makes predictions about the order of prefixes among themselves. Those predictions are the Mirror Principle’s order for prefixes (a higher head further left), and Rice’s generalization (wider scope nearer the stem, positions I–VIII) goes the other way for exactly the four orders the site lists as misfits. Grant’s notes also contain one plain factual slip, about Celtic (row 13), and two loose descriptions of Navajo (rows 6 and 9).
HYPOTHESISWhat headedness means for meaning. On Grant’s Merge-site view, headedness is a local linearization decision (notes §2, §3.1). The direction itself carries no meaning: scope is c-command, which is the same whether a head is pronounced before or after its complement. What carries meaning is the height of a Merge site, as the notes themselves say for applicatives (“Once the height is set, the semantic contribution and the available diagnostics (binding, scope, reconstruction, measuring-out) follow automatically”, §3.4). So in the clause, the choice between HS-1, HS-2 and HS-3 is linearization only, as long as the remnant VP of HS-2 lands below the subject. In the word the situation is different: once the direction of a domain is fixed, the order of prefixes is a readout of their heights, and so of their scope. That is why the word-level claims below are testable with scope facts, and why Rice’s scope data bear on them.
Order only for the verb’s place. Which noun comes first is interpreted: with two nouns the order fixes the roles “by default”, and swapping them changes the meaning or forces bi- (Willie 1991:60 (20), 71; theta-args D4).
WALS 81A (SOV), 83A (OV), 82A (SV), Navajo datapoints (Dryer; Young & Morgan 1987)
None.
2
Postpositions; possessor before noun; lexical V, P and N take their complement on the left.
§2.1; unified §2b
CORRECT
Order only.
WALS 85A (postpositions), 86A (genitive–noun); Hale 2003:32 (“an arrangement which extends to postpositional and nominal (possessive) projections as well”) and n. 4
None.
3
The verb word is prefixing, with the stem last; tense-aspect is prefixal; a small suffixal residue on the stem.
§2.2, §6; unified §3
CORRECT
Order only, until a claim about prefix heights is added (rows 7, 14).
WALS 26A (strong prefixing), 69A (tense-aspect prefixes); Hale 2003:14–15 ex. (10), after Young & Morgan 1987:37–38; Goad & Travis 2021:3 ex. (1); Hale 2003 n. 10 (suffix -ł)
None.
4
Prefixing plus verb-final order is typologically unusual.
Yes. The classifier spells a value of v⁰ on this site (H59), so whether it belongs to the stem or is its own head is a claim about argument structure, not just order.
Hale 2003:14–15 ex. (10) and Goad & Travis 2021:3 ex. (1) list the classifier (Voice, position IX) as its own position before the stem (X). Grant’s notes §7 say so too: the position-IX classifier “is a different morpheme”.
Read “stem (Root + classifier)” as “the classifier (IX) plus the stem (X)”. Not changed in Grant’s notes; stated here.
7
The word is “right-headed at the word level—the morphological analogue of a head-final projection” (notes §3.3), versus “morphology head-initial, phrasal syntax head-final” (Grant, 2026-11-15).
Only the second reading has interpretive content: it orders prefixes by height, so by scope. The first says only that the stem is last.
Williams 1981:248 (Righthand Head Rule: “in morphology, we define the head of a morphologically complex word to be the right-hand member of that word”, as quoted in the isiNdebele paper at doi:10.5842/56-0-788; Glottopedia, “Righthand head rule”). Hale 2003 n. 4: “the structure projected by the lexical category V is right-headed, while the structures projected by the higher nuclear elements (v, M, etc) is left-headed”. Hale 2001:680: Navajo’s functional projections “are taken to be left-headed, rather than right-headed as in Rice’s analysis”.
Unified §3.1 already separates the two readings; this row adds the published sides: right-headed word = Rice’s analysis (no flip); prefix heads initial = Hale’s (the flip). Note that Hale’s split is lexical versus functional, not word versus clause: on Hale’s assumption alone the clause’s functional heads would be initial too, which the final da, yaʼ and =go are not (§2.1). Grant’s word/clause split is therefore a genuinely different claim.
8
Prefixes are “ordered leftward by successive head movement or by templatic Merge” (notes §3.3).
Grant’s notes §3.3
LOOSE
Head movement itself is not taken to change scope here, but the order it predicts is a scope order (row 14).
Goad & Travis 2021, abstract and §3: “Athapaskan verbal morphology appears to violate the Mirror Principle of Baker (1985) in multiple ways”; the head-movement accounts they review need extra operations (Speas 1991b: lowering with suffixation and prefixation; Harley 2011: two head movements, Local Dislocation, morphological merger; their own: root prefixation plus prosody).
Successive head movement gives prefixes only if each head attaches to the left of the moved complex (with adjunction on the left, as Kayne 1994 requires, it gives suffixes), and it then predicts the Mirror order, which Navajo violates for positions I–VIII. Stated here; unified §3.1–3.2 already list the misfits.
9
“Higher in the clause, independent DPs are ordered largely by discourse and hierarchy factors (animacy, topicality, inverse) rather than by a transparent Spec–head–complement sequence” (notes §3.3).
Grant’s notes §3.3
LOOSE (overstated)
Yes, and this is the interpretive part of Navajo order: the two-noun order fixes roles, and the inverse puts the patient above the agent, so it scopes over it (unified §4, inverse row).
WALS 81A, 83A, 85A, 86A (rigid verb-final, postpositional, possessor–noun); Willie 1991:60 (20), 71 on the site (D4: “Navajo order is not free”)
Read as: the verb, postpositions and possessors are rigidly head-final; what animacy and topicality govern is which noun may come first and the yi-/bi- choice that goes with it, not free order.
10
“Headedness is not a language-wide parameter. It is a local linearization decision” (notes §2, §3.1).
Grant’s notes §2, §3.1; unified §3.1
GRANT'S HYPOTHESIS (a theory claim, not a fact)
By itself, linearization only; meaning comes from height (see above).
Published precedent for a domain split inside Navajo: Hale 2003 n. 4 (lexical right-headed, functional left-headed); Hale 2001:689 (“selection … is left to right, in accordance with the parametric head-initial character of the extended projection of the verb”)
Already labelled. None.
11
German: vP and TP head-final; the finite verb moves to C in root clauses, giving V2; embedded clauses with a complementizer lack the move (notes §3.2).
Grant’s notes §3.2; unified §3.1
CORRECT (the standard analysis)
Order only: “the movement merely carries already-valued features higher” (notes §3.2).
Dooley Collberg 1990:62 summarizes the V2 analyses: the verb lands in COMP, which is why V2 is restricted to main clauses (Koster 1975, Platzack 1986, as cited there)
None.
12
Preverb and stem are one lexical unit at opposite edges of the word, like Dutch and German particle verbs in V2 clauses.
§2.3; unified §3.2–3.3
CORRECT
Yes for theta/event structure: the preverb–stem pair is one lexical predicate, so the split is a linearization fact about one meaning.
Hale 2001:678 (“much in the manner of Dutch and German verb-particle arrangements in verb-second finite clauses”)
None.
13
“Many verb-initial languages (certain Celtic and Austronesian varieties) … verb movement produces VSO in matrix clauses, but embedded clauses often surface with different orders” (notes §3.2).
Grant’s notes §3.2 (not on the site)
WRONG for Irish; UNSOURCED for Austronesian
Order only, but it matters for where the verb lands (T, not C).
Dooley Collberg 1990:62, ex. (4a): in Irish “even [+fin] embedded clauses containing an overt lexical complementizer obligatorily show VSO order”; only non-finite clauses differ (4b–c). Dooley Collberg concludes that the Irish verb lands in T, not COMP (p. 62); McCloskey 1996 (NLLT 14) argues against V-to-C for Irish too (not opened)
Irish is not a parallel to German V2: its verb-first order holds in embedded finite clauses too. The point of §3.2 (verb movement to a particular site) survives, with Irish as movement to T rather than C.
14
“The mirror-image prefix order is the prediction, not a problem for head movement.”
Yes. The prediction is a scope order: the leftmost prefix should take widest scope. Rice’s generalization is the opposite: wider scope nearer the stem.
Baker 1985 as summarized in Goad & Travis 2021:2; Rice 2000:4 as quoted in Goad & Travis 2021:2 (“morphemes of greater scope appear to the right of morphemes within their scope”, said of positions 1–8); Hale 2001:689 (among adjuncts “scope is right to left”); Kibrik 2005 §0 (“nearly mirror-image” to the usual root–aspect–tense–mood–person order)
Reworded on all four pages. What is predicted is the order that mirrors the clause, with a higher head further left in the word. That is the order Baker’s Mirror Principle predicts for prefixes. Kibrik’s “nearly mirror-image” order is the opposite: Navajo against the Mirror Principle. Grant’s correction stands (the flip is the prediction, and it comes from Merge sites). What is OPEN is that the sourced order follows it in the conjunct core and goes against it in the four misfits, which are all instances of Rice’s one generalization.
15
The conjunct core (Qual > M/S > Voice > V) fits the prediction.
unified §3.2
CORRECT as stated, but analysis-dependent
Yes: it is a claim that the Qualifier scopes over Mode/Subject.
Hale 2001:680 (the Qualifier selects Mode/Subj, so “the c-command and scope relations between these elements are opposite to those assigned to them by Rice”); Goad & Travis 2021:4 (they take position 6 to be v, which is low)
Unified §3.2 now says the fit rests on Hale’s hierarchy, which Rice and Goad & Travis dispute.
16
Rice: greater scope to the right.
unified §3, §3.2; §2.3
CORRECT; slightly LOOSE without its domain
Yes: it is a scope claim.
Goad & Travis 2021:2: “Rice (2000: 4) states in discussing Positions 1-8” …; Rice’s position 9 (the classifier) is not covered
“(positions I–VIII)” added on the unified page.
17
If the stem counts as the head, every prefixing language comes out “right-headed” and the contrast disappears (§1, Convention).
§1 below
CORRECT
Neither: a definition.
Williams 1981:248 (row 7)
None.
18
The clause’s visible functional particles (da, yaʼ, =go, the relative enclitic) are final; the polar daʼ is initial.
§2.1; unified §2
CORRECT
The position of da fixes which clause is negated (NX-180/181), but that is c-command, not direction: order only.
Fernald, Chavez et al., Navajo Questions v3.2022 (as in §2.1); WALS 92A (polar question particles “in either of two positions”, Young & Morgan 1980)
None.
19
Sublexical V-to-T: the stem raises through word-internal heads.
unified §3.3
GRANT'S HYPOTHESIS, untested: labelled correctly
On the site’s assumptions it does not change scope; it predicts which heads can condition the stem’s shape (locality), and that a low preverb is stranded.
—
None.
20
“Cyclic head movement under the Mirror Principle predicts the reverse: the lower argument’s d (the object) should be nearest the stem.”
Goad & Travis 2021, abstract and §2; Hale 2003 ex. (10) (object IV left of subject VIII)
None (the essay’s “Rice 2000; not opened” can now cite Goad & Travis 2021:2).
OPENWhat would settle the word-level question. Since the word-level claims are scope claims, the decisive data are scope data: a pair of prefixes whose relative scope is independently known and whose order goes one way or the other. Rice’s and Hale’s cases (iterative and plural over the preverb; Qualifier over Mode/Subject) are the ones on record, and they point in opposite directions. No Navajo sentence is constructed here.
Sources opened for this pass: Grant’s notes (text copy); WALS online, Navajo datapoints for 26A, 69A, 81A, 82A, 83A, 85A, 86A, 90A, 92A (fetched 2026-11-15); Hale 2001 (LI 32:678–693, PDF); Hale 2003 (manuscript, notes 4 and 10, p. 32); Goad & Travis 2021 (preprint PDF, pp. 1–7); Kibrik 2005 (PDF, §0, §3.1); Dooley Collberg 1990 (Lund Working Papers 36:57–72, PDF); Williams 1981 through the quotation of p. 248 in the paper at doi:10.5842/56-0-788 and Glottopedia. Rice 2000, Baker 1985, Speas 1991b, Harley 2011, McCloskey 1996 and Kayne 1994 were not opened for this pass; they are cited only through the sources named.
History: “Rule set A vs rule set B: headedness” (sections 1–9, before the 2026-11-14 rebuild)
Everything below is the page as it stood before the rebuild, kept unchanged as a record, with its audit notes. Its question was: Can "ruleset A applies up to a boundary, then ruleset B" be stated precisely enough to fail? Where it differs from the sections above, the sections above supersede it.
2026-11-11 audit pass 1. All nine JSON figures were screened again, including these changeover/question/assembly rivals and the particle figures. Their current labels/specifiers passed. Three older figures elsewhere had malformed category levels and incomplete subject spines; they are corrected on the core derivation page and the light-d essay. This hygiene check does not establish the changeover or select a movement theory. Remaining old formal rule lists and missing LF proofs are OPEN for later passes.
2026-11-11 audit pass 3. Tree fix in the changeover comparison: the low directional AdvP is a VP adjunct, so it can only modify the event. Its LF is now OUT(e) instead of OUT(e,t), which had given the object a second theta role besides Theme. The headedness changeover is still not established, and V-to-T is still neither proven nor refuted. See the pass 3 report.
Status badges now use only the label set SOURCED, HYPOTHESIS, SPECULATIVE, OPEN, NONE and GRANT'S CONJECTURE; verdict or task words that had been styled as badges (for example Opened, PRE-VPISH, DONE, SUPPORTS, PARTLY ANSWERED) are now plain text or a label with a qualifier.
2026-11-11 audit pass 5. Tree panels that move V to T now each say that V→v→T is GRANT’S CONJECTURE: untested, neither proven nor refuted. (Superseded by audit pass 7, 2026-11-13: V→v→T is now the working default.) See the pass 5 report.
In the changeover figure, the head-final/head-initial split is now worded as a hypothesis, not established; panel B’s complex head is stated conditionally.
Placeholder-only trees are now titled “Schematic” so they are not read as analyses of a particular Navajo example.
2026-11-11 audit pass 6. Visible panel titles now carry the label GRANT’S CONJECTURE wherever the panel moves V to T. Before, the title said only “Hypothesis”, and the label was in the panel’s summary only. (Superseded by audit pass 7, 2026-11-13: V→v→T is now the working default.) See the pass 6 report.
2026-11-13 audit pass 7. Tree panels now draw V→v→T and subject raising Spec,vP→Spec,TP as the working default in tensed clauses, not as GRANT’S CONJECTURE. Panels that left V or the subject in place were redrawn with the default, or removed where they only repeated the same string and LF; the three panels that drew the verb word with no V raising were removed on Grant’s instruction (2026-11-13). Each panel now says what else has to move for the verb-final order. No Navajo form, gloss or judgment changed. See the 2026-11-13 note and the pass 7 report.
An earlier dated note on this page treats V→v→T and subject raising as Grant’s conjecture or a comparator to test against PF/template rivals. Since 2026-11-13 they are the working default in tensed clauses: VPISH frames where the arguments start so that syntax and semantics line up; it is not a word-order account. PF/template accounts are kept only as labelled contrast, and the open question is what else has to move. The earlier note is kept as written.
"a word is the changeover point where the rules of morphology no longer arrange elements and now the rules of syntax arrange words (as elements)" · "the syntax is right headed and the morpho is left headed" · "Navajo simply has the headedness parameter of words as different from the other levels. German modifies top level, Navajo bottom."
Page numbers. Hale 2003 is cited by the pages of the pre-publication manuscript we opened; the published chapter runs pp. 11–43. Example and note numbers are as in the manuscript. Convention (needed so the claim has content). "Head-initial" means a head comes before its complement (H⌢C). For morphology this follows Grant: the affix is the head and the stem or rest of the word is its complement. So a prefix is a head-initial affix and a suffix is a head-final one. If instead the stem is taken as head of the word, every prefixing language comes out "right-headed," and the contrast disappears. The Grok app made exactly that switch between its first and second answers.
2. Facts any formalization must cover SOURCED
2.1 Clause level
Fact
Source (opened)
Object–verb order: OV
WALS 83A, Navajo datapoint (Dryer; reference Young & Morgan 1987)
Adpositions: postpositions
WALS 85A, Navajo datapoint (Dryer; Young & Morgan 1987)
Inflection: strong prefixing
WALS 26A, Navajo datapoint (Dryer; Young & Morgan 1987)
Basic transitive order with two overt nominals: topic NP, then the other NP, then the verb, e.g. łį́į́ʼ tsé yiztał ‘The horse kicked the/a rock’
Willie & Jelinek 2000:273, ex. (61a)
Polar-question particle da’ is grammatical only clause-initially
Fernald, Chavez et al., Navajo Questions v3.2022, exx. (2), (3a–c)
Enclitic -ísh attaches after a word or phrase. It occurs once per question (8). On the first phrase it gives a neutral question; elsewhere it focuses the phrase it follows (10)–(11). More generally, focus and topic particles (-sh, lá, -shąʼ) are added "after the constituent (phrase) that is being emphasized"
Negation is a frame doo … da ("frame to negativize verbs and phrases"); cf. tʼáadoo … da in ex. (4) of the content-question page
same site, lexicon index and content questions (4)
-go: "verb and subordinating enclitic"; past -ę́ę at the right edge of a relative clause, [shizhéʼé tłʼóół yizhbizhę́ę] bee ‘with [the rope my father braided]’
Hale (2003:14–15, ex. (10)) gives the ten positions, after Young & Morgan (1987:37–38), using Rice's (2000) glosses. Position I is the leftmost, X the rightmost.
Disjunct vs conjunct: Hale 2003 n. 5, citing Li 1933:459–60. Hale (2003 n. 6) adds that many position-I elements are separate phonological words, so the unit is really a "verbal phonological phrase" (Tuttle's "greater verb"). Goad & Travis (2021:3), following Li 1946 and Kari 1976, take the template to be one syntactic word and one phonological domain.
(K2) Kibrik 2005, ex. (2), citing Young & Morgan 1987:628, 443
DERIVATION
INFLECTION
ROOT
ni-
hi-
di-
ghi-
s-
l-
tsił
TERMINATIVE-
SERIATIVE-
INCEPTIVE-
PROGR-
1SG.NOM-
TI-
move.sitting
DERIV.ASPECT
LEX.ASPECT
INFL.ASPECT
MODE
PERSON
VALENCY
ROOT
[nihideestsił] ‘I will move on the buttocks to a point’
Kibrik's own segmentation and labels. TI = transitivity indicator, i.e. the "classifier"; Kibrik prints it with arrow diacritics, omitted here. Kibrik notes that the inceptive and the TI "do not lend themselves easily to the derivation vs. inflection dichotomy". His point: inflection is closer to the root than derivation, and among inflectional categories the order outward from the root is "nearly mirror-image" to the common ROOT–aspect–tense–mood–person order (Kibrik 2005, §0; the PDF has no page numbers).
2.3 Facts about how the template is ordered
SOURCED Rice (2000:4, quoted in Goad & Travis 2021:2): "in Athapaskan languages morphemes of greater scope appear to the right of morphemes within their scope". The LINGUIST List review (Bird 2001) gives an example: subjects have scope over objects, so they appear to the right of objects.
SOURCED Iterative and distributive plural come in one order in Navajo and the reverse order in some other Athabaskan languages. Rice explains this: the iterative belongs to the event system and the plural to the argument system, so they enter no scope relation (Hale 2001:689; Bird 2001).
SOURCED Preverb and stem form one lexical unit (Rice 2000, Hale 2001:678), yet they sit at opposite edges of the word, "much in the manner of Dutch and German verb-particle arrangements in verb-second finite clauses" (Hale 2001:678). This is Grant's German parallel turned upside down, and it is a published one.
SOURCED Hale treats the stem itself as root plus a verbal host that carries aspectual inflection. In the future and progressive that host is the suffix -ł (Hale 2003 n. 10). So the word does have a small suffixal residue.
3. Where is the boundary?
HYPOTHESIS Grant's claim needs one changeover point. The sources give at least three candidates, nested one inside the next:
Candidate
Definition
Diagnostic
Source
B1 the "#" boundary
between the disjunct (I–III) and conjunct (IV–IX) sectors
conjunct phonology (only the neutral vowel underlyingly; coalescence) vs full disjunct phonology
Hale 2003 n. 5; Y&M's # as used in Hale 2003 ex. (15)
B2 the verb word
the extended projection of V (V, Voice, M/S, Qual) spelled into one bipartite skeleton, plus its proclitic adjuncts
the skeleton-filling domain; one prosodic word
Hale 2001:681–683; 2003:30–37; Goad & Travis 2021:3
B3 the Maximal Verb Complex
the verb plus any preceding postpositional phrases: "a noninterruptable constituent"
domain of the yi-/bi- voice alternation
Willie & Jelinek 2000:270, citing Willie 1991; page-checked 2026-10-11 in Willie 1991:49–50, (44): the “MVS (Maximal Verb-Sentence)”, “the domain of the expression of the subcategorizational features of the verb” (pleonastic & psych §7)
HYPOTHESISFormal version of the boundary claim. Let B range over {B1, B2, B3}. Write HD(H) ∈ {init, fin} for the direction in which head H takes its complement. Grant's claim is:
∃! B ∈ {B1,B2,B3} : ∀ functional heads H,
HD(H) = init ⟺ H is properly contained in B
A "word" is then defined as the B that satisfies this. The claim fails in two cases. (a) The direction switch falls inside one of the candidates (some heads inside B2 are final and others initial). (b) The switch matches no boundary that other diagnostics pick out.
Update 2026-10-20, 12:46pm (T41f). A new diagnostic for choosing among B1–B3 is on where the yes/no enclitic -ísh can go: can the yes/no enclitic -ísh follow the element? It follows the preverb ka (Hale 2003 n. 12), which sits in position I, left of “#”, and it follows a postposition word inside the MVS (NLA (6) Neésh hólǫ́?). No opened form puts it inside the skeleton word. So on the opened data the enclitic's boundary is B2, not B3. B1 vs B2 turns on whether a preverb that Hale prints inside the word (yisdá, (9)) can host it (T43c). This is a prediction (EN-1′), not a finding. Silence counts for nothing.
Update 2026-10-21, 4:46am (T43e). B2 vs B3 is sharpened on attached or separate postpositions. B2 (the verb word W) is a sealed spell-out unit; B3 (the MVS) is a syntactic domain without a seal: it is where the PPs are ordered before W and where the yi-/bi- choice is made (Willie p. 100: “the Inverse has scope over the entire MVS”). Writing a postposition onto the front of W is pronunciation after sealing (Hale 1997 PDF p. 31: “a phonological matter, not a syntactic one”), so procliticization is not a fourth boundary. The one postposition that goes inside W, the causative -y- (Hale 1997 PDF p. 53), is a case-realizing (functional) one, which fits a word edge defined by where functional material stops (HYPOTHESIS).
4. Five formalizations, side by side
Notation: H⌢C means H precedes C. "Hierarchy" is c-command order, highest first.
(i) Domain-indexed head parameter GRANT'S CONJECTURE → refined HYPOTHESIS
Statement (Grant, literal).
LIN(H, C) = H⌢C if H is inside B (rule set A: morphology)
LIN(H, C) = C⌢H if H is outside B (rule set B: syntax)
Immediate problem. The stem (lexical V) is inside B, yet its object comes first: the object prefix (IV) precedes the stem (X). On Hale's analysis the object pronoun is the complement of V (Hale 2003:11, tree (8)). So the literal version predicts V⌢Obj, and the surface shows Obj … V. Hale states the split himself: "the structure projected by the lexical category V is right-headed, while the structures projected by the higher nuclear elements (v, M, etc) is left-headed" (Hale 2003 n. 4). Lexical V takes its arguments on the left, "an arrangement which extends to postpositional and nominal (possessive) projections as well" (Hale 2003:32). SOURCED
Refined statement (i′)HYPOTHESIS
LIN(H, C) = H⌢C iff H ∈ F (functional) and H is inside B
LIN(H, C) = C⌢H iff H ∈ {V, P, N} (anywhere), or H ∈ F outside B
(i′) is Hale's footnote with Grant's word-domain index added. Hale himself attaches the head-initial setting to the extended projection of the verb: "selection (by a functional head of its complement) is left to right, in accordance with the parametric head-initial character of the extended projection of the verb" (Hale 2001:689). So (i′) and Hale differ only on functional heads outside the word: C-like particles and subordinators. Grant predicts these are final. Hale predicts they are initial, if C counts as part of V's extended projection.
Predicts. Inside B: Qual⌢M/S⌢Voice⌢V, which matches VI–VII/VIII–IX–X. Lexical heads are complement-final everywhere: Obj before V; NP before postposition, as in […] bee; possessor before noun. Outside B, functional heads are final.
Bearing data.ya’, -go and the relative-clause -ę́ę are final, which favours (i′). da’ is initial only, which favours Hale. -ísh and the focus/topic particles follow the constituent they mark, which favours (i′) if they are heads of focus/topic projections and is neutral if they are phrasal enclitics. OPEN Whether these particles are heads at all is undecided, and the NLA grammar does not assign them a category. If da’ is an operator in Spec,CP rather than C⁰, its initial position is neutral, because specifiers are initial under either head setting. If so, (i′) survives (see D6).
Falsified if. (1) A word-internal functional head follows its complement. The candidate is the aspectual suffix -ł on the stem (Hale 2003 n. 10), which (i′) can escape only by treating it as part of lexical V, as Hale does. (2) Clause-level functional heads turn out to be systematically initial.
Interaction with phenomenon 2.HYPOTHESIS If the Pronominal Argument Hypothesis is right, overt nominals are adjuncts (Hale 2003:5–7; Willie & Jelinek 2000). Then clause-level "OV" order is not a head–complement fact at all, and rule set B loses its most-cited evidence. Its direction would have to be read off clause-level functional particles, clausal complementation and postpositional phrases. Even in a postpositional phrase, the P-word itself carries a prefixed object pronoun (b-ee ‘3-with’), so a free NP before it may be an adjunct too.
(ii) Mirror Principle with prefixal exponence (Baker 1985)
Statement. Morphological derivations directly reflect syntactic derivations (Baker 1985, as summarized in Goad & Travis 2021:2). HYPOTHESIS With every affix a prefix, this means: if X is attached after Y, then X⌢Y. Read from left to right, the word lists the hierarchy from the top down.
Predicts with a standard hierarchy (Mood > Tense > Aspect > Voice > V; subject above object): Mode before aspectual and iterative material, and subject agreement before object agreement.
Template showsSOURCED Iterative (II) before Mode (VII), the reverse (template order, Hale 2003 ex. (10)). Object (IV) before 1st/2nd-person subject (VIII), the reverse. Derivational aspect is outermost and person/mode inner in (K2), the reverse (Kibrik 2005).
Verdict.SOURCED "Athapaskan verbal morphology appears to violate the Mirror Principle of Baker (1985) in multiple ways" (Goad & Travis 2021, abstract). To save it, one must either invert the hierarchy or add extra operations. Speas (1991b) starts from a head-final tree and adds iterative lowering with suffixation, then prefixation. Harley (2011) uses two head movements, Local Dislocation and morphological merger. Both are summarized in Goad & Travis 2021 §3.1–3.2; neither original was opened here.
Relation to Grant. Mirror plus prefixes treats the word as ordinary syntax read off the tree. It has no rule set A, so a failure of Mirror counts as evidence that something word-specific is going on.
(iii) Scope-based ordering (Rice 2000)
StatementSOURCED: if A has scope over B, then B⌢A, so wider scope sits further right (Rice 2000:4 via Goad & Travis; Bird 2001). Verbs are syntactic phrases and are head-final (Bird 2001; Goad & Travis 2021 n. 4). The stem raises: it right-adjoins to Voice, and Voice+V then right-adjoins to Mode/Subj. That is Hale's rendering of Rice in Hale 2001:679–680, trees (4)–(5). Where two morphemes have no scope relation, their order is free.
Predicts subject to the right of object ✓; Mode to the right of the iterative ✓; It/Pl order varying across languages ✓.
Falsified if a fixed order cannot be derived from scope, or a morpheme takes scope over material to its right. Hale's challenge: the Qualifier selects Mode/Subj, so Qual is higher than M/S even though it is to its left. "Consequently, the c-command and scope relations between these elements are opposite to those assigned to them by Rice" (Hale 2001:680).
Relation to Grant.HYPOTHESIS This is the main competitor. On Rice's account the word-internal syntax is head-final, the same direction as the clause. Nothing flips. The prefixal look comes from rightward stem movement plus scope-to-the-right. If Rice is right, rule set A and rule set B have the same headedness, and the word boundary is marked by verb raising, not by a parameter reset.
StatementSOURCED (Goad & Travis 2021:11, ex. (7); page numbers for this paper are those of the authors' preprint PDF, journal pp. 349–389). The hierarchy is head-initial underneath. Head movement starts from V and builds a complex head. At each step the new affix is prefixed "not to the left edge of the complex material to which it attaches, but rather, directly to the left edge of the root". Earlier affixes are pushed outward, a "tucking in" borrowed from Richards 1997/2001:
Consequence.HYPOTHESIS The last-attached (highest) affix ends up nearest the root. That is exactly the reverse of Mirror order: an "anti-mirror" produced by a single word-internal attachment rule. Goad & Travis identify position 6 (Qualifier) as v and position 9 as ExtArg. Position 9's position next to the root is put down to prosody, not syntax (2021:5, 12).
Predicts a right-branching bracketing [4 [5 [6 [7 [8 [6.5 [root]]]]]]] (their (31b)), and no hard prosodic-word edge between positions 8 and 9 (their §4.3).
Falsified if a hard PWd edge appears between 8 and 9 (that would favour their Morphological Merger alternative), or an order turns up that tuck-in cannot reverse.
Relation to Grant.HYPOTHESIS This is the closest published formal match to "rule set A". The hierarchy is the same in word and clause, but the word has its own linearization operation (attach at the root edge), and that operation does not apply to phrases. It is not a head-parameter value, though. Grant's "two syntax systems" becomes "one hierarchy, two attachment rules."
(v) Antisymmetry and roll-up (Kayne 1994; opened 2026-10-16, see A/B as spell-out §5)
Update 2026-10-16 (12:46pm, T27j). Kayne 1994 has now been opened. Applied to Hale 2003's hierarchy, it orders the four conjunct heads with no movement, but every prefix and preverb must raise above the Qualifier, and Hale's trees need extra silent heads (derivations N1′–N5′). Correction to the “Predicts” line below: a mirror-image order of stacked clause-final particles is predicted by a head-final account too, so it does not separate roll-up from (i′) (P4).
Statement. Asymmetric c-command maps onto precedence (the LCA). All heads are initial underneath, and head-final surface orders come from moving complements into specifiers ("roll-up"). Goad & Travis (2021:6) count a head-final structure as "a problem for those who espouse Kayne's antisymmetry hypothesis (Kayne 1994), as we do".
Applied to GrantHYPOTHESIS: the two rule sets become no roll-up inside the word (conjunct heads stay in underlying head-initial order, Qual M/S Voice V) and roll-up above the word (verb-final clause, postpositions). The object prefix before the conjunct heads needs a leftward specifier landing site. That is the Case-driven raising to Spec,M that Hale (2003:12) considered and rejected.
Predicts that if several clause-final particles co-occur, roll-up puts them in the mirror image of their hierarchy. OPEN Co-occurrence orders of -go, -ę́ę, ya’, lá, -sh are needed.
Falsified if the clause-final particles do not come out in reverse hierarchy order, or the object prefix shows no sign of a derived position.
5. Comparison
Real head-direction flip at the word?
Conjunct order Qual–M/S–Voice–V derived by
Object prefix left of M/S by
Main falsifier
(i′) Grant/Hale domain parameter
Yes, for functional heads
head-initial selection
phonological skeleton-filling (Hale 2003:33–36) or raising
a selector to the left with wide scope (Hale on Qual)
(iv) IRP
No parameter flip; distinct word-level attachment rule
tuck-in at root edge
lower attachment = farther from root
hard PWd edge between 8 and 9
(v) Kayne roll-up
Yes as ±roll-up
underlying order, no movement
raising to a left Spec
clause-final particle order not mirrored
The surface string does not settle head direction.HYPOTHESIS One string, Qual–M/S–Voice–V, fits two opposite hierarchies. It fits Qual > M/S > Voice > V with head-initial selection (Hale). It also fits M/S over Qual with head-final heads and rightward stem raising (Rice). The template alone cannot decide whether Navajo morphology is left-headed. What decides it is scope and selection inside the word, the diagnostics asked for in Grant's own open question 6 (reconstruction, binding, scope inside the verbal complex). The test needs, for example, judgments on whether a Qualifier-position meaning such as inceptive or semelfactive scopes over the mode/aspect meaning or under it.
6. Typology: Athabaskan against the suffixing tendency
SOURCED Greenberg 1963, Universal 27: "If a language is exclusively suffixing, it is postpositional; if it is exclusively prefixing, it is prepositional." Universal 28: "If both the derivation and inflection follow the root, or they both precede the root, the derivation is always between the root and the inflection." (Text from the course reprint opened this run.)
SOURCED Navajo is strongly prefixing, OV and postpositional (WALS 26A, 83A, 85A).
SOURCED (secondhand) Kibrik (2005, §3.1, problem (e)) lists as a known Athabaskan problem "the problem of Athabaskan languages being an exceptional example of consistently prefixing languages among the verb-final languages". He cites Bybee, Pagliuca & Perkins 1990: verb-final languages are typically suffixing, and in their sample Slave "turned out to be the only genuine exception to this generalization (1990:5-11)". Bybee et al. 1990 itself was not opened. Kibrik's reference list gives a different 1990 Bybee et al. title, "Back to the future". The pp. 5–11 citation fits "On the asymmetries in the affixation of grammatical material" (Croft, Kemmer & Denning eds., pp. 1–42), but that match has not been checked.
HYPOTHESISKeep the claims precise. Navajo is not exclusively prefixing (the stem has aspectual suffixal material, and there are enclitics), so it does not break the letter of U27. What it goes against is the statistical OV→suffixing tendency (Bybee et al.). On Kibrik's derivation/inflection classification, it also violates U28: both kinds of affix precede the root, but inflection sits between the root and derivation.
7. Strongest formalization reached in v0
Superseded in part 2026-10-16 (12:46pm, T27j): (i′) is restated as spell-out on A/B as spell-out. There, one ordering rule applied in two cycles (CYC: non-heads first, then heads, each highest-first; the finished word counts as one non-head in the clause) derives both the lexical-head-final pattern and the head-final look of the clause. The word becomes a spell-out cycle, not a parameter boundary. (i′) is kept below for the record.
HYPOTHESIS(i′) with B = B2. Inside the verb word (the extended projection of V spelled into Hale's bipartite skeleton), functional heads precede their complements. Lexical heads (V, P, N) follow their complements everywhere. Outside the word, functional heads are final. Non-heads inside the word (object pronoun, preverbs) are placed by skeleton-filling and by leftward-scope adjunction, not by the head parameter.
This statement fits every template fact listed above. It reduces Grant's conjecture to one testable difference from Hale (2001, 2003): the position of clause-level functional heads. In the sources opened, those particles split (da’ initial; ya’, -go, -ę́ę final). The serious rival is Rice's uniformly head-final account, which has no flip at all. The deciding evidence is scope inside the word, which v0 does not have.
8. Where does the order change? A head-by-head audit
Plain words — HYPOTHESIS / verdict. There is no demonstrated single point where Navajo changes from “left-headed” to “right-headed.” Some verb-building elements precede the stem, while the verb’s object precedes the stem too. Those are different relations inside the same word. Outside it, question marking can be initial, final, or attached to an earlier phrase, and negation has two edges. A limited Hale-style account survives: some nuclear heads select rightward; lexical V, P and N select leftward. That is a category-sensitive hypothesis, not an established word-edge switch.
8.1 Dated correction: what the earlier wording overstated
2026-11-11 — SOURCED / HYPOTHESIS. The older sections remain above as a record. Their “fits every template fact” does not establish the universal in (i′), and their B2 combines a skeleton, proclitics and a phonological word too freely. Hale explicitly allows separate-word preverbs (2003 ms.:64 n.6) and leaves the stopping point of word assembly unresolved (68–69 n.12). Nor does enclitic hosting identify a selecting head’s syntactic position. The earlier description of causative P as functional is not automatically a success for (i′): that P’s object is to its left inside the verb (Hale 1997 PDF:53, (40–41)).
2026-11-11 — SOURCED correction to §4(iii). “No scope relation” permits order variation; it does not require free order everywhere. Bird’s 2001 review expressly allows phonological, morphological and syntactic requirements to override scope. A fixed unexplained order alone therefore does not falsify Rice. Also, Hale’s “extended projection of V” should not be extrapolated into a demonstrated prediction about every C: the opened discussion establishes the nuclear verbal fragment, not a complete CP analysis (Hale 2001:681, 689).
OPEN / source limit. The old stem-suffix transcription is not independently re-certified here: the opened manuscript uses a legacy glyph. Its root-plus-suffix analysis is clear (Hale 2003 ms.:66–67 n.10); the glyph itself is not newly indexed or used as a Navajo leaf. The exact preverb-plus-question forms in n.12 also have damaged font encoding. They are not reconstructed.
8.2 What counts as a head? Fix the parameter before testing
HYPOTHESIS — operational definitions. A syntactic selecting head H is a category-bearing item which selects a sister phrase X; X, not any nearby word, is its complement. A theta-bearing argument is not thereby a functional head; neither is an adjunct or an agreement exponent. At the word level we must choose h:
h=A, affix-as-head: an affix selects the remaining base. A prefix then has initial head–base order by definition.
h=S, stem-as-word-head: the stem supplies the word’s category; modifiers/affixes precede it. The same prefixed word now has a final morphological head. This alone says nothing about a syntactic head’s complement.
h=C, selecting-head analysis: assign heads and complements independently. Hale’s nuclear heads, arguments and adjuncts are one such analysis; an agreement-head analysis is a rival. A pronounced prefix need not be the position of a selecting head.
HYPOTHESIS — strict surface measure. Let e(Y) be the set of overt token/segment positions realizing Y, excluding H from its complement X. Then:
HDsurf(H,X) = init iff max e(H) < min e(X)
= fin iff max e(X) < min e(H)
= split if both sets are known but interleaved
= OPEN if their realization or selection is unknown
= NA if no selecting H–X relation is posited.
HYPOTHESIS. A fused portmanteau does not give separate positions to its proposed subheads. Compare this strict measure with HDhh, the weaker order of H and the head of X, and with HDbase, underlying direction in a proposed derivation. They are not interchangeable. In NX-094 the object precedes Qual and M/S, while the stem follows them. If the object belongs to their full selected projection, their surface order to that whole complement is split, even though Qual–M/S–Voice–V is initial head-to-head order. Skeleton filling or movement can reconcile base and surface, but each is an extra hypothesis.
NONE / axioms. Only VPISH and theta roles are assumed here. Head movement, Mirror, light d, abstract agreement heads, phases and skeleton filling are not axioms. The category and selection definitions above are explicit modeling choices. No conclusion about an independently real syntactic head follows just from a hyphen.
8.3 Source examples: the checks that change the answer
SOURCED. These additions are indexed as NX-107–118. Navajo strings and source glosses are reproduced without grammatical repair; glottal stops are normalized to U+02BC. Source glosses are not editorial Leipzig analyses. Missing glosses stay missing. The full ten-position mapping of NX-107 is Hale’s (10), not a newly inferred segmentation. In NX-108 the source prints the final verb vowels without ogoneks; that spelling is retained.
NX-107 — SOURCED · Hale, Ken 2003, ms. p. 14. Yisdánídashizhdoołtééł.
“They (3a = 4th person) will get me back to safety.”
NX-108 — SOURCED · Willie, MaryAnn 1991, p. 184. Da-nihi-má doo kóó kééhatʼíí da.
Source gloss, unchanged: PL-1-mother NEG here PL-3-live NEG.
“Our mother does not live here.”
NX-109 — SOURCED · Willie, MaryAnn 1991, p. 171. doo
Source gloss, unchanged: FUT.
NONE: no source free-translation line assigned to this fragment.
NX-117 — SOURCED · Hale, Ken 2003, ms. p. 26. Jooł naamaas.
Source segmentation: jooł na#Ø-Ø-maas.
“The ball rolls around.”
NX-118 — SOURCED · Hale, Ken 2003, ms. p. 26. Mósí jooł neiłmaas.
Source segmentation: mósí jooł na#y-Ø-ł-maas.
“The cat rolls the ball around.”
SOURCED / limit on theta and scope. Hale says explicitly that in NX-118 the ball rolls around; the cat need not (2003 ms.:26, discussion of (20)). This supports a theme-oriented interpretation despite the preverb’s left-edge pronunciation. OPEN: it does not by itself establish a Qual–Mode scope order, a direction parameter, or that this preverb changes the argument count. NX-111 has a causee; the claim that its P incorporates is the author’s analysis, not a consequence of the translation alone.
8.4 Inventory: surface facts, selection hypotheses, and domains
HYPOTHESIS — domain conventions. B1 contains the conjunct sector and stem (IV–X); I–III are outside. B2 is an occurrence-specific verb word W, not a definition that guarantees the result. B3 is Willie’s verb plus its selected PPs, the MVS (1991:50); it is not the entire question or negative clause. I/O mean inside/outside, in the order B1 / B2 / B3. A noun’s own word is marked IN, but it is outside the verb word (OV). A clitic’s host is not its independently diagnosed structural site. OPEN is not a pass.
Candidate head
Checked source
Surface order
Head/complement and direction
B1 / B2 / B3
Disjunct / preverb (I)
NX-094: Hale 2001:678–679; NX-107: Hale 2003 ms.:14 (9–10); NX-117–118:26 (20).
SOURCED: preverb … object/other prefixes … stem. In (20b), the ball, not the cat, rolls around.
HYPOTHESIS: adjunct in Hale, not a nuclear selecting head. Affix-as-head is a different parameter. HD = NA in Hale’s model.
SOURCED: II precedes III, M/S and stem in the source template. REV and ITR glosses are not silently equated.
HYPOTHESIS: operator head versus adjunct. If a prefix-head taking the following base, head–base order is init; its selected complement and scope still require a model.
NX-107, Hale 2003 ms.:14 (10); nominal plural in NX-108, Willie 1991:184 (3).
SOURCED: verbal plural before object and stem; nominal PL before possessor and N. Willie says the nominal plural scopes over the possessor (183–184).
HYPOTHESIS: treating PL as a functional selecting head makes it a conditional B1 counterexample. Do not infer headhood merely from its prefix position.
Verbal: O / I / I. Nominal: O / O_V (I_N) / O
Object / object agreement (IV)
NX-094, Hale 2001:679 (3), 681–682; NX-107, Hale 2003 ms.:14 (10).
SOURCED: sh occurs before Qual, M/S, Voice and stem.
HYPOTHESIS: in Hale it is the object argument, not an agreement head. In an Agree-head model, its location need not be its pronounced exponent’s location. HD = NA versus OPEN, respectively.
I / I / I
Deictic subject (V)
NX-107, Hale 2003 ms.:14 (10); Hale 2001:689–692 (22–24).
SOURCED: deictic before Qual; ordinary subject in M/S.
SOURCED as author analysis: Hale treats deictics as adjunct nominals linked to third-person M/S, not another nuclear subject head. HYPOTHESIS: a distinct functional head is a rival.
I / I / I for the exponent; head site OPEN
Qualifier (VI)
NX-094, Hale 2001:679–681; NX-107, Hale 2003 ms.:14, 32.
SOURCED: Qual before M/S and stem, but after object/deictic exponents.
HYPOTHESIS: Qual selects M/S in Hale. Head-to-head order init; full-complement surface order is split if that complement includes the earlier object. The source’s syntactic hierarchy is not itself an axiom.
I / I / I
Thematic material / lexical theme
NX-094: d relates to arms/legs (Hale 2001:679); NX-089–093, Speas 1986:205–206.
SOURCED: theme pieces may be discontinuous across inflection.
OPEN: a thematic label is not another independently proved selecting head; do not count the same Qualifier twice. Idiomatic dependence does not determine direction.
I for the conjunct piece; disjunct pieces O / I or OPEN / I
Aspect / Mode (VII)
NX-094, Hale 2001:679, 682; NX-107, Hale 2003 ms.:14 (10).
SOURCED: M/S precedes Voice/stem; mode, aspect and person can be realized together.
HYPOTHESIS: M/S selects vP in Hale. Head-to-head init, strict full-complement order can be split. OPEN: direction of independently unpronounced subheads in a portmanteau.
SOURCED: first-person marking before the stem, after outer prefixes; source analyses place person and mode together or in an AStem.
OPEN: separate agreement head, argument pronounced on M/S, or portmanteau. Prefix order cannot choose among them. No independent Subj-head HD is scored.
I / I / I for the exponent
Classifier / Voice (IX)
NX-094, Hale 2001:679–682; NX-098–099, McDonough 2019:40 (8).
SOURCED: Voice before the stem; McDonough gives different valence analyses for the same surface form.
HYPOTHESIS: Voice selects VP in Hale, head-to-head init. A classifier is not invariably an independently compositional causative. Goad & Travis 2021 preprint:4–5 propose a different articulated-v mapping.
SOURCED: object exponent before stem; source also describes a root + suffixal verbal host.
HYPOTHESIS: V takes the object on the left in Hale; the R+V host is lexical V for him, but a functional categorizer/TAM head for a rival. The latter makes a conditional internal-final counterexample.
SOURCED: causee and b- precede P; overt d intervenes before y inside the verb.
SOURCED as author analysis: case-realizing P incorporates into E. HYPOTHESIS: if case-realizing P belongs to F, its final relation to its object counters all three candidates; if all P is lexical by stipulation, it is exempt. Agreement is not its selected complement.
SOURCED: object/person precedes P, PP precedes the verb in these examples. Willie explicitly distinguishes grammatical P with pronominal objects from lexical P.
HYPOTHESIS: selected object–P is fin in Hale’s analysis. Calling every P lexical is a category convention, not proof that grammatical P is outside the functional class.
O / O / I for selected PPs; non-selected adjuncts OPEN
N and possessor
NX-112, Willie 1991:183 (2a); NX-108:184; Hale 2003 ms.:32.
SOURCED: shi-má, gloss 1s-mother; plural precedes possessor and N.
HYPOTHESIS: N selects a left possessor in Hale. Stem-as-word-head calls this final; affix-as-head calls the possessive prefix initial. Same string, different head choice.
SOURCED: Díí precedes ʼashkii in NX-035. Willie allows a one-phrase reading and an independent-NP reading.
OPEN: D taking NP, modifier, or independent pronoun. If D selects following NP, this is an initial clause-side functional head; if modifier, no HD test. NONE: no light d is required by these data.
SOURCED: Daʼ before the statement; the other three placements are starred by the source.
OPEN: Q/C head versus operator phrase. Under the head analysis it is an initial functional counterexample outside B1/B2/B3. Operator status may not be chosen solely to rescue a final-head rule.
SOURCED: =sh follows DET in the wh-phrase; -ísh can follow the first P word, not only the whole clause. NLA distinguishes the two enclitics.
OPEN: right-edge exponence is not proof of head-final C. Neésh marks a host inside a PP/MVS but may realize a higher Q. Host position and head position must be separated.
NX-110, Willie 1991:133 (3); NX-015–016, Hale & Platero 1974:11 (6–7).
SOURCED: =go at the right edge of the subordinate predicate/clause.
HYPOTHESIS: if it is C selecting that clause, fin. If a phrasal enclitic, host-final alone does not locate the head. Not inferred to be internal merely because it is written onto a verb.
SOURCED: ending follows the relevant wh-material or relative predicate.
OPEN: D, nominalizer, C, or context-dependent categories. The same spelling does not prove one universal head. For an overt selected phrase preceding such a head, order is fin.
O / O / O for a head closing a containing phrase; PF verb/wh-word edge
SOURCED: doo … predicate … da; the subject precedes doo.
HYPOTHESIS: one Neg with two exponents gives split HD; two Neg heads give one initial and one final. OPEN: their selected complements and syntactic sites. Neither option is a uniform final surface pattern.
Clause sites outside B1/B2; relation to MVS edge OPEN
Free tense/future expressions
NX-109, Willie 1991:171 (42), final doo gloss FUT; NX-110:133 (3), final dooleeł gloss will; Hale 2001:681 n.3.
SOURCED: final in these contexts; future doo is not negative doo. Hale relates the future auxiliary to a fuller verbal form.
OPEN: T head, auxiliary V, or other category; a future translation does not assign T. NONE: no verified initial T-particle/complement pair added here.
Outside the preceding verb word; own auxiliary word possible; B3 inclusion OPEN
SOURCED: wh-words/phrases can precede V and need not be clause-initial; the enclitic may be on another word.
OPEN: heads of nominal/adverbial phrases, arguments, modifiers or specifiers in a clause. Being interrogative does not make a word a C head. HD to a clausal complement = NA without such selection.
O / O / O for these clause-level expressions; internal phrase structure OPEN
SOURCED. The interrupted preverb–stem dependency is especially clear in NX-094 (Hale 2001:678–679): preverb, object, Qual, M/S, Voice, stem. Prefixes can intervene between lexical partners; this is not evidence that a preverb freely moves around an English-style object. OPEN: the exact phonological status of every preverb token and which boundary independently blocks which operation.
8.5 Run the boundary test without changing the definitions
GRANT’S CONJECTURE — formal target. Choose headhood h, functional class F, complement relation X, direction measure m, and independently diagnosed membership IB first. Then test:
U(h,F,X,m): ∃! B ∈ {B1,B2,B3} ∀ H ∈ F:
HDm(H,X(H)) = init ⟺ IB(H).
Domain requirement: every tested HDm must be init or fin;
split / fused / OPEN cannot be scored as fin.
HYPOTHESIS — conditional finite tests. These are tests of explicit analyses of the checked fragment, not counts of all Navajo heads. The underlying inventory and conditional cases are also in the site’s source data, headedness-audit.json. Unknown cells are not silently completed.
Fixed analysis / measure
B1
B2
B3
What follows
Only Hale nuclear Qual, M/S, Voice; HDhh
Fits fragment
Fits fragment
Fits fragment
All three contain these heads. No unique boundary is identified.
Add iterative and distributive as functional prefix-heads over following base
Counterexample: initial heads outside # (NX-107)
No new contradiction in this fragment
No new contradiction in this fragment
Conditional on treating Hale’s adjuncts as selecting heads; B2 versus B3 still unresolved.
Add the stem’s suffixal V as a functional categorizer/TAM head
Final inside
Final inside
Final inside
No candidate survives this classification. Hale instead classes it as lexical V (ms.:66–67).
Count the incorporated case-realizing P as functional; compare P with its causee complement
Final inside
Final inside
Final inside
NX-111 is a conditional counterexample to all three; treating all P as lexical exempts it, at an explicit taxonomic cost.
Count initial daʼ as Q/C selecting the following clause
Initial outside
Initial outside
Initial outside
NX-114 defeats all three under this head analysis. Its headhood remains OPEN.
Count a selected external grammatical P as functional
Final outside: fits
Final outside: fits
Final inside: counterexample
Willie’s grammatical/lexical distinction matters. The PP inside B3 is not inside W.
Locate Q by the host of -ísh in NX-115
Would be final outside
Would be final outside
Would be final inside
That shortcut disfavors B3, but it is NOT an established head-site test. Host placement cannot determine U.
Strict HDsurf with the full complement in Hale’s argument model
Qual/M/S/Voice can be split: object before head, selected stem after head. Binary input condition fails.
No head-initial base structure is disproved by this; instead the surface universal is the wrong measurement for that claim.
One Neg head with doo…da
Split exponence, not a binary direction. Two-head analysis gives an initial and a final Neg instead.
Neither analysis licenses “every clause-side exponent is final.” Head sites/complements remain OPEN.
HYPOTHESIS / verdict on U. Under h=A much of “initial morphology” is true by definition; under h=S the same word is stem-final. Under h=C the nuclear fragment fails to identify a unique B. Broader independently justified F choices can falsify particular candidates or all of them. Thus U is not established; its truth cannot be inferred from the template. The original uniform word-inside/word-outside statement fails on its own Hale-style object–V relation inside W. The refined functional-only version is a conditional research hypothesis, not a demonstrated law.
Compatible with Hale’s proposed base selection and the head-to-head template. Extending “functional” to final =go / DET / relative heads or grammatical P is not automatic; if those are F and select left material, the global version fails. Strict surface complements still split.
Rice’s Scope Hypothesis
Within its scope-governed domain, wider scope is realized farther right; independent constraints may override it. Hale reports a head-final + stem-raising implementation.
Fits object-before-subject and some adjunct ordering. Hale assigns Qual scope over M/S despite its left position; that is a rival selection/scope analysis, not an independently settled falsification. NX-117–118 gives a theme reading but not the missing Qual/Mode minimal contrast. No full-book verification claimed.
Pure cyclic linearization / phase account
One proposed cycle spells nonheads before heads and heads highest-first; the completed W becomes one unit in the next cycle (existing CYC proposal).
Can make object-before-V and Qual–M/S–Voice order compatible without reversing base direction at W. Phase edge, hierarchy and head/nonhead division need independent evidence. Initial daʼ and split negation cannot be declared nonheads merely to save the rule. No unique phase proved.
HYPOTHESIS — our redrawn vP/V-to-T comparison. Both panels use the working default, V→v→T with the subject raised to Spec,TP. A tests initial functional T/v against final lexical V; with T⁰ initial, the verb-final order needs the remnant VP to move above T (HYPOTHESIS). B tests uniform-final base structure with the same V→v→T chain, where nothing else moves. Specifiers are left in both. These are not reconstructions of Hale or Rice: their data motivate the question, not its answer. Agreement/qualifier/mode slots are not silently equated with T or v. The word box belongs to a real complex head; there is no separate PF-output daughter. Each panel states its LF. OPEN: which category and boundary assignments survive independent tests.
Analysis A: Working default (V→v→T) inside the mixed-selection changeover hypothesis Hypothesis
Analysis B: Uniform-final rival with the same working-default V→v→T chain Hypothesis
(2) Our changeover test: initial functional heads or uniform-final structure? HypothesisExact word chʼíshidiníłdązh and its cited pieces: NX-094, Hale 2001:678–679(1–3). The two drawings are our vP/V-to-T derivations, not Hale/Rice facsimiles; previous authors’ ordering proposals are historical contrasts only (Hale 2001:679–682; 2003 manuscript p.32/n.4). Qualifier, agreement and mode slots are deliberately not relabeled as proven syntactic categories: their correspondence with T/v/AdvP is OPEN. The dashed box marks a complex head, not the boundary theorem B1/B2/B3. The LF is an event-level abstraction of the directional reading, not a telicity result.Bracket notation (copy)
DP (id S-high; landing site of Move from subject): agent a
T′
T⁰ (head; id T; hosts the “V+v+T” (V + v + T))
vP
DP (id subject; struck copy; moved to S-high): agent a
v′
v⁰ (head; id v; head-moves to T)
VP
AdvP: chʼí
VP (id lowVP)
V′
DP (id object): theme t
V⁰ (head; id V; head-moves to v): dązh
Analysis B
TP
DP (id S-high; landing site of Move from subject): agent a
T′
vP
DP (id subject; struck copy; moved to S-high): agent a
v′
VP
AdvP: chʼí
VP (id lowVP)
V′
DP (id object): theme t
V⁰ (head; id V; head-moves to v): dązh
v⁰ (head; id v; head-moves to T)
T⁰ (head; id T; hosts the “V+v+T” (V + v + T))
8.7 The most precise statement that survives
SOURCED — bounded descriptive result. For the source-mapped tokens NX-094 and NX-107, the published positional annotation supports the following partial order, not a language-wide head parameter:
Qual ≺ M/S ≺ Voice ≺ Stem; Object exponent ≺ Stem.
Checked P relations: selected object/person ≺ P.
Checked possessive relation (NX-112): possessor exponent ≺ N.
HYPOTHESIS — structural refinement, exact domain. In Hale’s nuclear verbal fragment, assume Qual → M/S → Voice → VP with rightward selection, and V → object with leftward selection; extend leftward lexical selection to the checked P and possessive N relations. This category-sensitive mixed structure is compatible with the data. Its pronunciation needs a separately tested realization mapping. It does not entail ∀F outside W: final, a unique boundary, or uniform head-final clauses.
OPEN / counterexamples absorbed, not erased. Initial daʼ, initial Díí and the left edge of doo…da defeat blanket final surface order outside W. They are conditional head-direction counterexamples until their selecting status and complements are fixed. Final clitics do not prove the opposite universal. Internal final lexical V/P/N is explicitly admitted by the limited account; a functional analysis of incorporated P or the stem suffix instead defeats its functional-initial generalization. The limited account does not claim those classifications have been resolved.
8.8 OPEN: what would decide more
OPEN: independent diagnostics for Q/C, D, Neg and T headhood, their selected phrases, and the distinction between a clitic’s host and its scope site. In particular, a future auxiliary is not automatically T.
OPEN: a source-verified Qual/Mode scope contrast with interpretation independently established, rather than read back from template order; NX-117–118 is not that contrast.
OPEN: verb-word, skeleton and phonological-word boundaries on the same tokens; many preverbs are not phonological parts of W. A nominal word’s behavior cannot silently redefine verb-specific B2.
OPEN: whether position II/III operators are selecting heads or adjuncts; whether every P is excluded from F despite its grammatical/Case role; whether the stem suffix is lexical V or a functional categorizer.
OPEN: separate realization of Mode and subject in portmanteaux, zero heads, co-occurrence/selection of final particles, and behavior under question, negation and passive in matched prefix–stem pairs. No missing cell is supplied by an invented judgment.
NONE: no proof of a unique B1/B2/B3, no claim that suffixing or prefixing by itself establishes base syntax, and no use of Mirror, head movement, light d or a phase edge as an axiom.
Reading record and source limits
SOURCED: Hale 2001, “Navajo Verb Stem Position and the Bipartite Structure of the Navajo Conjunct Sector,” Linguistic Inquiry 32:678–693: checked pp.678–682, 689–691. Hale 2003, The Significance of the Pronominal Argument Hypothesis: manuscript pages 14, 26, 32, 63–69, not the published chapter pagination (11–43). Young & Morgan’s template is mediated by Hale here, not claimed as a new direct reading of their book.
SOURCED: Hale 1997 workshop notes, Remarks on the Syntax of the Navajo Verb, PDF p.53 (40–41); Willie 1991, Navajo Pronouns and Obviation, pp.50, 74, 133, 171, 183–184, 211–212. See index entries for verification page images and source-specific spelling.
SOURCED: Fernald et al., Navajo Questions: Yes/No, v3.2022, §§1–4, especially (2–3), (6), (15). This is unpaginated HTML, checked as text, not assigned invented PDF pages.
SOURCED, secondary access to Rice: Goad & Travis 2021, “Phonological evidence for morpho-syntactic structure in Athapaskan,” author-preprint pp.2–5 (published The Linguistic Review 38:349–389), quoting Rice 2000:4; Bird 2001 review, unpaginated; Hale’s explicit rival derivations. The freely readable publisher summaries of Rice’s chapter 6 and chapter 15 were consulted in the preceding comparison; the complete book was not read. No paywalled text is represented as opened.
SOURCED, carried forward with page-verified index: Speas 1986:205–206 (NX-089–093); McDonough 2019:28, 40 (NX-095–099); Chee & Yazzie 2025:22, 34–35 (NX-100–106). Their analyses are not merged into one undisputed head taxonomy.
9. Questions, negation and movement: a harder test · 2026-11-11
SOURCEDThe new evidence settles one thing, but not everything: the negative frame can cover a clause, and changing its right edge changes the meaning. It is not merely a box around a verb word. Questions also show a real inverse–clitic interaction. Neither fact, however, demonstrates that the verb rises to T or C, or picks one unique headedness boundary.
HYPOTHESISWorking verdict. Start from Spec,vP plus V→v→T and subject raising, the working default in tensed clauses since 2026-11-13, and ask what else has to move to match the word-order facts and scope constraints. Remnant-VP movement, PF placement of prefixes and VP pied-piping are hypotheses about that “what else”, and each must preserve the reported roles and scope. A model that gets the string right but the meaning wrong fails. The candidate boundary remains OPEN, not disproved merely by the absence of an example.
HYPOTHESISFirst-principles regime. The only stipulated starting points are VPISH and theta roles: the role-bearing subject argument first merges in Spec,vP. VPISH frames where the arguments start, so that syntax and semantics line up; it is not a word-order account. Since 2026-11-13, V→v→T and subject raising Spec,vP→Spec,TP are the working default in tensed clauses (Grant: near-universal, to be assumed). An EPP/Case trigger, reconstruction, a phase edge, a Mirror Principle, an identity such as T = Mode/Subj, and any movement beyond the default are additional hypotheses. If overt nominals are adjuncts to pronominal arguments, as in the Pronominal Argument Hypothesis, raising an overt noun is not automatically raising the theta-bearing subject.
HYPOTHESISHeadhood parameter. A selecting head H is distinct from its pronounced exponent and from an enclitic’s host. HDbase(H) describes the head’s order with its selected complement before displacement; HDsurface describes its exponent’s observed order. They need not agree. Affix-as-head versus stem-as-word-head remains a parameter, not a silent change of definition. A clause-typing function is not itself proof that its exponent is C⁰. We apply the same tests to initial and final elements.
HYPOTHESISLF criterion. Every derivation below ends in an explicit logical form. At the interpreted level, an operator must c-command the expression it takes scope over; reconstruction is an extra, stated operation, not an excuse to ignore a mismatch. e ranges over eventualities, including states; ASK_POL(p) asks whether the distinguished prejacent p holds; ASK_x asks for values of x. LFs abstract from tense/aspect details unless relevant. The source translations are SOURCED; the role labels and LFs are HYPOTHESIS.
HYPOTHESIS A bare polar alternative set {p,¬p} equals {¬p,p}. It cannot by itself distinguish the scope of Q and Neg. Here negative-question wording and answer behavior motivate ASK_POL(¬p), not an independently established syntactic Q site. Ordinary propositional ¬ cannot apply directly to a question denotation: an alleged Neg>Q reading needs an extra, independently supported interpretation.
SOURCEDFidelity. New quotations are indexed as NX-119–202. The NLA HTML has numbered examples, not pages. The Perkins–Fernald PDFs are working drafts; chapter-local printed pages are cited because their pagination is inconsistent across versions. Source CI/NI/P/F, missing glosses, doubtful judgments and spelling warnings remain visible. Editorial Leipzig conventions use NEG, Q, SG, OBJ, SBJ and PFV only in separately labeled lines. They do not replace a source gloss or invent a Navajo decomposition.
9.2 The requested 2×2×2 grid
SOURCEDSix broad cells have examples; two do not. “Active/direct” includes an intransitive WHY question in the negative WH cell: it is not a matched transitive yi-/bi- pair. Restrict the grid to matched transitive predicates and that cell remains OPEN too. The inverse column is filled with bi- inverse, not silently relabeled as an argument-reducing passive. Different predicates across rows cannot establish a movement effect.
HYPOTHESIS The logical forms and role mappings are printed under each derivation in §9.6. +A means affirmative; −A means negative. Empty cells are not ungrammaticality judgments.
Keep the passive families apart
SOURCED Willie’s agentive passive has NX-020, NX-021 (1991:37, (21),(23)); her impersonal and simple-passive discussion includes NX-022, NX-023, NX-024, NX-054, NX-055, NX-056 (pp.38,57,99). Hale 2001:690 calls the agentive passive an impersonal subject construction rather than a true argument-reducing passive. The terminology is part of the analysis, not a reason to discard the forms.
OPEN; a passive English translation alone was not promoted
Passive inside a larger polar question
NLA Ch.8 p.187(74), gloss 2-tell.P.Pass
Present in a subordinate cause clause; main predicate is know
No matched negative/inverse version verified
OPEN The difficult full form in Ch.8 (74) is withheld from the index pending a complete diacritic check; its source’s passive gloss is recorded as a research lead, not a filled matrix-passive cell. A paid/got-paid English translation in Ch.6 p.87(28) likewise does not establish passive morphology.
9.3 “Not the event” is different from “not Mary”
SOURCED The draft explicitly distinguishes constituent correction with hanii from the negative frame (Ch.6 pp.92–95). It negates a selected focus while the rest of the assertion is taken for granted. In these examples gaʼ follows the correcting phrase.
Source translation: “It was not the goat that s/he roped, it was the sheep.”
HYPOTHESISConstituent-focus LF: roping occurred; the identity of its theme is corrected: ∃e∃x[rope(e) ∧ Agent(e,a) ∧ Theme(e,x) ∧ ¬(x=goat) ∧ x=sheep]. This is not ¬∃e[rope(e) …]. No V-crosses-sentential-Neg conclusion follows.
SOURCEDNX-168, NX-169, NX-170, NX-171, NX-172, NX-173 (Ch.6 p.85(17a–f)) put doo…da around nonverbal as well as verbal predicates. “It is not gold” does not show that an object NP in an affirmative event clause has been negated. There is no verified doo-[object NP]-da versus doo-[event]-da minimal pair with the same verb and contrasting argument/event scope here. That narrower contrast is OPEN.
The decisive scope pair: change only the right edge
Source gloss:John Mary neg car 3-2- Fut-buy 3-of-3-expect.Imp neg
Source translation: “John doesn’t expect Mary to buy a car.”
HYPOTHESIS Let pM = ∃e∃x[car(x) ∧ buy(e) ∧ Agent(e,Mary) ∧ Theme(e,x)], interpreted as the anticipated buying, not an asserted actual-world event. Then:
NX-180, lower Neg:∃s[expect(s) ∧ Exp(s,John) ∧ Content(s,⟦¬pM⟧)]. The lower Neg must c-command lower event closure, not matrix expect.
NX-181, matrix Neg:¬∃s[expect(s) ∧ Exp(s,John) ∧ Content(s,⟦pM⟧)]. Matrix Neg must c-command the expectation state. This does not entail that Mary will not buy a car.
HYPOTHESIS Content brackets are proposition-valued, intensional arguments; treating them as ordinary actual-world conjunctions would itself be an LF error. A fixed low doo-head interpretation that gives both strings lower negation fails NX-181. Two viable repairs remain: place the scope-bearing right-edge Neg at the appropriate clause edge and treat doo as concord; or give an initial Neg the appropriate higher scope and independently derive the preceding nominal order. The latter needs more than the location of doo alone. Neither repair tells us B1, B2 or B3.
SOURCED Other placement facts strengthen, but also limit, the diagnostic. NX-162, NX-163, NX-164 put a town PP and “yet” on different sides of doo with one reported translation; some speakers reject the last variant (pp.83–84). The cold idiom NX-165, NX-166 has an ordering restriction. Incorporated negative indefinites NX-182, NX-183 instead change who is understood as subject or object (p.100). So “doo can move anywhere without affecting meaning” would be false.
HYPOTHESISIndefinite LFs, not passive LFs: NX-182: ¬∃x∃e[person(x) ∧ see(e) ∧ Agent(e,x) ∧ Theme(e,boy)]; NX-183: ¬∃x∃e[person(x) ∧ see(e) ∧ Agent(e,boy) ∧ Theme(e,x)]. The data do not license replacing either with ∃x¬∃e[…], or with a question operator.
9.4 Headhood, scope, intervention: what the tests actually show
Dated correction, 2026-11-11. Trees redrawn; earlier versions mislabeled vP as “VPISH” and used non-category nodes. VPISH is the subject-inside-the-verbal-domain hypothesis, not a category: here its explicit implementation first-merges the subject DP in left Spec,vP. All specifiers are left; X⁰/X′/XP levels, head direction and scope are stated per panel. PF notes belong in captions, not syntax branches. The redraw also corrects the changeover and particle figures. These are our derivations to test, with earlier authors’ analyses kept as brief labeled contrasts.
Item / test
Source
SOURCED result
HYPOTHESIS / OPEN consequence
Initial daʼ
NX-114 versus NX-120–122; NX-140/141; NLA Ch.8 pp.174–175,180.
Fixed before ordinary clause constituents; may follow a vocative. Optional in other polar strategies. Co-occurs with -ísh (NX-132/156), but cannot itself host it (NX-133). Repetition in the cited coordination is rejected.
Clause-typing exponent is supported. Selecting C/Q head versus operator/specifier is still OPEN; fixedness and optionality alone do not decide.
Constituents may precede it; some reorders are dialect-dependent. It combines with verbal/nonverbal predicates. The draft even prints doubtful double doo, so a categorical no-stacking claim is unsupported.
Not demonstrated to be a fixed initial selecting head. Mobile concord, specifier/operator, and a head plus nominal displacement are rival hypotheses.
Right da / C-like endings
NX-167/178/179; Ch.6 pp.84–87,92.
da precedes the source’s postverbal past particle; it can combine with -go as dago, or the negative -góó form appears. The draft reports da absent with enclitic -ígíí; its own same-category conjecture is tentative.
A right-edge dependency is SOURCED. Neg head, concord, contextual exponence and C/Neg competition are not the same claim. None alone proves V crossed Neg.
Q can be on doo inside daʼ…; on an NPI after doo; and on final da in a specifically polite frame. NPI-Q without the negative frame is starred and spelling-flagged.
Host position is not Q’s selecting site. Interrogativity alone does not license this NPI. The polite frame is not a general interchangeability judgment.
The NLA draft distinguishes these. Content -sh may be on an initial non-WH topic or on a later WH; it cannot move to a later non-WH after an initial WH. Willie adds an inverse restriction.
A real host/intervention generalization, not an automatic V-to-C signature. Treating every sh-shaped item as the same head erases a source distinction.
Canonical inverse polar and WH questions exist. Initial non-WH-host relocation in the p.211 pair is blocked. Willie proposes a focus conflict.
Inverse must condition host licensing. A serial inverse-before-placement derivation and a parallel focus filter both fit; literal temporal precedence is HYPOTHESIS, not a proved V-movement block.
WH, D and P material
NX-113/137/139/191–202; NLA Ch.9 pp.187–191.
Question phrases can stay in situ; optional fronting, including rare long-distance cases, is described. DEM/Q words and P endings have different distributions.
A phrase can contain a P or D-like component without being C. Argument WH raises to Spec,CP only under an added analysis; WHY is an adjunct, not necessarily a raised subject.
Final tag yaʼ
NX-134/145–151; Ch.8 pp.187–188.
The tag follows an affirmative or negative proposition; it is incompatible with the cited daʼ, -ísh, WH and possibility-modal combinations.
Selection/compatibility is stronger evidence than bare finality, but a tag head versus a discourse-level adjunct is OPEN. Do not infer a universal CP-final head from one tag.
Verb before particles
NX-167; NX-110; Hale 2001:678–683; Ch.6 p.84.
The intact W precedes da/past material; mode and subject material occur inside W. No verified free Neg/adverb occurs between the stem and that inner material.
Both base-final structure and raising to a final head can give W-before-particle. Orthographic cohesion supports a word domain but does not identify a syntactic landing site.
HYPOTHESIS In the editorial line above, person/role labels follow the source translation; they are not a new prefix-by-prefix analysis. Compare the actual host Doósh with the uninflected host doo: neither asks us to move the whole verb to that position.
Analysis A: Final C/Neg with left concord (working default V→v→T; experiencer to Spec,TP) Hypothesis
Analysis B: Initial C/Neg with right-edge concord (working default V→v→T; experiencer to Spec,TP) Hypothesis
(3) One negative proposition; two candidate C/Neg positions HypothesisNX-153, Doósh shééhonísin da?, Perkins & Fernald draft ch.8 p.194(126). Exact component tokens doo, shééhonísin, da are in NX-159, p.195(132). The decomposition of surface Doósh into a host plus Q is an analysis, not a newly quoted form. AdvP concord and C⁰=[Q] are our category assignments, not established headhood. Both panels interpret Q > one sentential Neg > event/state closure. Leaves do not assert overt separate argument pronouns. No verb movement is evidenced by this pair.Bracket notation (copy)
DP (id S-high; landing site of Move from subject): experiencer: you
T′
vP
DP (id subject; struck copy; moved to S-high): experiencer: you
v′
v⁰ (head; id v; head-moves to T)
VP (id lowVP)
V′
DP (id object): theme: me
V⁰ (head; id V; head-moves to v): shééhonísin
T⁰ (head; id T; hosts the “verb word” (V + v + T))
Neg⁰ (head; id Neg): da
C⁰ (head; id C; [Q]; probe for Agree Q-host with host)
Analysis B
CP
C′
C⁰ (head; id C; [Q]; probe for Agree Q-host with host)
NegP
NegP
Neg′
Neg⁰ (head; id host; goal of Agree Q-host): doo
TP
DP (id S-high; landing site of Move from subject): experiencer: you
T′
vP
DP (id subject; struck copy; moved to S-high): experiencer: you
v′
v⁰ (head; id v; head-moves to T)
VP (id lowVP)
V′
DP (id object): theme: me
V⁰ (head; id V; head-moves to v): shééhonísin
T⁰ (head; id T; hosts the “verb word” (V + v + T))
AdvP: da
HYPOTHESISNegative interrogative LF for NX-153:ASK_POL(¬∃e[know(e) ∧ Exp(e,you) ∧ Theme(e,me)]). Both depicted head-direction hypotheses generate that LF. Interpreting doo and da as independent ¬ operators would instead cancel negation and contradict the reported question; that interpretation fails.
Analysis B: Inverse: intended derivation rejected (working default V→v→T; agent to Spec,TP; theme to Spec,CP) Hypothesis
(4) Our vP role mappings and the inverse-sensitive Q-host test HypothesisNX-085 and starred NX-086, Willie 1991:211(16–17): yiztał and biztał are exact verb tokens. The source reports the host restriction; DP fronting, C=[Q] and Agree are our HYPOTHESES, not her demonstrated movement steps. A serial inverse-before-host rule and a parallel compatibility filter remain rivals. The inverse agent is WH; the demonstrative is its theme. Neither English passive wording nor the failed host licenses a V-movement-block conclusion.Bracket notation (copy)
DP (id host-high; landing site of Move from S-high)
C′
TP
DP (id S-high; landing site of Move from subject; struck copy; moved to host-high): DEM agent
T′
vP
DP (id subject; struck copy; moved to S-high; goal of Agree host licensed): DEM agent
v′
v⁰ (head; id v; head-moves to T)
VP (id lowVP)
V′
DP (id object): WH theme x
V⁰ (head; id V; head-moves to v): yiztał
T⁰ (head; id T; hosts the “verb word” (V + v + T))
C⁰ (head; id C; [Q]; probe for Agree host licensed with subject)
Analysis B
CP
DP (id host-high; landing site of Move from object): DEM theme
C′
TP
DP (id S-high; landing site of Move from subject): WH agent x
T′
vP
DP (id subject; struck copy; moved to S-high): WH agent x
v′
v⁰ (head; id v; head-moves to T)
VP (id lowVP)
V′
DP (id object; struck copy; moved to host-high; goal of Agree host conflict): DEM theme
V⁰ (head; id V; head-moves to v): biztał
T⁰ (head; id T; hosts the “verb word” (V + v + T))
C⁰ (head; id C; [Q]; probe for Agree host conflict with object)
HYPOTHESISRole-sensitive LF check. NX-085 asks for the patient: ASK_x∃e[kick(e) ∧ Agent(e,DEM) ∧ Theme(e,x)]. The intended, rejected NX-086 asks for the agent: ASK_x∃e[kick(e) ∧ Agent(e,x) ∧ Theme(e,DEM)]. A grammar must explain why that host fails while inverse WH questions such as NX-069 succeed. Reversing who did the kicking to save the string is not an explanation.
9.5 The working default: V⁰→T⁰ and subject→Spec,TP
Working default (since 2026-11-13). In a tensed clause V raises (via v/Voice) to T; the subject separately raises from Spec,vP to Spec,TP, and, when required, a subject/WH/focus/topic phrase moves farther to Spec,CP. Remnant-VP movement / VP pied-piping is a hypothesis about what else moves (needed for verb-final order if T⁰ is initial), not a substitute for V-head raising. An object WH can raise to Spec,CP without being the subject; an in-situ WH need not do so overtly. These are distinct operations.
HYPOTHESIS VPISH supplies no word order by itself. To find out what else moves, fix independently (i) which head is T, (ii) where sentential Neg and an event-level adverb attach, (iii) which PPs are VP-internal, and (iv) where the MVS ends. Then compare what crosses those fixed landmarks. Calling every affix T or moving an adverb after observing the order removes the test.
Analysis A: Working default: V→v→T; subject→Spec,TP Hypothesis
Analysis D: Working default plus remnant-VP movement above T (HYPOTHESIS): gives the verb-final order Hypothesis
(5) Schematic (placeholder leaves; NX-094 motivates the test but its words are not drawn): Our vP-to-T derivations: which operation builds the word? HypothesisA is the working default (V→v→T, subject Spec,vP→Spec,TP); D adds remnant-VP movement above T (HYPOTHESIS). Generic placeholders only. Data motivating the tests: NX-094 (Hale 2001:678–679); no-head-movement template is a brief contrast (Hale 2001:681). PF alternatives (no longer drawn here): Goad & Travis 2021 preprint pp.5–8, secondary reports of Speas/Harley. These drawings are ours, not author facsimiles. Subject movement and optional Agree are held fixed to isolate word building; neither follows from VPISH alone. A later V→C or subject→Spec,CP requires its own trigger and is not claimed here. The affirmative LF abstracts from tense/aspect; adding sentential Neg/Q must preserve their higher interpreted scope.Bracket notation (copy)
DP (id S-high; landing site of Move from subject): agent a
T′
T⁰ (head; id T; probe for Agree φ (optional) with subject; hosts the “V+v+T” (V + v + T))
vP
DP (id subject; goal of Agree φ (optional); struck copy; moved to S-high): agent a
v′
v⁰ (head; id v; head-moves to T)
VP (id lowVP)
PP: low PP
VP
AdvP: low adverb
VP
V′
DP (id object): theme t
V⁰ (head; id V; head-moves to v): predicate
Analysis D
TP
DP (id S-high; landing site of Move from subject): agent a
T′
VP (id VP-high; landing site of Move from lowVP)
T′
T⁰ (head; id T; probe for Agree φ (optional) with subject; hosts the “V+v+T” (V + v + T))
vP
DP (id subject; goal of Agree φ (optional); struck copy; moved to S-high): agent a
v′
v⁰ (head; id v; head-moves to T)
VP (id lowVP; struck copy; moved to VP-high)
PP: low PP
VP
AdvP: low adverb
VP
V′
DP (id object): theme t
V⁰ (head; id V; head-moves to v): predicate
Rival
Operations: HYPOTHESIS
Discriminating prediction
Verdict
1. V-to-v/Voice-to-T + S raising (working default)
Merge VPISH roles; Merge T and a subject landing site; optional φ-Agree; move V through heads and S to Spec,TP. Subject-to-Spec,CP is an additional triggered step, not obligatory in every question.
If T is above a fixed low adverb/PP, V can cross it while the PP/adverb stays behind. S raising can strand a subject-associated element. Moving V alone does not move the MVS. An independent ban on crossing Neg would constrain the path; no such Navajo ban is demonstrated here.
The working default in tensed clauses (2026-11-13). Compatible with the intact-word data after extra assembly/linearization choices, which are what is still OPEN. Mode/Subj is not independently identified as T. Inverse Q examples do not show a V-raising block. LF target · HYPOTHESIS: ASK_POL(¬∃e[know(e) ∧ Exp(e,you) ∧ Theme(e,me)]).
2. T lowering / post-syntactic merger (labelled contrast)
Keep V syntactically low; allow S raising as a separate hypothesis; assemble T/inflection with V at PF, with explicit adjacency/domain conditions.
Low PPs/adverbs stay where syntax put them. A genuinely intervening head or nonlocal morphological pairing tests lowering/locality. No clause-level doo inside W is verified. An adverb outside the word does not by itself block a local PF rule.
Compatible, not established. Harley/Speas-type merger accounts are documented in Goad & Travis pp.5–8. Local Dislocation is a PF reordering hypothesis, not automatically T lowering, and not identical to all Morphological Merger accounts. LF target · HYPOTHESIS: ASK_POL(¬∃e[know(e) ∧ Exp(e,you) ∧ Theme(e,me)]).
3. Hale template / skeleton filling (labelled contrast)
Select Qual, Mode/Subj, Voice and V; transfer features into a prosodic skeleton. No traditional V head movement is required. A higher subject movement would need its own reason.
Nuclear material can stay contiguous despite a discontinuous preverb–stem theme. PPs may precede W in the MVS without being carried upward by W. Fixed template order need not mirror LF scope.
Compatible, not uniquely established. Hale 2001:681 explicitly rejects traditional head movement for this proposal. The current Q/Neg pairs do not distinguish it from a suitably local merger account. LF target · HYPOTHESIS: ASK_POL(¬∃e[know(e) ∧ Exp(e,you) ∧ Theme(e,me)]).
4. VP pied-piping / remnant VP (what else moves, if T⁰ is initial)
Use A’s V→v→T and separate subject→Spec,TP steps, then raise the remaining VP with its independently diagnosed low PPs/adverbs by left adjunction to T′. This added adjunction is a HYPOTHESIS, not a right specifier. The remnant VP carries its low adjuncts and theme, not the pronounced V a second time. The fourth tree displays this combined sub-option; reconstruction or a higher Neg must preserve its LF.
All material genuinely inside the moved VP travels together; outside/stranded material does not. MVS=VP would predict co-movement of that whole domain, but that identity is unproved. To preserve sentential ¬∃, Neg must be above the interpreted landing site, or the VP must reconstruct.
Still OPEN. The PP/yet/doo permutations are compatible with multiple attachment/scrambling analyses and do not establish movement of a single VP/MVS. Pied-piping with unreconstructed high event closure and low Neg fails the clausal negative reading. LF target · HYPOTHESIS: ASK_POL(¬∃e[know(e) ∧ Exp(e,you) ∧ Theme(e,me)]).
5. V-to-C / subject-to-Spec,CP when required
Add a distinct C target and a trigger, independently of T and of a Q enclitic’s host.
A V-to-C account predicts identifiable verb displacement relative to clause operators; fronting a WH phrase only predicts phrasal displacement. A Q clitic on a noun, P or doo need not carry V.
NONE verified for V-to-C in this set. Optional WH fronting is SOURCED; its exact movement analysis remains HYPOTHESIS. No verified inverse/passive comparison isolates the C step. LF target · HYPOTHESIS: ASK_POL(¬∃e[know(e) ∧ Exp(e,you) ∧ Theme(e,me)]).
HYPOTHESISLF requirement for every movement rival. For the negative know question, all successful versions must end with ASK_POL(¬∃e[know(e) ∧ Exp(e,you) ∧ Theme(e,me)]). V movement leaves the event predicate’s interpretation at its argument-taking position. PF lowering changes no LF. Template filling changes no LF. A pied-piped VP interpreted above low Neg must reconstruct, or select a higher Neg site. A high-VP analysis with ∃e[… ∧ ¬P(e)] cannot replace the source’s ¬∃e[P(e)]: the former may require an event and the latter denies one. “Raises to T” does not, by itself, tell us which LF is computed.
OPENDo the low PPs/adverbs and MVS actually travel? NX-162–164 vary PP/yet order relative to doo, but neither independently locates T nor identifies a single moved constituent. The source’s scope pair NX-180/181 tracks the clause edge marked by da, not a word’s motion. Q on a PP (NX-115/124) and optional fronted WH (NX-195/201/202) show host/phrase variation, not that V carries its adjuncts. A four-way direct/inverse × negative/positive series with the same low PP/adverb is missing; true passive counterparts are missing too. None of the proposed movement steps is shown to be blocked specifically by the inverse or passive.
English control: sentential not versus constituent not
SOURCED Pollock (1989:366–368), opened in the freely available original PDF, contrasts lexical-verb position relative to sentential negation and VP adverbs in English/French; his account is a head-movement HYPOTHESIS, not an axiom for Navajo. His p.367 star on “John likes not Mary” belongs to that sentential reading and setup.
HYPOTHESISConstructed standard-English control, not a quotation: “John kisses not Mary but Sue” is fine as constituent correction. Its LF permits a kissing event with Sue as theme; it is not the same test as “John does not kiss Mary”, ¬∃e[kiss(e) ∧ Agent(e,John) ∧ Theme(e,Mary)]. For “John often kisses Mary”, OFTEN(∃e[kiss(e) …]), low-adverb placement bears on V raising only after the adverb’s attachment is fixed. VP pied-piping above that adverb carries it only if it is inside the VP; pied-piping above sentential Neg needs reconstruction or a higher Neg. We do not assign a blanket star to every surface “V not NP”.
9.6 Derivations: eight cells, four headedness rivals, explicit LF
HYPOTHESIS These are auditable derivation schemas, not claims that all their operations occur. Each states Merge, Agree, Move, spell-out and a reading prediction. Empty cells receive conditional predictions only, never invented Navajo. For B, the B1/B2/B3 variants are all tested; these clause-level cases do not independently locate a head in the distinguishing zones. The verb-assembly rivals in §9.5 can be paired with any compatible clause analysis.
A · Uniform final clause: C/Neg/Q final; initial material is non-head · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the reported initiator/experiencer and theme distinct; do not infer a passive from the English word order. Merge final selecting Q/C; put initial daʼ/doo, if pronounced, in operator/concord positions, not automatically in heads. The scope-bearing position must c-command its semantic complement. No sentential Neg is merged. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Linearize selected complements before their final heads; apply the observed word template and a separate enclitic-host rule. Final Q may be silent while a nonfinal host receives its exponent.
Prediction / match. Conditional fit: final structure plus a PF host rule matches the cited string and LF. A raw “Q must follow the whole clause” rule would wrongly reject the nonverbal hosts.
B · Initial functional heads only inside B1/B2/B3 · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the reported initiator/experiencer and theme distinct; do not infer a passive from the English word order. Choose B1, B2 or B3. Merge initial nuclear functional heads only inside that candidate, final selecting clause heads outside it; lexical V/P remain final. If initial daʼ/doo is instead a selecting outside head, this version fails there. No sentential Neg is merged. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Spell nuclear material into W in its observed order; attach clause-edge and host exponents separately. For these cells B1/B2/B3 give the same PF once outside-head identity is fixed.
Prediction / match. All three B variants remain conditional, not uniquely selected. A clitic’s host cannot be substituted for its selecting-head site to make one boundary win.
C · Rice scope-based morpheme order · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the reported initiator/experiencer and theme distinct; do not infer a passive from the English word order. Build an LF hierarchy with the required role and operator scope. Apply Rice’s greater-scope-to-the-right hypothesis to its declared Athabaskan morphological domain, not indiscriminately to every free clitic. No sentential Neg is merged. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Use the explicitly chosen stem-assembly hypothesis to obtain W; give clause-level Q/Neg their independently stated realization and host conditions. Scope alone does not supply their phonological hosts.
Prediction / match. Not contradicted by a host before the verb unless the law is extended beyond its word-morpheme domain. An unrestricted “every wider-scope exponent is farther right” version fails the initial/nonfinal operator exponents.
D · Cyclic linearization / phase-specific spell-out · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the reported initiator/experiencer and theme distinct; do not infer a passive from the English word order. Choose phase/spell-out domains independently; keep semantic Q over the interpreted proposition. A PF phase direction cannot invert LF scope. No sentential Neg is merged. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Linearize the declared word phase and clause phase separately, then apply a stated host-placement rule. No fixed phase inventory is supplied by these sentences alone.
Prediction / match. Can fit after directions and host rules are specified; without independently fixed phase boundaries it is a parameterized description, not a unique explanation.
SOURCEDNX-053: “Did the lightning kill the horse?”.
A · Uniform final clause: C/Neg/Q final; initial material is non-head · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Merge final selecting Q/C; put initial daʼ/doo, if pronounced, in operator/concord positions, not automatically in heads. The scope-bearing position must c-command its semantic complement. No sentential Neg is merged. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Linearize selected complements before their final heads; apply the observed word template and a separate enclitic-host rule. Final Q may be silent while a nonfinal host receives its exponent.
Prediction / match. Conditional fit: final structure plus a PF host rule matches the cited string and LF. A raw “Q must follow the whole clause” rule would wrongly reject the nonverbal hosts. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles.
B · Initial functional heads only inside B1/B2/B3 · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Choose B1, B2 or B3. Merge initial nuclear functional heads only inside that candidate, final selecting clause heads outside it; lexical V/P remain final. If initial daʼ/doo is instead a selecting outside head, this version fails there. No sentential Neg is merged. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Spell nuclear material into W in its observed order; attach clause-edge and host exponents separately. For these cells B1/B2/B3 give the same PF once outside-head identity is fixed.
Prediction / match. All three B variants remain conditional, not uniquely selected. A clitic’s host cannot be substituted for its selecting-head site to make one boundary win. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles.
C · Rice scope-based morpheme order · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Build an LF hierarchy with the required role and operator scope. Apply Rice’s greater-scope-to-the-right hypothesis to its declared Athabaskan morphological domain, not indiscriminately to every free clitic. No sentential Neg is merged. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Use the explicitly chosen stem-assembly hypothesis to obtain W; give clause-level Q/Neg their independently stated realization and host conditions. Scope alone does not supply their phonological hosts.
Prediction / match. Not contradicted by a host before the verb unless the law is extended beyond its word-morpheme domain. An unrestricted “every wider-scope exponent is farther right” version fails the initial/nonfinal operator exponents. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles.
D · Cyclic linearization / phase-specific spell-out · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Choose phase/spell-out domains independently; keep semantic Q over the interpreted proposition. A PF phase direction cannot invert LF scope. No sentential Neg is merged. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Linearize the declared word phase and clause phase separately, then apply a stated host-placement rule. No fixed phase inventory is supplied by these sentences alone.
Prediction / match. Can fit after directions and host rules are specified; without independently fixed phase boundaries it is a parameterized description, not a unique explanation. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles.
A · Uniform final clause: C/Neg/Q final; initial material is non-head · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the reported initiator/experiencer and theme distinct; do not infer a passive from the English word order. Merge final selecting Q/C/Neg; put initial daʼ/doo, if pronounced, in operator/concord positions, not automatically in heads. The scope-bearing position must c-command its semantic complement. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Linearize selected complements before their final heads; apply the observed word template and a separate enclitic-host rule. Final Q may be silent while a nonfinal host receives its exponent.
Prediction / match. Conditional fit: final structure plus a PF host rule matches the cited string and LF. A raw “Q must follow the whole clause” rule would wrongly reject the nonverbal hosts. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading.
B · Initial functional heads only inside B1/B2/B3 · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the reported initiator/experiencer and theme distinct; do not infer a passive from the English word order. Choose B1, B2 or B3. Merge initial nuclear functional heads only inside that candidate, final selecting clause heads outside it; lexical V/P remain final. If initial daʼ/doo is instead a selecting outside head, this version fails there. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Spell nuclear material into W in its observed order; attach clause-edge and host exponents separately. For these cells B1/B2/B3 give the same PF once outside-head identity is fixed.
Prediction / match. All three B variants remain conditional, not uniquely selected. A clitic’s host cannot be substituted for its selecting-head site to make one boundary win. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading.
C · Rice scope-based morpheme order · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the reported initiator/experiencer and theme distinct; do not infer a passive from the English word order. Build an LF hierarchy with the required role and operator scope. Apply Rice’s greater-scope-to-the-right hypothesis to its declared Athabaskan morphological domain, not indiscriminately to every free clitic. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Use the explicitly chosen stem-assembly hypothesis to obtain W; give clause-level Q/Neg their independently stated realization and host conditions. Scope alone does not supply their phonological hosts.
Prediction / match. Not contradicted by a host before the verb unless the law is extended beyond its word-morpheme domain. An unrestricted “every wider-scope exponent is farther right” version fails the initial/nonfinal operator exponents. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading.
D · Cyclic linearization / phase-specific spell-out · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the reported initiator/experiencer and theme distinct; do not infer a passive from the English word order. Choose phase/spell-out domains independently; keep semantic Q over the interpreted proposition and Neg over event closure. A PF phase direction cannot invert LF scope. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Linearize the declared word phase and clause phase separately, then apply a stated host-placement rule. No fixed phase inventory is supplied by these sentences alone.
Prediction / match. Can fit after directions and host rules are specified; without independently fixed phase boundaries it is a parameterized description, not a unique explanation. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading.
OPEN No verified Navajo form or source reading. The LF below is only the specified target of a future test.
A · Uniform final clause: C/Neg/Q final; initial material is non-head · OPEN, prediction only
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Merge final selecting Q/C/Neg; put initial daʼ/doo, if pronounced, in operator/concord positions, not automatically in heads. The scope-bearing position must c-command its semantic complement. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Linearize selected complements before their final heads; apply the observed word template and a separate enclitic-host rule. Final Q may be silent while a nonfinal host receives its exponent.
Prediction / match. UNTESTED: no source datum for this cell. Conditional fit: final structure plus a PF host rule matches the cited string and LF. A raw “Q must follow the whole clause” rule would wrongly reject the nonverbal hosts. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading. This is not evidence that the candidate target LF is available.
B · Initial functional heads only inside B1/B2/B3 · OPEN, prediction only
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Choose B1, B2 or B3. Merge initial nuclear functional heads only inside that candidate, final selecting clause heads outside it; lexical V/P remain final. If initial daʼ/doo is instead a selecting outside head, this version fails there. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Spell nuclear material into W in its observed order; attach clause-edge and host exponents separately. For these cells B1/B2/B3 give the same PF once outside-head identity is fixed.
Prediction / match. UNTESTED: no source datum for this cell. All three B variants remain conditional, not uniquely selected. A clitic’s host cannot be substituted for its selecting-head site to make one boundary win. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading. This is not evidence that the candidate target LF is available.
C · Rice scope-based morpheme order · OPEN, prediction only
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Build an LF hierarchy with the required role and operator scope. Apply Rice’s greater-scope-to-the-right hypothesis to its declared Athabaskan morphological domain, not indiscriminately to every free clitic. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Use the explicitly chosen stem-assembly hypothesis to obtain W; give clause-level Q/Neg their independently stated realization and host conditions. Scope alone does not supply their phonological hosts.
Prediction / match. UNTESTED: no source datum for this cell. Not contradicted by a host before the verb unless the law is extended beyond its word-morpheme domain. An unrestricted “every wider-scope exponent is farther right” version fails the initial/nonfinal operator exponents. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading. This is not evidence that the candidate target LF is available.
D · Cyclic linearization / phase-specific spell-out · OPEN, prediction only
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Choose phase/spell-out domains independently; keep semantic Q over the interpreted proposition and Neg over event closure. A PF phase direction cannot invert LF scope. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge polar ASK over the proposition.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Working default (§9.5): V→v→T, and the subject Spec,vP→Spec,TP. No V-to-C step follows from adding a polar marker.
Spell-out. Linearize the declared word phase and clause phase separately, then apply a stated host-placement rule. No fixed phase inventory is supplied by these sentences alone.
Prediction / match. UNTESTED: no source datum for this cell. Can fit after directions and host rules are specified; without independently fixed phase boundaries it is a parameterized description, not a unique explanation. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading. This is not evidence that the candidate target LF is available.
A · Uniform final clause: C/Neg/Q final; initial material is non-head · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the reported initiator/experiencer and theme distinct; do not infer a passive from the English word order. Merge final selecting Q/C; put initial daʼ/doo, if pronounced, in operator/concord positions, not automatically in heads. The scope-bearing position must c-command its semantic complement. No sentential Neg is merged. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Linearize selected complements before their final heads; apply the observed word template and a separate enclitic-host rule. Final Q may be silent while a nonfinal host receives its exponent.
Prediction / match. Conditional fit: final structure plus a PF host rule matches the cited string and LF. A raw “Q must follow the whole clause” rule would wrongly reject the nonverbal hosts.
B · Initial functional heads only inside B1/B2/B3 · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the reported initiator/experiencer and theme distinct; do not infer a passive from the English word order. Choose B1, B2 or B3. Merge initial nuclear functional heads only inside that candidate, final selecting clause heads outside it; lexical V/P remain final. If initial daʼ/doo is instead a selecting outside head, this version fails there. No sentential Neg is merged. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Spell nuclear material into W in its observed order; attach clause-edge and host exponents separately. For these cells B1/B2/B3 give the same PF once outside-head identity is fixed.
Prediction / match. All three B variants remain conditional, not uniquely selected. A clitic’s host cannot be substituted for its selecting-head site to make one boundary win.
C · Rice scope-based morpheme order · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the reported initiator/experiencer and theme distinct; do not infer a passive from the English word order. Build an LF hierarchy with the required role and operator scope. Apply Rice’s greater-scope-to-the-right hypothesis to its declared Athabaskan morphological domain, not indiscriminately to every free clitic. No sentential Neg is merged. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Use the explicitly chosen stem-assembly hypothesis to obtain W; give clause-level Q/Neg their independently stated realization and host conditions. Scope alone does not supply their phonological hosts.
Prediction / match. Not contradicted by a host before the verb unless the law is extended beyond its word-morpheme domain. An unrestricted “every wider-scope exponent is farther right” version fails the initial/nonfinal operator exponents.
D · Cyclic linearization / phase-specific spell-out · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the reported initiator/experiencer and theme distinct; do not infer a passive from the English word order. Choose phase/spell-out domains independently; keep semantic Q over the interpreted proposition. A PF phase direction cannot invert LF scope. No sentential Neg is merged. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Linearize the declared word phase and clause phase separately, then apply a stated host-placement rule. No fixed phase inventory is supplied by these sentences alone.
Prediction / match. Can fit after directions and host rules are specified; without independently fixed phase boundaries it is a parameterized description, not a unique explanation.
A · Uniform final clause: C/Neg/Q final; initial material is non-head · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Merge final selecting Q/C; put initial daʼ/doo, if pronounced, in operator/concord positions, not automatically in heads. The scope-bearing position must c-command its semantic complement. No sentential Neg is merged. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Linearize selected complements before their final heads; apply the observed word template and a separate enclitic-host rule. Final Q may be silent while a nonfinal host receives its exponent.
Prediction / match. Conditional fit: final structure plus a PF host rule matches the cited string and LF. A raw “Q must follow the whole clause” rule would wrongly reject the nonverbal hosts. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles.
B · Initial functional heads only inside B1/B2/B3 · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Choose B1, B2 or B3. Merge initial nuclear functional heads only inside that candidate, final selecting clause heads outside it; lexical V/P remain final. If initial daʼ/doo is instead a selecting outside head, this version fails there. No sentential Neg is merged. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Spell nuclear material into W in its observed order; attach clause-edge and host exponents separately. For these cells B1/B2/B3 give the same PF once outside-head identity is fixed.
Prediction / match. All three B variants remain conditional, not uniquely selected. A clitic’s host cannot be substituted for its selecting-head site to make one boundary win. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles.
C · Rice scope-based morpheme order · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Build an LF hierarchy with the required role and operator scope. Apply Rice’s greater-scope-to-the-right hypothesis to its declared Athabaskan morphological domain, not indiscriminately to every free clitic. No sentential Neg is merged. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Use the explicitly chosen stem-assembly hypothesis to obtain W; give clause-level Q/Neg their independently stated realization and host conditions. Scope alone does not supply their phonological hosts.
Prediction / match. Not contradicted by a host before the verb unless the law is extended beyond its word-morpheme domain. An unrestricted “every wider-scope exponent is farther right” version fails the initial/nonfinal operator exponents. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles.
D · Cyclic linearization / phase-specific spell-out · conditional fit
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Choose phase/spell-out domains independently; keep semantic Q over the interpreted proposition. A PF phase direction cannot invert LF scope. No sentential Neg is merged. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Linearize the declared word phase and clause phase separately, then apply a stated host-placement rule. No fixed phase inventory is supplied by these sentences alone.
Prediction / match. Can fit after directions and host rules are specified; without independently fixed phase boundaries it is a parameterized description, not a unique explanation. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles.
A · Uniform final clause: C/Neg/Q final; initial material is non-head · conditional fit
HYPOTHESIS
Merge. Merge go with its agent in Spec,vP; attach a reason variable to the negative proposition, not as the subject of go. Merge final selecting Q/C/Neg; put initial daʼ/doo, if pronounced, in operator/concord positions, not automatically in heads. The scope-bearing position must c-command its semantic complement. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Linearize selected complements before their final heads; apply the observed word template and a separate enclitic-host rule. Final Q may be silent while a nonfinal host receives its exponent.
Prediction / match. Conditional fit: final structure plus a PF host rule matches the cited string and LF. A raw “Q must follow the whole clause” rule would wrongly reject the nonverbal hosts. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading.
B · Initial functional heads only inside B1/B2/B3 · conditional fit
HYPOTHESIS
Merge. Merge go with its agent in Spec,vP; attach a reason variable to the negative proposition, not as the subject of go. Choose B1, B2 or B3. Merge initial nuclear functional heads only inside that candidate, final selecting clause heads outside it; lexical V/P remain final. If initial daʼ/doo is instead a selecting outside head, this version fails there. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Spell nuclear material into W in its observed order; attach clause-edge and host exponents separately. For these cells B1/B2/B3 give the same PF once outside-head identity is fixed.
Prediction / match. All three B variants remain conditional, not uniquely selected. A clitic’s host cannot be substituted for its selecting-head site to make one boundary win. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading.
C · Rice scope-based morpheme order · conditional fit
HYPOTHESIS
Merge. Merge go with its agent in Spec,vP; attach a reason variable to the negative proposition, not as the subject of go. Build an LF hierarchy with the required role and operator scope. Apply Rice’s greater-scope-to-the-right hypothesis to its declared Athabaskan morphological domain, not indiscriminately to every free clitic. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Use the explicitly chosen stem-assembly hypothesis to obtain W; give clause-level Q/Neg their independently stated realization and host conditions. Scope alone does not supply their phonological hosts.
Prediction / match. Not contradicted by a host before the verb unless the law is extended beyond its word-morpheme domain. An unrestricted “every wider-scope exponent is farther right” version fails the initial/nonfinal operator exponents. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading.
D · Cyclic linearization / phase-specific spell-out · conditional fit
HYPOTHESIS
Merge. Merge go with its agent in Spec,vP; attach a reason variable to the negative proposition, not as the subject of go. Choose phase/spell-out domains independently; keep semantic Q over the interpreted proposition and Neg over event closure. A PF phase direction cannot invert LF scope. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Linearize the declared word phase and clause phase separately, then apply a stated host-placement rule. No fixed phase inventory is supplied by these sentences alone.
Prediction / match. Can fit after directions and host rules are specified; without independently fixed phase boundaries it is a parameterized description, not a unique explanation. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading.
OPEN No verified Navajo form or source reading. The LF below is only the specified target of a future test.
A · Uniform final clause: C/Neg/Q final; initial material is non-head · OPEN, prediction only
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Merge final selecting Q/C/Neg; put initial daʼ/doo, if pronounced, in operator/concord positions, not automatically in heads. The scope-bearing position must c-command its semantic complement. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Linearize selected complements before their final heads; apply the observed word template and a separate enclitic-host rule. Final Q may be silent while a nonfinal host receives its exponent.
Prediction / match. UNTESTED: no source datum for this cell. Conditional fit: final structure plus a PF host rule matches the cited string and LF. A raw “Q must follow the whole clause” rule would wrongly reject the nonverbal hosts. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading. This is not evidence that the candidate target LF is available.
B · Initial functional heads only inside B1/B2/B3 · OPEN, prediction only
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Choose B1, B2 or B3. Merge initial nuclear functional heads only inside that candidate, final selecting clause heads outside it; lexical V/P remain final. If initial daʼ/doo is instead a selecting outside head, this version fails there. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Spell nuclear material into W in its observed order; attach clause-edge and host exponents separately. For these cells B1/B2/B3 give the same PF once outside-head identity is fixed.
Prediction / match. UNTESTED: no source datum for this cell. All three B variants remain conditional, not uniquely selected. A clitic’s host cannot be substituted for its selecting-head site to make one boundary win. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading. This is not evidence that the candidate target LF is available.
C · Rice scope-based morpheme order · OPEN, prediction only
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Build an LF hierarchy with the required role and operator scope. Apply Rice’s greater-scope-to-the-right hypothesis to its declared Athabaskan morphological domain, not indiscriminately to every free clitic. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Use the explicitly chosen stem-assembly hypothesis to obtain W; give clause-level Q/Neg their independently stated realization and host conditions. Scope alone does not supply their phonological hosts.
Prediction / match. UNTESTED: no source datum for this cell. Not contradicted by a host before the verb unless the law is extended beyond its word-morpheme domain. An unrestricted “every wider-scope exponent is farther right” version fails the initial/nonfinal operator exponents. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading. This is not evidence that the candidate target LF is available.
D · Cyclic linearization / phase-specific spell-out · OPEN, prediction only
HYPOTHESIS
Merge. Merge the predicate with its theta-role arguments inside vP. Keep the patient/topic first at PF without exchanging agent and theme. Inverse role mapping must be licensed before or jointly with the host filter; serial timing is not established. Choose phase/spell-out domains independently; keep semantic Q over the interpreted proposition and Neg over event closure. A PF phase direction cannot invert LF scope. Merge one sentential Neg operator above event closure; realize its two exponents without two semantic negations. Merge the WH abstraction/reason operator above the relevant proposition; it need not front the lexical V.
Agree. Agree is an optional φ/feature-licensing hypothesis; the cited words alone do not prove its probe or landing site. Host licensing is separately conditioned by focus/inverse compatibility.
Move. Keep WH in situ for the displayed pair; optional WH/topic movement is a separate hypothesis, not V movement.
Spell-out. Linearize the declared word phase and clause phase separately, then apply a stated host-placement rule. No fixed phase inventory is supplied by these sentences alone.
Prediction / match. UNTESTED: no source datum for this cell. Can fit after directions and host rules are specified; without independently fixed phase boundaries it is a parameterized description, not a unique explanation. It must admit the canonical inverse question but reject the incompatible non-WH-host configuration exemplified by NX-086, without swapping roles. Any derivation giving affirmative event closure, two cancelling negations, or only argument correction fails the specified negative reading. This is not evidence that the candidate target LF is available.
The higher combination: Neg inside versus outside expect
HYPOTHESISA: Merge the lower VP, its Neg above closure, and expect above that negative proposition for NX-180; for NX-181 Merge Neg above expect instead. Agree is optional; V→v→T applies by default and does not change which clause Neg takes. Spell da at the right edge of the negated clause and doo by its licensing rule. B: the same semantic Merge choices are required in all three boundary variants; initial versus final nuclear heads does not alter which clause Neg takes. C: keep the two semantic scopes distinct and apply the morpheme-order hypothesis only in its domain; the free frame still needs a clause-realization rule. D: declare which clause is spelled out under Neg; changing phase direction alone cannot substitute for moving the semantic Neg site. In every rival the LF must be NX-180 ∃s[expect(s) ∧ Exp(s,John) ∧ Content(s,⟦¬∃e∃x[car(x) ∧ buy(e) ∧ Agent(e,Mary) ∧ Theme(e,x)]⟧)], versus NX-181 ¬∃s[expect(s) ∧ Exp(s,John) ∧ Content(s,⟦∃e∃x[car(x) ∧ buy(e) ∧ Agent(e,Mary) ∧ Theme(e,x)]⟧)]. Assigning both strings the first LF is a real failure, even if their words are spelled correctly.
9.7 Verdict: what survives, and the most useful missing test
HYPOTHESISPlain answer: we have a clause-sized negative dependency and a voice-sensitive question-host rule, and no uniquely located head-direction switch. V-to-T plus subject raising is the working default (2026-11-13), not something these data had to prove; the no-head-movement analyses are kept only as labelled contrast. What else moves is OPEN: VP pied-piping has an additional co-movement/reconstruction burden. The clause is not proved uniformly head-final merely by calling every inconvenient left element a non-head.
Retained conditional generalization, not a unique-boundary theorem:
one semantic Neg may have two exponents;
interpreted Q > sentential Neg > event/state closure;
inverse preserves the reported agent/theme mapping;
host(Q) ≠ selecting-site(Q) in the definition of headedness.
Still unestablished:
∃! B ∈ {B1,B2,B3} ∀ functional H:
HD_base(H)=initial ↔ H is inside B.
Neither HD_surface nor an enclitic’s host can replace HD_base in that formula.
HYPOTHESIS The literal “everything is initial inside the word, final outside” remains too strong: lexical V/P and the affix/stem-head parameter already break that shorthand. The functional-only revision remains conditional on headhood and base order. If initial daʼ or doo is independently shown to be a selecting functional head outside all three B domains, the revision fails; if they are operators/concord, the narrow nuclear fragment still fits all three boundaries and selects none uniquely. Rice’s account and phase-specific linearization have not been eliminated, but they must be specified, not invoked as automatic derivations.
OPENSingle most discriminating missing experiment: one scope-controlled, matched negative polar paradigm with a transitive predicate, an independently low VP adverb and a selected PP, in direct and inverse (then classifier-passive) forms. Independently identify the T landing site and fix sentential ¬∃e, not constituent correction. Does the verb cross that fixed low material while leaving it behind (head-raising prediction), does the whole diagnosed VP/MVS cross together (pied-piping), or is only the word assembled without displacement (PF/template prediction)? Ask for scope and binding/reconstruction judgments as well as order. No Navajo forms are constructed here and no speaker has been contacted. Without independent T/attachment diagnostics, even this paradigm would not uniquely distinguish B1/B2/B3; no single unanchored surface string can do that.
OPEN Negative inverse polar and WH forms; matched transitive negative WH; matrix passive polar/WH and negative-passive-question combinations for each passive family.
OPEN Independent selection, complement-size, binding and intervention diagnostics separating C/Q/Neg heads from operators/concord and discourse tags; T’s identity versus Mode/Subj; overt subject versus pronominal argument.
OPEN Word-internal free-adverb/Neg intervention, displacement of a whole independently diagnosed MVS, and whether inverse/passive blocks any particular movement rather than just a host/focus pairing.
OPEN A same-verb doo…da argument-focus versus event-negation pair; productive Q/Neg scope ambiguity, not just a negative question’s English wording; contemporary dialect judgments for the draft’s disputed variants.
HYPOTHESISDated correction / qualification, 2026-11-11. The older §3 inference that an enclitic’s observed host “is B2, not B3” is too strong as an established boundary result. It is a host-domain prediction, not a demonstrated selecting-head or phase boundary. Likewise, the older blanket “once per question” needs the source’s coordination exception (NLA Ch.8 p.177(25)); daʼ and -ísh can co-occur, whereas the tag exclusions are separately sourced. The older claim that the unassigned categories are not supplied by a source should not conceal the new scope and compatibility evidence. These qualifications preserve the older text as a research record rather than silently rewriting it.
9.8 Research audit and opened sources
SOURCED The 892 structured harvest entries and 2,343 italic-form candidates were screened, with 114 structured and 948 form candidates retained by the question/negation/passive filter. Candidate IDs are not verification IDs and were not copied into the index. Existing NX-001–118 were cross-checked for relevant contrasts; Willie’s pp.37,92,204,211 page images were reopened for the decisive passive, polar, WH and host data. Concordances were generated over Willie, Speas, Hale–Platero, Jelinek, Platero and Chee–Yazzie material; the new NLA negation, polar, content and complement chapters were read and their quoted pages checked as images. This is a substantial harvest, not a claim to have exhausted every Navajo source: OCR and source-availability gaps remain.
SOURCED Perkins, Ellavina T., and Theodore B. Fernald. A Grammar of Navajo, NLA Working Papers 1, title page “Revised 2014”; working-draft Ch.6 Negation, especially pp.83–88,92–95,100,107,117–118,121; Ch.8 Yes/No Questions, pp.173–195; Ch.9 Content Questions, pp.187–191; Ch.5 Clausal Complements, pp.43–45 on direct/inverse and passive embedding. Source caveats and draft notes matter; a cited Y&M/Reichard page is not claimed as independently opened.
SOURCEDNavajo Questions, yes/no, numbered examples1–15; content questions, examples1–5, v3.2022. These HTML quotations are unpaginated and may differ from PDF-draft spellings.
SOURCED Willie1991, Navajo Pronouns and Obviation, pp.37–38,57,71,92,99–100,201–215. The index retains source judgments; a starred string is not a ban on its entire voice family.
SOURCED Hale2001, Navajo Verb Stem Position and the Bipartite Structure of the Navajo Conjunct Sector, pp.678–683,689–690; Hale2003 manuscript p.32 and n.4, for the rival nuclear architecture. Goad & Travis2021, authors’ preprint pp.2–3,5–8, for Rice’s scope hypothesis and documented movement/merger alternatives. Rice2000 and original Harley/Speas merger papers are secondary reports here, not newly opened originals.
NONE No usable local Schauber1979 question monograph was found; the freely listed Schauber1975 MIT precursor’s PDF download returned HTTP429 and was not quoted from snippets. Files named for Barss1989 and Perkins1978 proved to contain, respectively, Eddington–Lachler on verb stems and an unrelated nonnative-English study: neither was used as the named primary source. NLA’s reports of Barss/Schauber are identified as secondary.
OPEN Withheld rather than repaired: the full embedded passive question Ch.8(74); the difficult negative paid clause’s voice classification; unverified Schauber snippet forms; and the missing negative inverse/passive combinations. The spelling-flagged NPI contrast is quoted as a provisional source judgment, not a corrected new Navajo form.
See also the prefix/particle comparison. This update does not propagate an unproved changeover claim to other topics.
Restated on the unified structure (2026-11-15). Nouns are restated as arguments merged in their theta positions (VP, Spec,vP) and, for the subject, raised to Spec,TP by the working default; a lone noun is linked by the H4″ default. Adjunction and pronominal-argument analyses below are labelled contrast; where nouns sit for Condition C is to be re-run on the unified spine. Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Status on the spine: see unified structure, §2 and §4, first rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.
2026-11-11 audit pass 3. Verdict words on this page (SUPPORTS, AGAINST, CONTESTED, Neutral) are our tallies against the variants of the dropped thesis, not status labels: read SUPPORTS and AGAINST as HYPOTHESIS (our assessment) and CONTESTED as OPEN. See the pass 3 report.
The configurational objection to Grant's thesis, read in the original. Margaret Speas's MIT dissertation (Adjunctions and Projections in Syntax, 1986) argues that Navajo nouns sit in argument positions and that the pronominal prefixes are not the verb's arguments. This page checks every Navajo datum in her Chapter 2 §2.2.9, Chapter 3 and Chapter 4 against the feature-bundle hypothesis (FB, on Not quite a word). Her data and her analysis are kept apart throughout. Written 2026-10-14, 8:46pm run (PT). Closes FB-a / TA-a for the 1986 version; the revised 1990 book chapter is still unread.
Ground rules (Grant's). Speas, Jelinek, Willie and Platero are each treated as possibly partly right. A source that does not state FB is not evidence against it. Only data that contradict FB count against it. Dowty's (1991) proto-roles are the baseline. The Navajo rank scale (FB-4, κ) is treated as an extension of Dowty: a weighting of proto-agent properties plus topicality. Every Navajo form below is copied from a page image of the dissertation, with its example number and page. None is invented.
Bottom line
Update 2026-10-15 (12:46pm): Hale 1999 “Conflicting Truths” checked (page). HYPOTHESIS Hale's (10) and (11a,b) show Condition C excluding a reading where both participants are people, so rank plays no part (pp. 172–173). That supplies the T22b case. Whether it counts against FB-5's exemption is conditional on assumptions A1–A10 (audit); 1:07pm correction: the 12:46pm wording (“contradicts”, “withdrawn”) was too strong. The D6 verdict is compatible, undecided: “SUPPORTS FB” is no longer claimed, because Hale's (11a) and Speas's (4.119) have the same shape and opposite judgments (T24a), and neither is secured. FB's core is untouched; LINK-IN and FB-5′/FB-6′ are conditional additions. (4.15a) is COMPATIBLE, not SUPPORTS. (4.119)–(4.121) are OPEN. Tally: 0 against, 1 supports, 16 compatible, 1 open.
Update 2026-11-08 (8:46pm PT):§10, Speas’s yi-/bi- sentences re-derived with Grant’s light determiners. HYPOTHESIS With “d stays, only the DP moves”, the bi- clause is the one where the object DP has moved above the subject DP; with Platero’s NP-PRO this derives every reading in (4.6)–(4.9). It shows that a DP moves, not that d stays (T104a–c).
Update 2026-11-09 (4:46am PT):§11, Platero’s relative-clause paradigm (4.15) re-run with the same rules. HYPOTHESIS NP-PRO over DP positions gets 12 of 14 readings but wrongly excludes (4.15h), as Speas says of Platero’s rule (p. 257); Speas’s subject-to-subject, object-to-object parallelism, restated over the d positions (ID-7), plus rank for (4.15d), gives all 14. Nothing here separates “d stays” from “d moves” (T104c answered; T105a–d).
Update 2026-11-09 (12:46pm PT):§12, bi- relatives (4.53)–(4.55) with the same rules. HYPOTHESIS Parallelism has to pair the noun phrases by where they are after the bi- movement, not by the d head they were merged with: in (4.54) the horse is kicked in one clause and bites in the other. With that restatement (ID-7′) and a rule that a bi- relative cannot be headed by its subject (ID-8, (4.53b)), all printed judgments come out; “d stays” is still untested (T105d answered; T106a–c).
Update 2026-11-09 (8:47pm PT):§13, which noun phrase heads a relative clause. HYPOTHESIS The §12 rule that the subject jumped over by a bi- object can never head a bi- relative (ID-8) holds only when both participants are named by nouns inside the relative; with null arguments Willie’s bi- relatives are headed by exactly that subject, and Hale & Platero’s (42) needs it too (ID-8′). Willie’s (29) confirms that parallelism follows positions after the inverse movement (ID-7′). “d stays” is still untested (T106a, T106c answered; T106b narrowed; T107a–c).
Update 2026-11-10 (4:50am PT):§14, a noun after the relative marker. HYPOTHESIS Hale & Platero’s good (20) and Willie’s starred (24b) differ in the noun after the relative marker: for Hale & Platero it is the relative’s external head, while Willie says her speech has no such heads and calls this a possible dialect difference from Platero (p. 158). Willie’s definiteness account would rule out both, so it does not explain the contrast. “d stays” is still untested (T107b answered; T108a–c).
Update 2026-11-10 (12:47pm PT):§15, raising or matching for a noun after the relative marker. HYPOTHESIS Hale & Platero’s parse rule (44) cannot tell raising from matching, because it reads only the finished string (their p. 24). Their own conclusion is raising, not the deletion analysis §14 attributed to them (correction noted in §14). If the noun after the relative is a separate main-clause noun (Willie’s analysis), (20) needs exactly the configuration Willie stars in (24b), so T108a also decides between the analyses. Relativization leaves the verb’s light determiners in place, which fits “d stays” but does not test it (T108c answered for (44), otherwise reduced to T34a; T108a sharpened; T109a–b).
HYPOTHESISSpeas's Navajo data do not contradict FB, T-MID or T-WEAK. None of her Navajo data (18 groups of sentences and readings, §7) needs a noun in an argument position. Her case that nouns are arguments rests on two tests in Ch. 2, the absence of weak crossover (2.95) and grammatical multiple questions (2.96c). Both are aimed at one particular version of the Pronominal Argument Hypothesis, in which nouns are adjuncts bound by A-bar operators. They do not bear on FB, which does not bind nouns as operators. Her Ch. 4 relative-clause data point the other way. A null matrix subject can refer to a noun inside the object clause, (4.15a). It can also refer to a noun inside an embedded clause with no relative clause at all, (4.119)–(4.121). In her own structure both of those are Condition C violations, and she has to add across-the-board (ATB) representations to let them through. She concludes that Navajo nominals “behave like adjuncts insofar as they show ATB effects, although they are in argument positions” (p. 323). (8:46pm reading; since 1:07pm, (4.119) is one side of T24a, the conflict with Hale's (11a).)
HYPOTHESISWhat FB has to borrow. FB does not get her relative-clause paradigm for free. It needs two positional rules that restate Platero's descriptive generalizations, which Speas reports (pp. 250, 285): LINK-NEAR (§6) and, for the equal-rank case (4.52), the same subject-to-subject, object-to-object parallelism that her ATB constraint states (p. 264). Restated over bundles, these rules plus the rank rule FB-4 derive all five good readings of Platero's 14-way paradigm and exclude the nine bad ones, with no Condition C and no ATB (§5, §6).
Version
Effect of Speas 1986's data
T-STRONG (nouns are never arguments; roles alone decide links)
Still fails. Two more data against it: a single noun links by position, not role (4.8); relative-clause parallelism matches external with external and internal with internal even when the participant's role in the action differs, as when the horse is kicked in one clause and bites in the other (4.54).
T-MID / FB (the arguments are role-bearing φ-bundles spelled as prefixes; nouns link to them)
Survives, with no contradicting datum. Condition C (D6) is compatible, undecided (Hale 1999 and Speas 1986 give opposite judgments; audit). It now carries two explicit borrowed rules (FB-3a LINK-NEAR, FB-6 PAR).
T-WEAK (Pronominal Argument Hypothesis)
Survives on the data. Speas's two Ch. 2 tests do hit the specific A-bar-adjunct version of it (Saito's ECP argument).
Speas's own analysis (nouns in θ-positions; prefixes as non-θ tails of a CHAIN)
Not refuted, but not forced either. No datum in Ch. 3–4 distinguishes “θ-role on the noun, prefix as tail” from “θ-role on the bundle, noun linked”. Her own conclusion assigns the θ-role to the noun-plus-clitic CHAIN (p. 324).
Not beyond contest yet.OPEN The decisive tests are not in this source: Condition C with two participants of equal rank, weak crossover with a question word or quantifier as antecedent, and quantifier scope. The revised Navajo chapter of Speas 1990 (pp. 202–281) and Barss, Hale, Perkins & Speas 1989 are still unread.
SOURCED Speas, Margaret Jean 1986. Adjunctions and Projections in Syntax. PhD dissertation, MIT. Scan, 333 pp. Read: §2.2.9 “Navajo Overt Nominals are in Argument Positions” (pp. 196–197, with pp. 190–195); Ch. 3 “Pronominal Argument Infixes” (pp. 198–243); Ch. 4 Navajo sentence structure (pp. 244–324), including §4.3 Parallelism in Relative Clause Constructions (p. 254), §4.4 the ATB treatment (p. 263), §4.5 Huang-style pro-drop (pp. 270–285), §4.6 Mapping Conventions (p. 285) and the end of §4.7 (pp. 319–324).
Page mapping. PDF page = printed page. Checked on pp. 196, 197, 201, 249, 250, 254, 264, 276, 278, 282, 284, 319 and 324, where the printed folio matches the PDF index.
OCR. The scan's text layer is unusable for Navajo. It was used only to find sections. Every form below was read on page images (90–300 dpi).
Whose data. Nearly all Navajo sentences are Speas's citations of Platero (1974, 1978, 1982) or of Young & Morgan. (2.96) reports Hale's informants and her own. The judgments are theirs, and Platero notes that one parsing is “difficult to elicit” (p. 281).
Spelling and gloss slips (kept as printed). (4.15) is reprinted as (4.38) in plainer spelling (hastiin liiʼ yizloh-ee yiʼdiilid). We quote the (4.15) spelling. In (4.6b) the word glosses read ‘girl boy’ under ashkii atʼééd, but the translation ‘The girl saw the boy’ fits ashkii = boy. biztał is glossed ‘bit’ in (4.53b)/(4.54) but translated ‘kicked’, and yishxash is glossed and translated ‘bit’. In (4.15) the tone mark on the second vowel of -ę́ę is faint. (4.6a) prints a gap in yi yiiłtsą́. In (4.119) the nasal hooks in naayáhą́ą are faint, and the first letters of the last word are blotted (“…anaa'”, glossed ‘would-that’).
1986 vs 1990. Speas 1990 (Phrase Structure in Natural Language) revises this material. Nothing here should be read as her final position.
2. What Speas gets right, and what she misses
Gets right (FB keeps)
Misses or leaves undone
The prefixes are not phonological decoration. They must be visible to syntax, as “infixes” in a discontinuous lexical item (pp. 200–202).
The real typological line is between languages “in which sub-parts of a word fill argument ‘positions’” and those in which they do not (p. 192). FB says Navajo is on the first side.
Platero's paradigm is real and hard: 14 possible readings of one string, only 5 good (pp. 254–255). Simple control and pro-drop accounts fail on it (pp. 270–285).
Parallelism holds over grammatical relations, subject with subject and object with object (p. 264), not over nouns. The noun-less (4.39) shows this, since the restriction holds with no nouns at all (p. 269).
The possessive prefix bi- “may be + or − argument” (p. 279). That concedes FB-1 for the possessor slot.
She raises, then declines, the reading that “the pronominal argument morphemes may be in argument positions in syntax, and may simply move into the verb at Phonetic Form” (p. 201). Two of her three reasons are theory-internal (typology, morphology). The third is the Ch. 2 tests, which do not bear on FB (§4).
Her Condition C case (4.15d)/(4.38d) is confounded with rank: a horse acting on a man in a yi- verb. FB-4 excludes it whatever the coreference (§5.3).
Her own data make nouns look adjunct-like (Condition C exemptions, ATB). She absorbs this with ATB representations and a “generalized transformation” view of adjunction (pp. 320–324) rather than by moving the θ-role off the noun.
No quantifier, scope or wh-crossover data with a non-referential antecedent.
She treats verbal prefixes as “embedded within an unstructured string of morphemes” (p. 278) after arguing in Ch. 3 that they are syntactically accessible. The tension is not resolved (T22f).
3. Chapter 3: prefixes as infixes, and how that relates to FB-1
SOURCED The problem (pp. 199–200): the Navajo prefixes that mark grammatical functions sit inside an unstructured string of prefixes, so verb-internal configurations cannot define subject and object. Speas argues that the subject and object prefixes must be infixed. At some level they are “external to a discontinuous lexical item, to be subsequently inserted into the affix string” (pp. 200–201).
SOURCED Her key passage (p. 201): “If an infixation model for Navajo should turn out to be correct, then … the pronominal argument morphemes may be in argument positions in syntax, and may simply move into the verb at Phonetic Form. I will not draw this conclusion, however.” Her reasons are wrong typological predictions, loss of the affixes' morphological properties, and the Ch. 2 tests showing “NPs in Navajo are not in adjoined positions, but are themselves in argument positions”. So “although the pronominal prefixes are accessible to the syntax, they are not actually arguments as far as the Projection Principle is concerned” (pp. 201–202).
SOURCED Her solution (p. 202; §3.0.5, pp. 241–243): a weak Lexicalist Hypothesis and an extended CHAIN, in which “the head of a CHAIN may occupy a theta-position while the tail of the CHAIN occupies a non-theta position, as long as the tail of the CHAIN is a sub-part of a word and hence not subject to the Projection Principle”. A morpheme's person/number features are independent of its category, so they can sit on an NP or on an inflectional morpheme (p. 242). The final picture (4.124, p. 324) pairs syntactic NPi, NPj with morphological [V S-cli [O-clj [stem]]], and “the theta role is assigned to the CHAIN”.
HYPOTHESISFB-1 is the option Speas declines on p. 201. In VPISH terms it is: Merge the bundle in Spec,v or Comp,V (clause-level syntax), then spell it out inside the verb word (rule set A, word-internal linearization; see headedness). Her CHAIN and FB pair the same two objects, a noun and a prefix, with the same index. They differ only on which member holds the θ-position. Speas: the noun (head), with the prefix as non-θ tail. FB: the bundle, with the noun linked. Her stated reasons for choosing the noun:
Typology (pp. 193–194): if pronouns could be arguments, some language should have obligatory independent pronouns with nominals adjoined (her (2.91)). Does not apply to FB. FB-5 makes the argument φ-level, and φ-level proforms are typically affixes or clitics (Déchaine & Wiltschko 2002, as on the FB page). FB predicts no such language.
Morphological properties of the affixes (p. 201). Theory-internal. No Navajo datum.
The Ch. 2 tests: the only Navajo data offered. Checked in §4. They do not contradict FB.
4. Chapter 2: her two argumenthood tests
4.1 No weak crossover
SOURCED (2.95), p. 196:
Jáani bii-zhé'é yiyiiłtsą́ (= Jáani bii zhé'é [proi pro yiyiiłtsą́])
John 3.POSS-father 3OBJ.3SUBJ.saw
'Johni saw hisi father'
Speas: if overt nominals were adjoined, they would A-bar bind both the possessive bi and the subject pronominal, and weak crossover would make (2.95) bad. It is “fully grammatical” (p. 196).
HYPOTHESISCOMPATIBLE with FB. FB-3 linking is feature matching between a noun and a bundle, not operator–variable binding. Jáan is a name, not a quantifier or question word. Referential antecedents are the case where crossover is weakest even in English (“weakest crossover”, Lasnik & Stowell 1991; cited, not opened). Speas herself treats the possessive bi- as possibly an argument (p. 279). The test that could hurt FB would use a question word or quantifier as the subject (‘whoi saw hisi father?’). No such sentence is in the source (T22a).
4.2 Multiple questions
SOURCED (2.96), p. 197 (from Saito 1984; judgments of Hale's informants and Speas's own):
a. *Háí lá ha'át'íí lá yiyiiłtsą́? who-PRT what-PRT saw 'Who saw what'
b. *Jáan-ga' at'ééd-ga' yiyiiłtsą́ John-PRT girl-PRT saw 'John saw the girl.'
c. Háí-lá ha'át'íí yiyiiłtsą́? 'Who saw what?'
Speas: Saito's adjunct analysis predicts that multiple questions are out, since two wh-adjuncts moved to COMP would leave two adjunct traces. (a) is out only because “multiple focus is out” (b). (c) is fine, so “nominals are not adjuncts in Navajo” (p. 197).
HYPOTHESISCOMPATIBLE with FB. The argument refutes one mechanism, adjunct traces subject to the ECP. Under FB each question word links to its own bundle (háí to the external one, haʼátʼíí to the internal one), and the bundles are inside the verb, so there is no unlicensed trace in an argument position. This matches the site's wh-questions page (WH-SLOT). (2.96a, b) show a one-focus limit, which FB does not address. That is silence, not contradiction.
5. Chapter 4: yi-/bi-, relative clauses, Condition C
5.1 yi-/bi- with two, one and no overt nouns (pp. 249–251)
(4.6) a. ashkii at'ééd yi yiiłtsą́. boy girl yi-saw 'The boy saw the girl'
b. ashkii at'ééd b-iiłtsą́. [boy girl] bi-saw 'The girl saw the boy'
(4.7) a. yi-yiiłtsą́. 'S/he saw him/her'
b. b-iiłtsą́. 'S/he saw him/her' (talking about him/her)
(4.8) a. At'ééd yiiłtsą́. girl yi-saw 'Sh/he saw the girl' NOT 'The girl saw him/her'
b. At'ééd b-iiłtsą́. girl bi-saw 'The girl saw him/her' NOT 'Sh/he saw the girl'
(4.9) Ashkii yiyiiłtsą́. *'The boyi saw himi' (fine as 'Hei saw the boyj', n. 5)
(4.10) Ashkii ádiiłtsą́. 'The boy saw himself' (Platero 1978:108)
SOURCED (4.8a) is printed yiiłtsą́ with word gloss ‘yi-saw’ (we keep the print). Speas: in (4.6) the interpretation of grammatical relations is “switched” by the object prefix (SOV vs OSV, p. 249). Platero's NP-PRO constraint covers (4.8): “If PRO immediately follows an overt noun phrase NP', it must be coreferential with NP'” (Platero 1982:288, quoted p. 250). Speas's n. 4 (p. 251) allows that in the reflexive “the verbal morphology removes the object theta role”.
HYPOTHESIS FB reading. A single noun always links to the internal bundle. Under yi- that is the patient (4.8a). Under bi-, FB-4 has made the patient external, so the internal bundle is the agent (4.8b). That is one positional rule (FB-3a below), and it needs no null NP. Coreference between the two arguments is decided inside the word, by the reflexive prefix (4.10), not by a noun binding a pronoun (4.9). All COMPATIBLE. None of these needs a noun in an argument position.
No Condition C: the matrix bundle does not c-command a linked noun (FB-5)
b) ‘He/shej branded the horse that the mani roped’
✓
normal parse
null bundle, free reference (4.7)
c) ‘He/shej branded the man that roped the horsei’
✓
normal parse
as (b), head = man
d) *‘Iti branded the man that roped the horsei’
*
Condition C; not ATB
FB-4: horse external over a human under yi-
ii) man [[pro horse roped] REL] branded
e) ‘The mani branded the horse hei roped’
✓
NP-PRO coindexes pro with man
LINK-NEAR(a): hastiin links into the first verb (= a, same truth conditions)
f) *‘The mani branded the horse he/shej roped’
*
NP-PRO
LINK-NEAR(a)
g) *‘The mani branded the onej who roped the horse’
*
NP-PRO
LINK-NEAR(a)
iii) [[man horse roped] REL] pro branded
h) ‘The man who roped the horsei branded iti’
✓
ATB
LINK-NEAR(b): closest noun = internal bundle of both verbs
i) *‘… branded him/herj’
*
NP-PRO; not ATB
LINK-NEAR(b)
j) *‘The horse that the mani roped branded himj’
*
NP-PRO; not ATB
FB-4 (horse over a human) + LINK-NEAR(b)
k) *‘The horse that the mani roped branded himi’
*
NP-PRO; not ATB
FB-4 + LINK-NEAR(b)
iv) *man [[horse pro roped] REL] branded
l) *‘The mani branded the horse that roped himi’
*
NP-PRO in all interpretations
FB-4 (horse ropes a human) + LINK-NEAR(a)
m) *‘… the horse that roped him/herj’
*
FB-4 + LINK-NEAR(a)
n) *‘The mani branded the onej that the horse roped’
*
LINK-NEAR(a); FB-4 too if j is human
SOURCED Speas's ATB constraint (p. 264): “In a parallel structure P, made up of component sentences S1 and S2, the subject of S1 and the subject of S2 must be coreferential, and the object of S1 and the object of S2 must be coreferential.” Platero's generalization (p. 285): “no matter how the sentence is parsed, the NP which is string-closest to the verbs must be the object of both verbs.” (4.39) Yizloh-ee yiʼdiilid ‘S/he branded the one that s/hei,?j roped’ shows the restriction with no nouns at all (p. 269).
HYPOTHESIS The FB column uses only FB-4 (rank), FB-5 (no Condition C from a φ-bundle onto a linked noun) and LINK-NEAR (two clauses restating Platero, §6). It gets the same 5 good and 9 bad readings as Speas without ATB or Condition C. Rank alone does a lot of the work. Every bad reading in which the horse acts on a person (d, j, k, l, m) is excluded by FB-4 before coreference is considered. Dowty's proto-agent properties are the baseline here, and the rank weighting is the Navajo extension (eight ranks, who acts on whom).
5.3 The Condition C facts
SOURCED Speas's two problems (pp. 256–257): (1) why is (4.15a) good, “where the empty pronominal precedes, is coindexed with and c-commands (therefore binds) the referring expression man, in violation of standard coreference principles?” (2) why is (4.15d) bad? Outside relative clauses (pp. 319–320):
(4.119) Kóhoot'éédą́ą́' shínaaí Naakaii Bikéyahgóó naayáhą́ą yee nihił hólne …anaa'
last year my-brother Mexico-to went-NOM about-it us tell(OPT) would-that
'I wish my brother would tell us about his going to Mexico last year'
(4.121) parsed as: 'I wish hei would tell us about my brotheri's going to Mexico last year.'
Correction 2026-10-16: the verb is printed hólne with no final glottal stop (p. 319, page image at 300 dpi); earlier versions of this page had hólneʼ. Site-by-site re-run of this sentence against Hale's (11a): Condition C by site.
Because ‘last year’ modifies the trip, shínaaí is inside the embedded clause, “so that the apparent violation of condition C which was encountered in the relative clause case is found here too” (p. 320). She concludes that “the ATB effect is not related to the fact that relative clauses are internally headed” (p. 320).
HYPOTHESISVerdict: SUPPORTS FB over a plain configurational structure. Replaced 2026-10-15 (12:46pm), restated 1:07pm as compatible, undecided: Hale 1999 (10), (11a,b) supply an equal-rank case with the opposite judgment from (4.119), but the inference from them to noun placement is conditional on assumptions A1–A10 (derivation page), and (4.119) pulls the other way (T24a). Original text of the 8:46pm verdict: A null matrix argument coreferent with a noun it c-commands is just what a φ-level bundle and a linked (not merged) noun predict. Speas needs a separate device, ATB plus generalized transformation, to allow it. Her one Condition C exclusion, (4.15d), is a horse acting on a man in a yi- verb (Speas restricts her discussion to yi- forms, p. 251 n. 6), which FB-4 excludes on rank. So her data contain no case where Condition C, rather than rank or position, is the only thing that rules a reading out. OPEN A minimal pair with two participants of equal rank would decide it (T22b).
5.4 bi- relatives, proper names and equal rank (pp. 278–284)
(4.52) Asdzą́ą́ Kii ya'níłtsoodę́ę yich'ahóóshkeed (Platero 1982:293)
woman Kii 3-fed-3-REL 3-scolded-3
*'The woman scolded Kii who she fed.' 'The woman who fed Kiii scolded himi,*j'
(4.53) a. łį́į́' dzaanééz yiztał-ę́ę horse mule yi-bit-REL 'The horse that kicked the mule' OR 'The mule that the horse kicked'
b. łį́į́' dzaanééz biztał-ę́ę horse mule bi-kick-REL 'The horse that was kicked by the mule' / *'The mule that the horse was kicked by'
(4.54) łį́į́' dzaanééz biztał-ę́ę yishxash horse mule bi-bit-REL yi-bit
'The horse that was kicked by the mulei bit iti' *'The horse bit the mule that it was kicked by'
(4.55) *ashkiii at'ééd yiyiiłtsanę́ę proi yizts'ǫs boy girl saw-REL kissed
'The girl who the boyi saw kissed himi'
SOURCED In (4.52) the name cannot head the relative, so the relative is the matrix subject and the null object must be Kii (p. 281). Speas's revised control rule for Navajo (4.57) is, in her words, “completely ad-hoc” and still leaves Platero's problem unsolved (p. 284). Linear control also fails for -go clauses and coordination:
(4.58) Hastiin deezghal-go nídii'na' man he.awoke-COMP he.got.up
a) 'When the mani awoke, hei got up.' b) 'When the mani awoke, Hej got up.' (Platero 1978:15)
(4.59) Ashkii at'ééd yí'déesol dóó hastiin yiyiiłtsą́ dóó deezgo'.
boy girl whistled.at and man saw and fell
'The boyi whistled at the girl, hei saw the man and hei fell.' (P 1978:66)
HYPOTHESIS (4.52) and (4.55) are the cases where rank cannot help: woman/Kii and boy/girl are equal. LINK-NEAR(b) still decides them. The noun closest to the verbs (Kii, atʼééd) is the internal bundle of both verbs, so ‘scolded Kii’ is forced and ‘the girl kissed him’ is out. (4.54) is the important one for Grant's thesis. The horse is the patient of ‘kick’ but the agent of ‘bite’, yet it is the external participant in both clauses (external under bi-, external under yi-). So parallelism matches bundle positions after voice, not the participant's role in the action. That cuts against T-STRONG (roles alone) and fits FB, whose bundles have a position as well as a role. (4.58) shows that FB's linking rule must link overt nouns and must not force reference on null bundles, and that is how LINK-NEAR is stated.
5.5 Possessive bi- (pp. 278–279)
(4.49) ashkii [NP PRO bizhé'é] yiyiiłtsą́ 'The boy saw hisi,j/her father'
(4.50) ashkii bizhé'é yáłti' 'The boy's father is speaking'
SOURCED “if we suppose that the possessive clitic pronoun bi- may be + or − argument, then we need not assume the existence of PRO” (p. 279). It is the only third-person possessive form, with no yi-form (n. 16). This clitic, “unlike the verbal prefixes, is not embedded within an unstructured string of morphemes” (p. 278). HYPOTHESISSUPPORTS FB-1 for the possessor slot. Speas grants that the prefix can itself be the argument, with ashkii linked to it (4.50). What keeps her from saying the same of verbal prefixes is the “unstructured string”, which her own Ch. 3 infix model was built to dissolve (T22f).
6. Rules FB needs (borrowed from Platero and Speas, restated over bundles)
FB-3a LINK-NEAR (a) An overt noun links to an open bundle of the nearest
following verb (a nominalized clause counts as a verb's
domain); it may not skip an open bundle of that verb.
(b) The noun string-closest to a verb cluster links to, or
identifies, the INTERNAL bundle of every verb in the
cluster ("internal" = after voice, FB-4).
Null bundles are otherwise free in reference (4.7), (4.58).
FB-6 PAR When a nominalized clause C is linked to a bundle of V, null
bundles of V may be identified with the bundle of C's verb in
the same position (external↔external, internal↔internal).
Rule
Status
Basis
FB-3a(a)
HYPOTHESIS
Restates Platero's NP-PRO constraint (1982:288, via Speas p. 250) as linking of the overt noun, so (4.58b) stays possible. Revised 2026-10-15: a default, not a hard rule (FB-3a(a′)). Hale & Platero 1974 (44) is relaxed by some speakers in context ((41)–(42); page)
FB-3a(b)
SOURCED as Platero's generalization (p. 285), and already in Hale & Platero 1974:17, (28); bundle form HYPOTHESIS
(4.8), (4.15h–k), (4.52), (4.55)
FB-6
SOURCED as Speas's ATB constraint (p. 264); bundle form HYPOTHESIS
(4.39), (4.54). Overlaps FB-3a(b) on the paradigm; kept because (4.39) has no nouns for LINK-NEAR to use
HYPOTHESISIs this simpler than Speas? Partly. FB needs no Condition C exemption and no ATB representation for (4.15a) or (4.119). But FB-3a and FB-6 are stipulations, just as Platero's rules were. Speas's objection to Platero-style rules was that, with null pronouns in argument positions, they clash with Condition B (pp. 250, 283–284). FB has no null pronouns in argument positions (the bundles are the arguments), so that particular objection does not carry over. Whether LINK-NEAR is grammar or a parsing preference is open (T22c).
7. Datum-by-datum table
Speas datum (page)
Speas's conclusion
FB prediction
Verdict for FB
(4.6a/b) two nouns, yi-/bi- (p. 249)
Object prefix “switches” relations
FB-4: bi- marks external ≠ ρA; first noun links to the external bundle
COMPATIBLE
(4.7) no nouns (p. 249)
Both arguments may be null
FB-2: bundles complete without nouns
COMPATIBLE (not discriminating; D1)
(4.8) one noun (p. 250)
NP-PRO constraint
FB-3a(b): lone noun = internal bundle
COMPATIBLE
(4.9)/(4.10) reflexive (p. 251)
Condition B on pro; perhaps the morphology removes a θ-role (n. 4)
Identity of two bundles set inside the word
COMPATIBLE
(2.95) no weak crossover (p. 196)
Nominals are in argument positions
Linking is not operator binding; name antecedent
COMPATIBLE; decisive version OPEN (T22a)
(2.96c) multiple questions (p. 197)
Nominals are not adjuncts (contra Saito)
Each wh-word links to its own bundle; no adjunct trace
COMPATIBLE
(4.15a) ‘Hei branded the horse the mani roped’ ✓ (pp. 255–256)
Condition C violation; needs ATB
FB-5: no Condition C from a bundle onto a linked noun Revised 2026-10-15 (12:46pm): FB-6′ parallel link (Hale 1999 (13); CT4)
SUPPORTSCOMPATIBLE
(4.15d) *horse brands man (p. 268)
Condition C
FB-4 rank, whatever the coreference
COMPATIBLE (confounded; T22b)
(4.15f,g,i–k) and Platero's generalization (pp. 268–269, 285)
NP-PRO + ATB
FB as stated is silent; needs FB-3a
OPEN → COMPATIBLE with FB-3a
(4.15l–n) horse acts on a person (p. 269)
NP-PRO
FB-4 (and FB-3a(a))
COMPATIBLE
(4.39) no nouns, restricted readings (p. 269)
Only one parse; ATB indistinguishable
Restriction must be stated on slots (FB-6)
COMPATIBLE (favors a slot statement)
(4.52) Kii, equal rank (p. 281)
Relative must be subject; object = Kii
Rank cannot decide; FB-3a(b)
COMPATIBLE with FB-3a; shows FB needs more than κ
(4.53)/(4.54) bi- relative under ‘bite’ (pp. 281–282)
Empty-object parse must exist; Huang-style account fails
Parallelism over external/internal, not over role
COMPATIBLE; CUTS AGAINST T-STRONG
(4.55) *girl kissed himi (p. 283)
Parallelism, not closeness
FB-3a(b): atʼééd is the internal bundle
COMPATIBLE
(4.49)–(4.51) possessive bi- (pp. 278–279)
The clitic “may be + or − argument”
FB-1 for the possessor slot
SUPPORTS
(4.58)/(4.59) -go, coordination (p. 284)
Linear control rule is wrong
Null bundles are free; LINK-NEAR links overt nouns only
COMPATIBLE
(4.119)–(4.121) Condition C without a relative (pp. 319–320)
ATB effect is general to adjunction
FB-5 Revised 2026-10-15 (12:46pm): FB-5′ and FB-6′ predict disjointness, like Hale's (11a), which has the same shape and the opposite judgment
SUPPORTSOPEN (T24a)
Ch. 4 conclusion (pp. 323–324)
Nominals are in argument positions but adjunct-like; θ-role on the CHAIN
θ-role (ρ) on the bundle; noun linked
COMPATIBLE (no datum separates the two)
Revised tally 2026-10-15 (12:46pm), after Hale 1999: 0 CUTS AGAINST FB's core; 1 SUPPORTS; 16 COMPATIBLE; 1 OPEN (why). Original 8:46pm tally:
Tally over 18 rows: 0 CUTS AGAINST FB; 3 SUPPORTS; 15 COMPATIBLE (one only with the new FB-3a); open points T22a–g. The one CUTS AGAINST is for T-STRONG, which had already failed.
8. VPISH-aware derivations
Conventions as on the thesis page and derivations: Σ = schema step; Merge/Agree/Move at clause level; A = word-internal linearization (rule set A, head-initial, the prefix string); B = clause level (rule set B: nouns, clauses, particles). V-P = VPISH raising of the external bundle to Spec,M/S inside the word. Proto-role labels are ours (Dowty 1991 baseline).
SP1. (4.8a) Atʼééd yiiłtsą́ ‘Sh/he saw the girl’, NOT ‘The girl saw him/her’ (p. 250)
Σ Σ(√SEE) = {β1[ρ experiencer: sentient], β2[ρ stimulus]} HYPOTHESIS (labels)
Merge β2 in Comp,V ; β1 in Spec,v VPISH base positions
Agree β1 = [π3], β2 = [π3]; κ(β1) ≥ κ(β2) → yi- (FB-4) SOURCED (yi- gloss)
Move V-P: β1 → Spec,M/S HYPOTHESIS
A spell-out yiiłtsą́ : yi- (β2, 3OBJ) … ∅ (β1, 3SUBJ) … stem SOURCED (form, p. 250)
B at'ééd adjoined at the left edge; FB-3a(b): lone noun → β2 SOURCED (NOT-reading)
Result: the noun's role comes from the bundle it links to; β1 needs no
noun and no PRO. Speas: at'ééd Merged in Comp,V; null pro in Spec,v.
Σ as SP1: {β1[ρ experiencer], β2[ρ stimulus]}
Agree κ(boy) = κ(girl) (equal rank) → either voice; bi- chosen SOURCED (Creamer p. 32; form p. 249)
Move bi-: β2 is external (topic); V-P raises β2 → Spec,M/S HYPOTHESIS
A spell-out b-iiłtsą́ SOURCED
B ashkii (first) → external β2 ; at'ééd (next to V) → internal β1
by FB-3a(b) SOURCED (reading)
Result: the girl, a proto-agent by Dowty's entailments, is the internal
bundle; position after voice, not role, places the nouns.
SP3. (4.15a) hastiin łį́į́ʼ yizloh-ę́ę yíʼdiiłid ‘Hei branded the horse that the mani roped’ (pp. 254–256)
Inner verb (roped)
Σ Σ(√ROPE) = {e1[ρA], e2[ρP]} ; κ(man) > κ(horse) → yi- HYPOTHESIS / SOURCED (yi- forms, p. 251 n. 6)
Merge e2 Comp,V ; e1 Spec,v ; V-P e1 → Spec,M/S
A spell-out yizloh ; B + -ę́ę (REL, clause-level) SOURCED
B FB-3a(a): hastiin → e1 ; łį́į́' → e2 (SOV default) SOURCED (reading)
head = łį́į́' (internally headed relative)
Outer verb (branded)
Σ {m1[ρA], m2[ρP]}
Merge m2 Comp,V ; m1 Spec,v ; V-P m1 → Spec,M/S
A spell-out yí'diiłid
B [hastiin łį́į́' yizloh-ę́ę] linked to m2 (head horse) ; m1 null, free
m1 = man : no Condition C, since m1 is φ-level and hastiin is
linked inside the relative, not Merged in an A-position FB-5 HYPOTHESIS
Result: reading (a) without ATB. Speas: Condition C violated in (4.17);
rescued only by the ATB representation (pp. 256, 266–267).
SP4. Where the bad readings die: (4.15d) and (4.15f)
(4.15d) *'Iti branded the man that roped the horsei'
B relative linked to m2 with head = hastiin ; m1 = horse
Agree FB-4: κ(m1 = horse) < κ(m2 = man) but voice = yi- → ✗ HYPOTHESIS (rank: eight ranks)
(crash at Σ/Agree, before any binding theory)
(4.15f) *'The mani branded the horse he/shej roped'
B hastiin → m1 would skip the open e1 of the nearer verb yizloh
→ violates FB-3a(a) → ✗ HYPOTHESIS (Platero's NP-PRO restated)
Result: one bad reading is excluded by rank (an extension of Dowty's
proto-agent weighting), one by position. Neither needs Condition C.
Inner {e1[ρA], e2[ρ recipient]} ; asdzą́ą́ → e1 ; Kii → e2 SOURCED (gloss 3-fed-3)
head ≠ Kii (a proper name cannot head a relative) SOURCED (p. 281)
Outer {m1[ρA], m2[ρP]} ; relative (head woman) → m1
B FB-3a(b): Kii is string-closest → identifies m2 HYPOTHESIS
κ(woman) = κ(Kii): FB-4 decides nothing here
Result: 'The woman who fed Kiii scolded himi,*j'. Rank cannot do
this; position does. This is the case that makes FB borrow Platero's rule.
9. Open questions T22a–g
#
Question
What would decide it
T22a
Weak crossover with a non-referential antecedent: is ‘whoi saw hisi father?’ (háí with bizhéʼé) good?
Speaker judgments, or Barss, Hale, Perkins & Speas 1989
T22b
A Condition C minimal pair with two participants of equal rank (e.g. two people), so rank cannot be what excludes (4.15d)-type readings
Speaker judgments; Platero 1978 originals. Partly answered 2026-10-15 (12:46pm, corrected 1:07pm): Hale 1999 (10), (11a,b) give a case with two people, which is in tension with FB-5's exemption conditional on A1–A10; Speas's (4.119) points the other way (derivation and audit)
T22c
Is LINK-NEAR (FB-3a) grammar or a parsing preference? Platero says one parse is “difficult to elicit” (p. 281). Partly answered 2026-10-15: Hale & Platero 1974 state the same rule as a parsing principle (44), “employed, with varying degrees of strictness”, and print a speaker split ((40)–(43)). FB-3a(a) becomes the default FB-3a(a′) (page)
Judgments with context that favors the excluded reading
T22d
Does Speas 1990 Ch. 4 keep the CHAIN analysis, drop ATB for the Parallel Function Constraint (1990:232), or add new tests?
Read Speas 1990 pp. 202–281 (FB-a remainder)
T22e
Gloss and spelling slips: biztał ‘bit’ vs ‘kicked’; (4.6b) word glosses; (4.15) vs (4.38) spelling; (4.8a) yiiłtsą́
Platero 1974/1978/1982 originals
T22f
Speas lets possessive bi- be an argument (p. 279) but not verbal yi-/bi-, because verbal prefixes are in an “unstructured string” (p. 278). Does her own infix model (Ch. 3) remove that difference?
Theory; Preminger-style clitic tests (FB-c)
T22g
Does the (4.119) Condition C exemption also hold with an overt matrix subject coreferent with the embedded noun?
Speaker judgments
Sources: Speas 1986 (page images, box copy, not committed). Platero 1974, 1978, 1982; Young & Morgan; Saito 1984; Huang 1984; Woolford 1986: cited through Speas, not opened. Lasnik & Stowell 1991: cited, not opened. Dowty 1991, Creamer 1974, Déchaine & Wiltschko 2002: opened in earlier runs (see sources).
10. Update 2026-11-08 (evening): Speas’s yi-/bi- sentences re-derived with Grant’s light determiners — the inverse is where “the DP moves, d stays” does real work (T104a–c)
HYPOTHESIS In plain words: take Grant’s 2026-11-05 version of the light-determiner idea literally — each argument is a small head d with a full noun phrase as its complement, d stays put, and only the noun phrase may move. Then the bi- sentence is simply the one where the object’s noun phrase has moved up past the subject’s, and the object d left behind is spelled bi- instead of yi-. Add one rule Speas quotes from Platero (an unpronounced pronoun right after a noun must mean the same person as that noun) and every reading and every “NOT” reading in Speas (4.6)–(4.9) comes out, with no extra rule about lone nouns. The same sentences also show a limit: the prefix string is identical whether d moved or stayed, so these data support the “noun phrase moves” half of the proposal, not the “d stays” half.
Notation as on the rest of this page: B = clause syntax (rule set B, right-headed: complements and specifiers before heads); A = word spell-out (rule set A, the left-headed prefix string). The word boundary is the changeover from B to A. Merge builds structure, Move is internal Merge of something already built, Agree only values features. This section uses Grant’s light-determiner analysis (essay, analysis D as revised 2026-11-05: stationary d), not the feature-bundle rules of §§6–8, so that the two can be compared on the same data.
10.1 Sources opened this run
SOURCED Speas 1986, pp. 249–251, re-read on the page images of the box copy (not the text layer, which garbles the Navajo): examples (4.6a,b), (4.7a,b), (4.8a,b), (4.9), (4.10), footnotes 2, 4, 5, 6, and the NP-PRO Constraint as Speas quotes it from Platero 1982:288 (p. 250).
Not opened: Platero 1978, 1982 (cited through Speas only); Young & Morgan p. 170 (source of (4.7c–e), not used here). Site results reused, not re-derived: the deep-inverse verdict on deep or shallow inverse (HYPOTHESIS there) and Saxon 2003’s two OSV constructions on two ways to front the object.
10.2 Rules first
ID-1 STRUCTURE (Grant, 2026-11-05). Object: [dP d_obj DP_obj], complement of V. GRANT'S CONJECTURE
Subject: [dP d_subj DP_subj] in Spec,vP (or, on the Syntax bot's split, / HYPOTHESIS
DP_subj in Spec,vP with d_subj a spine head; nothing below depends on which).
d stays where it is merged. Only a DP may be internally merged elsewhere.
Null arguments are syntactically present as a null pronominal DP (pro).
ID-2 INVERSE MOVE. In a bi- clause DP_obj is internally merged above DP_subj HYPOTHESIS
(outer Spec,vP or the subject position; landing site OPEN, T104a). d_obj
stays inside VP. In a yi- clause there is no such movement. (Which clauses
must or may do this is the rank condition, FB-4 / RCC; not restated here.)
ID-3 EXPONENT OF d_obj (rule set A). With two third persons, d_obj is spelled bi- HYPOTHESIS
if the highest copy of its own DP c-commands DP_subj, and yi- otherwise.
This restates Speas p. 249 ("switched" due to the object prefix) over
positions. It is not local: d_obj is no longer next to its DP (T104b).
ID-4 CLAUSE ORDER (rule set B). DPs precede the verb word; a higher DP precedes SOURCED (orders)
a lower one. Speas p. 249: (4.6a) is SOV with yi-, (4.6b) OSV with bi-. HYPOTHESIS (the rule)
ID-5 NP-PRO (Platero 1982:288, quoted Speas p. 250): "If PRO immediately follows SOURCED
an overt noun phrase NP', it must be coreferential with NP'." Speas n. 2:
Platero's PRO is a generic empty element (here: pro). Applied to the
string of DP positions after ID-2 and ID-4.
ID-6 NO COREFERENCE WITHOUT THE REFLEXIVE. A null object pro of a non-reflexive SOURCED (Speas pp. 250–251:
verb cannot be coreferential with the clause-mate subject; the reflexive Condition B; (4.9) vs (4.10))
verb is needed, (4.10).
10.3 Data (Speas 1986, pp. 249–251, page images)
Glosses are Speas’s word glosses in Leipzig layout; she segments only the yi-/bi- prefix, so the rest of each verb is glossed as one piece. Spacing and hyphens are as printed.
Ex. (page)
Navajo
Gloss
Translation as printed
(4.6a) p. 249
ashkiiatʼéédyi yiiłtsą́
boy girl yi-saw
‘The boy saw the girl’ (SOV)
(4.6b) p. 249
ashkiiatʼéédb-iiłtsą́
boy girl bi-saw
‘The girl saw the boy’ (OSV)
(4.7a) p. 249
yi-yiiłtsą́
yi-saw
‘S/he saw him/her’
(4.7b) p. 249
b-iiłtsą́
bi-saw
‘S/he saw him/her’ (talking about him/her)
(4.8a) p. 250
Atʼéédyiiłtsą́
girl yi-saw
‘Sh/he saw the girl’, NOT ‘The girl saw him/her’
(4.8b) p. 250
Atʼéédb-iiłtsą́
girl bi-saw
‘The girl saw him/her’, NOT ‘Sh/he saw the girl’
(4.9) p. 251
Ashkiiyiyiiłtsą́
boyi he-saw-himi
*‘The boyi saw himi’; n. 5: fine as ‘Hei saw the boyj’
(4.10) p. 251
Ashkiiádiiłtsą́
boy he-saw-himself
‘The boy saw himself’ (Platero 1978:108)
SOURCED Print notes (adds to T22e): (4.6a) prints the verb with a space, yi yiiłtsą́; (4.7a) has a hyphen, yi-yiiłtsą́; (4.9) has neither, yiyiiłtsą́; (4.8a) prints yiiłtsą́ with one yi. In (4.6b) the printed word-gloss line reads “girl boy bi saw” under ashkiiatʼééd, the reverse of the words; the gloss above follows the words. The translation (‘The girl saw the boy’) and Speas’s “OSV” fix the reading.
10.4 Derivations (schematic structure; only the Navajo strings are data)
DER-1 (4.6a) ashkii atʼééd yi yiiłtsą́ 'The boy saw the girl'
B Merge [dP d_obj [DP atʼééd]] with V; v; [dP d_subj [DP ashkii]] in Spec,vP ID-1
B Move none (yi- clause) ID-2
B order ashkii > atʼééd > [verb word] SOV ID-4
-- word boundary: the d heads and the spine heads go to rule set A --
A spell d_obj: DP_obj does not c-command DP_subj → yi- ID-3
NP-PRO no null DP: vacuous ✓ as printed
DER-2 (4.6b) ashkii atʼééd b-iiłtsą́ 'The girl saw the boy'
B Merge [dP d_obj [DP ashkii]] with V; v; [dP d_subj [DP atʼééd]] in Spec,vP
B Move DP ashkii → above DP atʼééd; d_obj stays in VP ID-2
B order ashkii > atʼééd > [verb word] OSV ID-4
A spell highest copy of DP_obj c-commands DP_subj → bi- ID-3
✓ as printed
DER-3 (4.8a) Atʼééd yiiłtsą́
parse P1 [pro_subj] > [DP_obj atʼééd] > V, no Move, yi-
pro precedes the noun: NP-PRO vacuous ✓ 'Sh/he saw the girl'
parse P2 [DP_subj atʼééd] > [pro_obj] > V, no Move, yi-
pro immediately follows atʼééd → must be coreferential ID-5
coreference without the reflexive is out ID-6
✗ NOT 'The girl saw him/her'
DER-4 (4.8b) Atʼééd b-iiłtsą́
parse P1 pro_obj moved: [pro_obj] > [DP_subj atʼééd] > V, bi- ID-2, ID-3
pro precedes the noun: NP-PRO vacuous ✓ 'The girl saw him/her'
parse P2 atʼééd moved: [DP_obj atʼééd] > [pro_subj] > V, bi-
pro immediately follows atʼééd → coreference forced ID-5
non-reflexive verb ID-6
✗ NOT 'Sh/he saw the girl'
DER-5 (4.9) Ashkii yiyiiłtsą́: as DER-3. 'boy saw him' = P2 ✗ with any index;
'He saw the boy' = P1 ✓ (Speas n. 5).
DER-6 (4.7a,b) no overt noun: both DPs are pro, NP-PRO never applies; yi-/bi- is fixed
by ID-2/ID-3 alone. (4.7b)'s "talking about him/her" fits a moved object DP
with topic-like status (HYPOTHESIS; A vs A′ landing site is T104a).
DER-7 (4.10) Ashkii ádiiłtsą́: reflexive. If there is a null object, P2's forced
coreference is exactly what the reflexive allows (ID-6 satisfied). Speas n. 4
leaves open that the morphology removes the object role instead (no DP_obj).
10.5 What this shows and what it does not
HYPOTHESISCoverage. ID-1–ID-6 give all six printed readings of (4.6)–(4.9) and exclude both NOT readings of (4.8) and the starred reading of (4.9). The lone-noun rule FB-3a(b) of §6 is not needed: under bi- the lone noun is the agent because the moved object DP, null or not, precedes it, and a null DP cannot follow an overt one with a different referent. The price is that null arguments must be real DPs (pro) in the syntax, which FB does not need.
HYPOTHESISThe DP-movement half does work; the stationary-d half is not tested. In every derivation above the object prefix is pronounced in the same slot of the word whether its DP moved or not. A moving-d version (d_obj head-moves into the word, its DP moves separately) gives the same strings and the same readings. So these sentences, like the prefix tables (testability report, 2026-11-05), do not show that d stays; they show that a DP moves in bi- clauses, given ID-4.
HYPOTHESISCorrection to older text. The light-determiner essay’s analysis D says the complement noun “is stranded in argument position, so no extra DP movement is needed.” For bi- clauses that is too strong: with nouns stranded in their base positions, OSV (4.6b) and the NOT reading of (4.8b) need the object DP above the subject DP, i.e. the same movement as ID-2, unless the inverse is base-generated with the object high (OPEN, T104a). The essay text is left unchanged in this run; this note supersedes that sentence for bi- clauses.
SOURCEDLimit. Speas p. 251: Platero himself showed that NP-PRO is inadequate for more complex sentences, especially those where one argument is a relative clause. ID-5 is therefore only established here for simple clauses (T104c).
GRANT'S CONJECTURE The light-determiner structure itself (ID-1) remains Grant’s conjecture; nothing in Speas states it.
ID
Question
What would settle it
T104a
In a bi- clause, is the object DP moved above the subject (and to an A or A′ position), or base-generated high? (Same question as T53a, now with ID-1’s DP-only movement.)
Reconstruction or binding connectivity under bi-: can the fronted object’s possessor or quantifier be read low? Saxon 2003 / Willie 1991 data, or speaker judgments.
T104b
Locality of ID-3: under stationary d, d_obj is spelled bi- although its DP has left. Is bi- the exponent of Agree between a probe above vP and the moved DP (with d_obj as the place it is pronounced), or chain-sensitive vocabulary insertion? (Deep-or-shallow found no person-hierarchy, shallow side.)
A case where the moved DP and d_obj could disagree in features, e.g. number with da-, or a possessed object; theory comparison with Oxford 2023’s deep inverse.
T104c
Can ID-5, stated over DP positions, be extended to Platero’s relative-clause paradigm (4.15) the way LINK-NEAR (§6) was, or do the relative-clause counterexamples Speas mentions on p. 251 break it?
Re-run SP3/SP4 (§8) with ID-1–ID-6; Speas pp. 254–269 (on disk, page images). Answered 2026-11-09 in §11: not by itself; fails on (4.15h).
11. Update 2026-11-09 (morning): Platero’s relative-clause paradigm (4.15) re-run with Grant’s light determiners — NP-PRO over DP positions fails on exactly one good reading, (4.15h), and parallelism has to come back (T104c)
HYPOTHESIS In plain words: §10 showed that “d stays, only the noun phrase moves”, plus Platero’s rule that an unpronounced pronoun right after a noun phrase means the same thing as that noun phrase, gets every reading of Speas’s simple sentences. This section runs the same rules on Platero’s harder sentence hastiin łį́į́ʼ yizloh-ę́ę yíʼdiiłid, which has fourteen possible readings, five of them good. The rules get twelve of the fourteen right. They wrongly throw out the good reading ‘The man who roped the horse branded it’ (h), which Speas herself names as a violation of the same rule (p. 257), and on their own they wrongly let in ‘It branded the man that roped the horse’ (d). Fixing (h) needs Speas’s subject-with-subject, object-with-object parallelism, restated below over the d positions; fixing (d) needs the rank rule or Condition C. All of these sentences use yi- verbs, so the inverse movement of §10 never applies, and whether d stays or moves makes no difference anywhere.
Notation as in §10: B = clause syntax (rule set B, right-headed); A = word spell-out (rule set A, the left-headed prefix string); the word boundary is the switch from B to A. Merge builds structure, Move is internal Merge, Agree only values features. Structures are schematic; only the Navajo strings and the judgments are data.
11.1 Sources opened this run
SOURCED Speas 1986, page images of the box copy (not the text layer): p. 251 (the NP-PRO limit and n. 6, yi- forms only, SOV basic), pp. 254–257 ((4.15) i–iv with readings a–n, n. 11, (4.16), structures (4.17) and (4.18) with her problem statements), pp. 264–269 (Constraint on ATB Interpretation, NP-PRO restated as a c-command control rule, (4.26)–(4.40), the reason-by-reading list (4.38)).
Not opened: Platero 1974, 1978, 1982 (cited through Speas only). Site rules reused, not re-derived: ID-1–ID-6 (§10), the rank rule FB-4 (§6; eight ranks), the FB analysis of the same paradigm (§5.2, §8 SP3/SP4).
11.2 Data (Speas 1986, pp. 254–255, 268–269)
The string, as printed on p. 254 (Speas’s word glosses in Leipzig layout; she segments only -ę́ę):
SOURCED Speas’s four parses and fourteen readings (p. 255), with her reasons (pp. 268–269). Five readings are good: a, b, c, e, h ((4.16)). n. 11: readings where subject and object of one verb corefer are left out, “since the verb I am using is not marked as reflexive.” In (4.38) on p. 268 the same sentence is typed without tone marks, nasal hooks or barred l (hastiin liiʼ yizloh-ee yiʼdiilid); the p. 254 spelling is used here (T22e).
Parse (p. 255)
Reading
Judg.
Speas’s reason (pp. 268–269)
i) pro [NP[S man horse roped] REL] branded
a) ‘Hei branded the horse that the mani roped’
✓
ATB representation (4.31)
b) ‘He/shej branded the horse that the mani roped’
✓
plain parse (4.33)–(4.34)
c) ‘He/shej branded the man that roped the horsei’
✓
plain parse (4.35)
d) *‘Iti branded the man that roped the horsei’
*
Condition C; not generable in ATB format
ii) man [[pro horse roped] REL] branded
e) ‘The mani branded the horse hei roped’
✓
NP-PRO coindexes pro with man (4.36)–(4.37)
f) *‘The mani branded the horse he/shej roped’
*
NP-PRO
g) *‘The mani branded the onej who roped the horse’
*
NP-PRO
iii) [[man horse roped] REL] pro branded
h) ‘The man who roped the horsei branded iti’
✓
ATB representation (4.30)
i) *‘The man who roped the horsei branded him/herj’
*
NP-PRO; not generable in ATB format
j) *‘The horse that the mani roped branded himj’
*
NP-PRO; not generable in ATB format
k) *‘The horse that the mani roped branded himi’
*
NP-PRO; not generable in ATB format
iv) *man [[horse pro roped] REL] branded
l) *‘The mani branded the horse that roped himi’
*
NP-PRO in all interpretations
m) *‘The mani branded the horse that roped him/herj’
*
n) *‘The mani branded the onej that the horse roped’
*
SOURCED Speas on parse iii, (4.18), p. 257: “Why is a null object allowed here, in violation of the NP-PRO constraint? That is, what accounts for the grammaticality of (4.15.h)”. On p. 264 she restates NP-PRO as “a completely general control rule, which always coindexes an empty pronominal with a c-commanding overt NP”, which rules out parses iii and iv in plain format ((4.26), (4.27)); (4.15h) and (4.15a) come back only through the ATB format (4.30), (4.31) (pp. 265–266). Constraint on ATB Interpretation (p. 264): “In a parallel structure P, made up of component sentences S1 and S2, the subject of S1 and the subject of S2 must be coreferential, and the object of S1 and the object of S2 must be coreferential.” Head construal (p. 265): if the relative’s subject is its head, the relative is the matrix subject; if its object is the head, the relative is the matrix object.
11.3 Rules first
ID-1–ID-6 are as in §10 and are not changed. Two consequences for this sentence, then one new rule.
ID-4′ YI- CLAUSES KEEP BASE ORDER. Both verbs are yi- forms (Speas p. 251 n. 6: SOURCED (forms, n. 6)
SOV basic). By ID-2 no DP moves over another in either clause, so by ID-4 HYPOTHESIS (consequence)
the earlier DP of a clause is its subject. In parse iv, horse is therefore the
subject of 'roped' and pro its object.
ID-5 (unchanged) NP-PRO, applied to the string of DP positions. A relative clause SOURCED (rule; p. 250)
is an overt DP (DP_rel), so a pro right after DP_rel must corefer with DP_rel. HYPOTHESIS (DP_rel counts)
Same verdicts as Speas's c-command version (pp. 264,
268–269) on every parse of (4.15).
ID-7 PARALLEL DERIVATION (restates Speas's Constraint on ATB Interpretation, p. 264, SOURCED (Speas's constraint)
and her head construal, p. 265, over light-determiner positions). HYPOTHESIS (restatement)
A string with a relative clause may also be derived as two components: S2 =
the relative clause, S1 = the matrix clause with DP_rel replaced by pro.
(a) The complement of d_subj in S1 corefers with the complement of d_subj in
S2; likewise for d_obj.
(b) The DP of S2 construed as head fixes the function of the relative in S1:
head = complement of d_subj in S2 → relative is S1's subject; d_obj → object.
(c) ID-5 and ID-6 are checked inside each component, on its own DP string.
A reading is good if a plain derivation or a parallel derivation converges.
RANK The rank rule (FB-4 in §6; HYPOTHESIS there) or Speas's Condition C, needed HYPOTHESIS / SOURCED
for (d) only. Which of the two does it is T22b, not decided here. (Speas p. 268)
11.4 Derivations (schematic structure; only the Navajo string is data)
Common steps for every parse (both verbs):
B Merge [dP d_obj DP] with V; v; [dP d_subj DP] in Spec,vP ID-1
B Move none: both verbs are yi- forms, no inverse ID-2, ID-4′
B REL -ę́ę closes the inner clause; the result is DP_rel, the complement of
one matrix d (internally headed relative; head = man or horse)
-- word boundary: d heads and spine heads go to rule set A --
A spell d_obj of both verbs: DP_obj does not c-command DP_subj → yi- ID-3
A words yizloh-ę́ę, yíʼdiiłid (as printed, p. 254). d stays or moves: same strings.
REL-1 parse i pro_subj > [DP_rel man > horse > roped-REL] > branded
plain pro comes first; inside DP_rel there is no pro: ID-5 vacuous
head horse, pro free → (b) ✓; head man → (c) ✓
head horse, pro = man → (a) ✓ if no Condition C is assumed
head man, pro = horse → (d) generated ✗ WRONG without RANK
parallel S1: pro_i pro_j branded | S2: man_i horse_j roped; head = horse (d_obj)
→ relative = S1 object → (a) ✓ (this is Speas's (4.31))
RANK (d): horse as subject over a man under yi- → ✗ (or Condition C, Speas)
REL-2 parse ii DP man > [DP_rel pro > horse > roped-REL] > branded
plain pro immediately follows man → pro = man ID-5
→ (e) ✓; (f) pro ≠ man ✗; (g) head = pro ≠ man ✗
parallel S1: man pro branded: pro follows man → pro = man, so man branded ID-5, ID-6
himself without the reflexive ✗ (Speas's (4.32) fails too)
REL-3 parse iii [DP_rel man > horse > roped-REL] > pro_obj > branded
plain pro immediately follows DP_rel → pro = DP_rel, i.e. the subject of ID-5
'branded' and its object corefer without the reflexive ID-6
→ (h) ✗ WRONG; (i), (j), (k) ✗ correctly
parallel head = man (d_subj in S2) → relative = S1 subject ID-7(b)
S1: pro_i pro_j branded | S2: man_i horse_j roped
ID-5 vacuous in S1 (no overt DP) and in S2 (no pro) ID-7(c)
→ (h) ✓ (Speas's (4.30)); (i) pro ≠ horse breaks ID-7(a) ✗
(j), (k): head horse would make the relative S1's object, not its
subject → no parallel derivation of these readings ✗
REL-4 parse iv DP man > [DP_rel horse > pro > roped-REL] > branded
horse is subject of 'roped', pro its object ID-4′
plain pro immediately follows horse → pro = horse: the horse roped itself ID-5
without the reflexive → ✗ for (l), (m), (n) alike ID-6
parallel S1 subject man must corefer with S2 subject horse → ✗ ID-7(a)
11.5 What this shows and what it does not
Rule set
Good readings (a b c e h)
Bad readings (d f g i j k l m n)
Errors
ID-1–ID-6 only (no Condition C)
a b c e ✓; h ✗
f g i j k l m n ✗; d ✓
2 of 14: h out, d in
+ ID-7 (parallel)
all 5 ✓
8 ✗; d ✓
1 of 14: d in
+ ID-7 + RANK (or Condition C)
all 5 ✓
all 9 ✗
0 of 14
HYPOTHESIST104c answered: no, not by itself. ID-5 stated over DP positions handles parses ii and iv completely and is idle in parse i, but in parse iii it forces the matrix null object to corefer with the relative-clause DP and so excludes the good reading (4.15h). This is the same failure Speas names for Platero’s rule (p. 257), so moving the rule onto light-determiner positions does not repair it. With ID-7 added, and rank (or Condition C) for (d), all 14 judgments come out as printed.
HYPOTHESISThe cost. Because ID-1 makes null arguments real DPs (pro), the light-determiner analysis has nothing like FB-5’s “no Condition C from a bundle” (§6). If standard Condition C applies to pro, (4.15a) is out on the plain parse too and survives only through ID-7. On this paradigm the light-determiner analysis therefore needs the same two devices Speas uses, NP-PRO and parallelism. FB needed LINK-NEAR and rank instead (§5.2). Neither of these results says anything for or against nouns being outside the verb word.
HYPOTHESISWhere the d positions help, and where they do not. ID-7(a) has a natural statement here: the arguments that must corefer across the two clauses are the complements of the same light determiner (d_subj with d_subj, d_obj with d_obj). Each d’s merge site records the function whether or not the DP later moves. But a moving-d analysis keeps a copy of d at the same site and can state ID-7 the same way. So (4.15) neither supports nor weakens “d stays”. Correction 2026-11-09 (afternoon): the merge-site statement is wrong for bi- clauses; parallelism has to follow the moved DPs (ID-7′, §12). And since there is no bi- verb in the paradigm, the DP movement of §10 (ID-2) is never used; what matters is that it is absent, because that fixes łį́į́ʼ as the subject in parse iv (ID-4′).
SOURCED(4.39), no overt nouns (p. 269): Yizloh-ee yiʼdiilid (typed without tone marks or barred l there) ‘S/he branded the one that s/hei,?j roped’. Speas: the only parse obeying all the principles is the one with the relative as matrix object. HYPOTHESIS ID-5 gives the same result: with the relative as subject, the pro object comes right after an overt DP_rel and must corefer with it, as in REL-3. But ID-1–ID-7 give the plain object-relative parse free subject reference, so the “?” on the disjoint reading j is not derived (T105b).
GRANT'S CONJECTURE The light-determiner structure (ID-1) remains Grant’s conjecture. Speas states neither it nor ID-7 in this form.
ID
Question
What would settle it
T105a
Is ID-7 coindexing of two pro’s with the S2 DPs (Speas’s ATB format), or sharing of one DP as the complement of two d heads, one in each clause? Sharing would leave no matrix pro in (4.15a) and (4.15h) at all, so Condition C and NP-PRO would have nothing to apply to.
Condition C cases without a relative, (4.119)–(4.121) (pp. 319–320), and Hale 1999 (11a,b) (T24a); compare the predictions of the two versions.
T105b
Why is the disjoint subject reading of (4.39) Yizloh-ee yiʼdiilid marked “?j” (p. 269) when the plain parse leaves the subject free?
Platero 1978 originals (the source of the judgment); speaker judgments with context.
T105c
ID-5 is stated over string adjacency; Speas’s p. 264 version uses c-command. No datum on pp. 249–269 separates them. Does any?
A sentence where an overt DP comes just before a pro without c-commanding it, for example a possessor inside a DP ((4.49)–(4.50), pp. 278–279) or a -go clause ((4.58), p. 284).
T105d
With bi- relatives, ID-2 does apply. Do ID-2 + ID-7 derive (4.53)–(4.55) (pp. 281–283), where the parallelism matches positions, not roles?
Re-run §5.4 with ID-1–ID-7 on the page images pp. 281–283 (on disk). Answered 2026-11-09 in §12: ID-7 by merge site fails on (4.54); ID-7′ over moved DP positions works, plus ID-8 for (4.53b).
Sources: Speas 1986 (box page images, not committed), pp. 251, 254–257, 264–269. Platero 1978 (the paradigm’s source) cited through Speas, not opened.
12. Update 2026-11-09 (afternoon): bi- relatives (4.53)–(4.55) with Grant’s light determiners — parallelism must match the noun phrases’ positions after the inverse movement, not the d heads’ merge sites (T105d)
HYPOTHESIS In plain words: §11 restated Speas’s parallelism (“subject goes with subject, object with object”) over the places where the two light determiners are merged, and claimed this was a natural fit because d stays put. Speas’s bi- sentence (4.54), ‘The horse that was kicked by the mule bit it’, shows that this is wrong: the horse is the object of ‘kick’ but the subject of ‘bite’, so matching by merge site gives the wrong pairing. What has to match is where the noun phrases end up after the bi- movement of §10: the higher noun phrase in one clause goes with the higher one in the other. With that change (ID-7′) the rules still get all 14 judgments of (4.15), plus (4.54) and (4.55). One more datum, that a bi- relative cannot be headed by its subject (4.53b), needs a further rule (ID-8) that nothing so far derives. The stationary-d half of Grant’s proposal is again neither helped nor hurt; the DP-movement half is used and is now doing work in two places.
Notation as in §§10–11: B = clause syntax (rule set B, right-headed); A = word spell-out (rule set A, the left-headed prefix string); the word boundary is the switch from B to A. Merge builds structure, Move is internal Merge, Agree only values features. Structures are schematic; only the Navajo strings and the judgments are data.
12.1 Sources opened this run
SOURCED Speas 1986, page images of the box copy (not the text layer): pp. 280–286. That is: the Huang-style “closest nominal element” and cases b, c (p. 280); (4.52) and Platero’s second argument, “Headless relatives with bi-verbs, unlike those with yi-verbs, are unambiguous as to which NP is to be construed as the head” (p. 281); (4.53a,b) (pp. 281–282); (4.54) with “If we embed such a relative clause under another verb, we can get only the reading in which there is an empty object” (p. 282); (4.55) and (4.56) (p. 283); (4.57)–(4.59) (p. 284); Platero’s generalization “no matter how the sentence is parsed, the NP which is string-closest to the verbs must be the object of both verbs” (p. 285).
Not opened: Platero 1978, 1982 (cited through Speas only). Site rules reused, not re-derived: ID-1–ID-7 (§§10–11), FB-6 PAR (§6), the earlier FB reading of the same data (§5.4).
12.2 Data (Speas 1986, pp. 281–283)
Speas’s word glosses in Leipzig layout (she segments only yi-/bi- and -ę́ę); spelling as on the page images. The gloss of the -ztał verb varies in print (‘bit’ in (4.53a) and (4.54), ‘kick’ in (4.53b)); the translations all say ‘kick’ (T22e).
Ex. (page)
Navajo
Gloss
Translation and judgment as printed
(4.53a) p. 281
łį́į́ʼ dzaanééz yiztał-ę́ę
horse mule yi-bit-REL
‘The horse that kicked the mule’ OR ‘The mule that the horse kicked’
(4.53b) pp. 281–282
łį́į́ʼ dzaanééz biztał-ę́ę
horse mule bi-kick-REL
‘The horse that was kicked by the mule’; *‘The mule that the horse was kicked by’
(4.54) p. 282
łį́į́ʼ dzaanééz biztał-ę́ę yishxash
horse mule bi-bit-REL yi-bit
‘The horse that was kicked by the mulei bit iti’; *‘The horse bit the mule that it was kicked by’
(4.55) p. 283
*ashkiiiatʼééd yiyiiłtsanę́ę proiyiztsʼǫs
boy girl saw-REL kissed
*‘The girl who the boyi saw kissed himi’
SOURCED Speas on (4.54) (p. 282): it “violates the Huang-style account … since the nominal with which the pro ends up coindexed neither c-commands pro nor is string adjacent to pro”. On (4.55) (p. 283): if relatives were simply exempt from NP-PRO, coindexing pro with the non-head noun “should also be permitted”, but (4.55) is ungrammatical. Her language-particular repair (4.56)/(4.57) is, in her words, “completely ad-hoc” and leaves Platero’s problem unsolved (p. 284). On p. 285 she credits the ATB account of her section 2 with capturing Platero’s string-closest generalization.
12.3 Rules first
ID-1–ID-6 (§10) and ID-4′, ID-5 (§11) are unchanged. ID-7 is replaced; one new rule is added.
ID-7′ PARALLEL OVER DERIVED DP POSITIONS (replaces ID-7 of §11). SOURCED (Speas's constraint,
A string with a relative clause may also be derived as two components, p. 264; Platero's generalization,
S2 = the relative clause, S1 = the matrix with DP_rel replaced by pro. p. 285)
(a) Take each component's DPs after ID-2 (highest copies). The higher DP of HYPOTHESIS (restatement)
S1 corefers with the higher DP of S2; the lower with the lower.
(b) Head construal: if the head is the higher DP of S2, the relative is S1's
higher DP; if the lower, S1's lower DP.
(c) ID-5 and ID-6 are checked inside each component, as before.
In yi- clauses (no ID-2 movement) higher = complement of d_subj and lower =
complement of d_obj, so ID-7′ and ID-7 give the same results on (4.15).
ID-8 HEAD ACCESS. The subject DP of a bi- relative (the DP crossed by ID-2 movement) SOURCED (datum (4.53b),
cannot be construed as the head. In a yi- relative either DP can be the head. p. 281: yi- ambiguous, bi- not)
HYPOTHESIS (the rule; why: T106a)
Correction 2026-11-09 (evening): ID-8 is too strong. It holds only when both DPs of the bi- relative are overt nouns; with a null higher DP the crossed subject can head (Willie 1991 (15), (19), (30); Hale & Platero 1974 (42)). Replaced by ID-8′, §13. REL-6 stands under ID-8′(b); the star in REL-8 is not derived by head access, see §13.6.
12.4 Derivations (schematic structure; only the Navajo strings are data)
REL-5 (4.53a) łį́į́ʼ dzaanééz yiztał-ę́ę
B Merge [dP d_obj [DP dzaanééz]] with V; v; [dP d_subj [DP łį́į́ʼ]] in Spec,vP ID-1
B Move none (yi-): horse is subject, mule object ID-2, ID-4′
B REL -ę́ę: head = either DP (ID-8 is idle without ID-2 movement)
A spell d_obj: its DP does not c-command DP_subj → yi- ID-3
head horse → 'the horse that kicked the mule' ✓
head mule → 'the mule that the horse kicked' ✓ both readings, as printed
REL-6 (4.53b) łį́į́ʼ dzaanééz biztał-ę́ę
B Merge [dP d_obj [DP łį́į́ʼ]] with V; v; [dP d_subj [DP dzaanééz]] in Spec,vP
B Move DP łį́į́ʼ → above DP dzaanééz; d_obj stays in VP ID-2
B REL head = łį́į́ʼ (moved object) → 'the horse that was kicked by the mule' ✓
head = dzaanééz (crossed subject) → blocked ID-8
A spell highest copy of DP_obj c-commands DP_subj → bi- ID-3
✓ as printed, * as printed
REL-7 (4.54) łį́į́ʼ dzaanééz biztał-ę́ę yishxash, good reading
parse [DP_rel łį́į́ʼ dzaanééz biztał-ę́ę] > pro_obj > yishxash
plain pro immediately follows DP_rel → pro = DP_rel (the horse): the horse ID-5
bit itself without the reflexive ✗ ID-6
parallel S2: łį́į́ʼ (higher, moved object of 'kick') > dzaanééz (lower, subject) ID-2
S1: pro_high (subject of 'bite', yi-, no move) > pro_low (object)
ID-7 (old, by merge site) S1 d_subj ↔ S2 d_subj: biter = kicker = mule;
head łį́į́ʼ is S2's d_obj complement → relative = S1 object (ID-7(b));
but DP_rel precedes pro in a yi- clause, so it is S1's subject (ID-4′)
→ no parallel derivation; the printed good reading is lost ✗ WRONG
ID-7′ S1 higher ↔ S2 higher: biter = horse; S1 lower ↔ S2 lower: bitten = mule
head łį́į́ʼ = S2 higher → relative = S1 higher (subject of 'bite') ID-7′(b)
→ 'The horse that was kicked by the mule_i bit it_i' ✓
A spell 'kick': bi- (ID-3, object DP high); 'bite': yi- (no move) ✓ yishxash as printed
REL-8 (4.54) starred reading *'The horse bit the mule that it was kicked by'
parse DP łį́į́ʼ > [DP_rel pro dzaanééz biztał-ę́ę] > yishxash,
pro = moved object of 'kick' (= the horse), dzaanééz = subject of 'kick', head
plain pro follows łį́į́ʼ → pro = horse (ID-5), different clauses (ID-6 idle)
head = dzaanééz, the crossed subject of a bi- relative → blocked ID-8 ✗
parallel S1: łį́į́ʼ (higher) > pro (lower) | S2: pro (higher) > dzaanééz (lower);
the head would again be the crossed subject ID-8 ✗
✗ as printed
REL-9 (4.55) *ashkii atʼééd yiyiiłtsanę́ę pro yiztsʼǫs 'The girl who the boy_i saw kissed him_i'
parse [DP_rel ashkii atʼééd yiyiiłtsanę́ę] > pro_obj > yiztsʼǫs, head = atʼééd
plain pro immediately follows DP_rel → pro = the girl: girl kissed herself ID-5, ID-6 ✗
parallel head atʼééd is S2's lower DP → relative must be S1's lower DP (object),
but in this reading the relative is the kisser (S1 higher) ID-7′(b) ✗
(ID-7 by merge site gives the same ✗: both verbs are yi- forms)
✗ as printed
SOURCED (4.52) Asdzą́ą́ Kii yaʼníłtsoodę́ę yichʼahóóshkeed (p. 281) is not re-derived here: Speas glosses both verbs “3-…-3” without marking yi- or bi-, and its verdict depends on a separate fact Speas reports, that proper names cannot head relative clauses. HYPOTHESIS If both verbs are of the yi- type, ID-7 and ID-7′ agree on it and it comes out as in §8 SP5.
12.5 What this shows and what it does not
HYPOTHESIST105d answered: yes, but only after ID-7 is restated. ID-2 + ID-7 (merge-site parallelism, §11) derive the wrong pairing for (4.54). ID-2 + ID-7′ (parallelism over DP positions after the inverse movement) derive the good reading of (4.54) and exclude (4.55); ID-8 is needed for (4.53b) and also excludes the starred reading of (4.54). On (4.15), which has only yi- verbs, ID-7′ and ID-7 coincide, so §11’s 14/14 is unchanged.
HYPOTHESISCorrection to §11. §11.5 said ID-7 has “a natural statement” over light-determiner positions because “each d’s merge site records the function whether or not the DP later moves”. (4.54) shows that this is the wrong thing to record: the pairing follows the moved DP, not the d it was merged with. This is what FB-6 PAR (§6) already said for bundles (“after voice”), and it fits Platero’s string-closest generalization (p. 285): in (4.54) the string-closest noun, dzaanééz, is the kicker, so “object of both verbs” there must mean the lower position after bi-, not the object role.
HYPOTHESISWhat it says about Grant’s proposal. The “DP moves” half now does work twice: it fixes bi- (ID-3) and it fixes which DPs are parallel (ID-7′). The “d stays” half is still not tested: a moving-d analysis whose DPs move the same way gives the same strings and the same pairings. If anything, the d positions themselves turn out to be the wrong place to state parallelism. So the inverse-moved object behaves like a subject for parallelism and for heading a relative, which is some support for an A-type landing site (T104a), not proof.
OPENPossible overgeneration in (4.54). Speas says embedding the bi- relative gives “only the reading in which there is an empty object” (p. 282). ID-1–ID-8 also allow a matrix null subject with the relative as object, ‘S/he bit the horse that the mule kicked’, just as (4.15b) is good with yi-. Speas stars only one reading explicitly, so it is unclear whether this one is meant to be out (T106b).
GRANT'S CONJECTURE The light-determiner structure (ID-1) remains Grant’s conjecture. Speas states neither it, ID-7′ nor ID-8 in this form.
ID
Question
What would settle it
T106a
Why can’t the crossed subject head a bi- relative (ID-8)? Is it a locality effect (the moved object DP intervenes), a freezing effect, or does relativization of the subject require the subject to be the highest DP?
bi- relatives in other sources (Platero 1978; Willie 1991; Hale & Platero 1974 on the box), and whether a yi-/bi- relative with a third participant shows the same block. Answered 2026-11-09 (evening) in §13: none of the three as stated; ID-8 is too strong (Willie (15), (19), (30); H&P (42)) and holds only when both DPs are overt (ID-8′(b)); why overtness matters is T107c.
T106b
Is ‘S/he bit the horse that the mule kicked’ (matrix pro subject, the (4.54) relative as object) really out, as “only” on p. 282 suggests? If so, what excludes it, given that (4.15b) is good?
Platero 1978 originals; speaker judgments with context. Narrowed 2026-11-09 (evening) in §13: would follow if parallelism is obligatory here (Willie (27)); open as T107a.
T106c
ID-7′ predicts that in a yi- relative embedded under a bi- matrix verb, the relative’s subject is parallel to the matrix’s moved object. Is that right?
A sentence of that shape from a source (none on Speas pp. 249–286), or speaker judgments. Answered 2026-11-09 (evening) in §13: yes for Willie’s speech, (29) p. 166; merge-site ID-7 gives the meaning of (28) instead.
Sources: Speas 1986 (box page images, not committed), pp. 280–286. Platero 1978, 1982 cited through Speas, not opened.
13. Update 2026-11-09 (evening): which noun phrase heads a relative clause — ID-8 is too strong; the crossed subject of a bi- relative can head when the moved object is not an overt noun (T106a, T106c)
HYPOTHESIS In plain words: §12 added a rule (ID-8) saying that the subject a bi- object has jumped over can never be what a bi- relative clause is about. Two more sources on the box show that this is true only when both participants are named by nouns inside the relative clause. With no nouns at all, Willie’s bi- relatives are about exactly that jumped-over subject (‘the one who kicked him’), and Hale & Platero print a sentence, ‘The boy shot the snake he was bitten by’, that they analyse the same way. So the rule is about how many nouns are spelled out, not about the inverse movement as such. Willie’s data also confirm the prediction left open in §12 (T106c): when a yi- relative sits under a bi- main verb, the participants pair up by position after the inverse movement (ID-7′), not by the d heads they were merged with. As before, nothing here tells “d stays” from “d moves”.
Notation as in §§10–12: B = clause syntax (rule set B, right-headed); A = word spell-out (rule set A, the left-headed prefix string); the word boundary is the switch from B to A. “Higher” and “lower” DP mean the positions after ID-2 (the bi- movement of the object DP over the subject DP). Structures are schematic; only the Navajo strings, glosses and judgments are data.
13.1 Sources opened this run
SOURCED Willie 1991, Ch. 6, page images of the box copy, printed pp. 153–155 and 159–168: (2) and the claim that the relativizing enclitic is the head and “binds a pronoun that is incorporated into the relativized verb” (p. 153); (13)–(15) (p. 159); generalization (16), “The heads of relative clauses are absolutives”, and “with the Inverse voice, the agent is the internal argument” (p. 160); (18)–(20a) (p. 161); (20b), (21), the note on Platero 1982, and (22) (p. 162); (24) with its explanation (p. 164); (25)–(26) (p. 165); Parallel Processing (27) and (28)–(30) (p. 166); (31) and the yi-/bi- statement (p. 167); (32)–(35) (p. 168).
Reused, not re-read: Speas 1986 (4.53)–(4.55) and p. 282 (§12); H&P (17) (p. 15, read 2026-10-15); Platero 1978/1982 still not opened. The same Willie and H&P data were read in the feature-bundle framework on this topic’s relative-clause sections; here they are re-run with ID-1–ID-7′.
13.2 Data
Glosses are the authors’ own, in Leipzig layout (INV = inverse bi-; 3.3 = third-person object and subject; REL = relativizer). Spelling as on the page images, including Willie’s plain l in yiyiiltsą́ (30), (32) and biiltsą́ (33), and H&P’s tlʼiish and yił adeesdǫǫh (38).
Ex. (page)
Navajo
Gloss
Translation and judgment as printed
W (14) p. 159
yizloh=ę́ę
3.3.roped=REL
‘the one that was roped’
W (15) p. 159
bizloh=ę́ę
INV3.3.roped=REL
‘the one who roped him’
W (18) p. 161
yiztsʼǫs-yę́ę yóóʼeelwod
3.3.kissed=REL 3.ran.away
‘The one he kissed ran away.’ (“patient-headed”)
W (19) p. 161
biztsʼǫs=yę́ę yóóʼeelwod
INV3.3.kissed=REL 3-ran.away
‘The one who kissed him ran away.’ (“agent-headed”)
W (20a) p. 161
ʼatʼééd ʼashkii yiztsʼǫs=yę́ę yóóʼeelwod
girl boy 3.3.kissed=REL 3.ran.away
‘The girl who kissed the/a boy ran away.’
W (20b) p. 162
ʼatʼééd ʼashkii biztsʼǫs=yę́ę yóóʼeelwod
girl boy INV3.3.kissed=REL 3.ran.away
‘The girl who was kissed by the boy ran away.’
W (22) p. 162
ʼashkii yiztsʼǫs=yę́ę yóóʼeelwcd [sic]
boy 3.3.kissed=REL 3.ran.away
a. ‘The one who kissed the boy ran away.’ b. ‘The boy, who was kissed by her, ran away.’
W (24) p. 164
ʼashkii biztsʼǫs=yę́ę yóóʼeelwod
boy INV3.3.kissed=REL 3.ran.away
a. ‘The boy, the one who kissed her, ran away.’ b. *‘The one who was kissed by the boy ran away.’
W (28) p. 166
yiztał-yę́ę yiyiiłtsą́
3.3.kicked=REL 3.3.saw
‘He saw the one he kicked.’
W (29) p. 166
yiztał=yę́ę biiłtsą́
3.3.kicked=REL INV3.3.saw
‘The one he kicked saw him.’
W (30) p. 166
biztał=yę́ę yiyiiltsą́
INV3.3.kicked=REL 3.3.saw
‘He saw the one that kicked him.’
W (31) p. 167
biztał=yę́ę biiłtsą́
INV3.3.kicked=REL INV3.3.saw
‘The one that kicked him saw him.’
H&P (20) p. 15
Ashkii biiłtsą́-(n)ę́e atʼééd yidloh.
boy INV.saw-REL girl laugh
‘The girl that was seen by the boy is laughing.’
H&P (38) p. 19
Ashkii tlʼiish bishxash-ę́ę yił adeesdǫǫh.
boy snake bit-REL shot
‘The boy shot the snake he was bitten by.’ Source (39), p. 20: ashkii [ashkii tlʼiish bishxash-ę́ę] tlʼiish yiłʼadeesdǫǫh, both lower copies deleted.
some speakers (42): ‘When the girl yelled out, the boy kicked the dog that bit her.’ Others, only (43): ‘… the dog that bit him.’ (pp. 20–21)
SOURCED Willie on (24) (p. 164): “The Inverse relative with one nominal has only one reading”; she derives the star on (24b) from definiteness: “the Inverse voice requires a definite subject, which cannot be subjected to further definiteness marking by means of a restrictive relative.” On (20) (p. 162): for some speakers “(see Platero 1982) Ex. (20) has two readings, where either argument may be the head; this is not true of my speech.” On (29)–(31) (p. 167): “If yi- is present, the head of the relative is an adjunct to the patient argument of the main clause verb; if bi- is present the relative is an adjunct to the agent argument”; with an overt noun, as in (32)–(35), (27) “does not apply, since the noun marks disjoint reference”.
13.3 Rules first
ID-1–ID-6, ID-4′, ID-5 and ID-7′ are unchanged. ID-8 (§12) is replaced. “Overt” means that the DP is spelled out by a noun inside the relative clause; “null” means pro.
ID-8′ HEAD CHOICE (replaces ID-8). Take the relative's DPs after ID-2. SOURCED (each cell's data,
(a) DEFAULT. If at least one DP is null, the lower DP can be the head: in see 13.4 table)
yi- the object, in bi- the crossed subject (Willie's "internal HYPOTHESIS (the rule)
argument", p. 160).
(b) BOTH OVERT. If both DPs are overt, the higher DP can be the head. The
lower one cannot in bi- (Speas (4.53b); Willie (20b), (21)). In yi- it
can for Platero's and H&P's speakers ((4.53a); H&P (17)), not for
Willie ((21), "not true of my speech").
(c) LOWER OVERT, HIGHER NULL. yi-: the null higher DP can head (W (22a)).
bi-: it cannot head when there is no head noun after the relative
(W (24b)); with a head noun after it, H&P (20) is good (T107b).
ID-8 of §12 is the bi- half of (b) only.
[Note 2026-11-10: the bi- half of (c) is refined as ID-8′(c′) in §14:
H&P (20) has an external head noun, which Willie's grammar lacks.]
ID-5° NP-PRO AS A DEFAULT. ID-5 holds for H&P's strict speakers ((43)); lenient SOURCED (H&P pp. 20–21, 23)
speakers let context override it ((42)). Same status as LINK-NEAR, HYPOTHESIS (extension to ID-5)
FB-3a(a′) on the H&P section of this topic.
13.4 Which DP heads: the data by cell
Nouns inside the relative
Voice
Lower DP as head
Higher DP as head
none
yi-
✓ W (14), (18), (28), (29)
not printed (W: “patient-headed”, p. 161)
bi-
✓ (crossed subject) W (15), (19), (30), (31)
not printed (W: “agent-headed”, p. 161)
lower overt, higher null
yi-
✓ non-restrictive only, W (22b)
✓ W (22a)
bi-
✓ (crossed subject) H&P (38) as analysed in (39); (41) on reading (42); W (24a) non-restrictive
* W (24b); ✓ H&P (20), with a head noun after REL (T107b)
both overt
yi-
✓ Speas (4.53a), H&P (17); * for Willie ((21), p. 162)
✓ W (20a), (4.53a)
bi-
* Speas (4.53b); W (21)
✓ W (20b), (4.53b)
SOURCED Every ✓ and * in the table is a printed judgment or an author’s explicit statement; “not printed” cells are gaps, not judgments. The configuration “higher overt, lower null” is not printed in these pages.
13.5 Derivations (schematic structure; only the Navajo strings are data)
HC-1 W (19) biztsʼǫs=yę́ę yóóʼeelwod 'The one who kissed him ran away'
B Merge [dP d_obj [DP pro_j]] with V; v; [dP d_subj [DP pro_k]] in Spec,vP ID-1
B Move DP pro_j → above DP pro_k; d_obj stays in VP ID-2
B REL =yę́ę; head = lower DP pro_k (the kisser) ID-8′(a)
(ID-8 of §12 would block this: pro_k is the crossed subject) ✗ ID-8
A spell highest copy of DP_obj c-commands DP_subj → bi-; word biztsʼǫs=yę́ę ID-3
B matrix DP_rel = subject of yóóʼeelwod → 'the one who kissed him ran away' ✓ as printed
HC-2 H&P (38) Ashkii tlʼiish bishxash-ę́ę yił adeesdǫǫh, H&P's structure (39)
parse DP ashkii > [DP_rel pro_boy > tlʼiish bishxash-ę́ę] > yił adeesdǫǫh
B Merge inner: d_obj + DP pro (the bitten); d_subj + DP tlʼiish (the biter) ID-1
B Move DP pro → above DP tlʼiish (bi-, as printed) ID-2
B REL head = lower DP tlʼiish (crossed subject; higher DP null) ID-8′(a)
plain pro immediately follows ashkii → pro = the boy ID-5
A spell bi- (ID-3); yił adeesdǫǫh as printed
→ 'The boy shot the snake he was bitten by' ✓ as printed
Note: the ID-8-respecting parse [DP_rel ashkii tlʼiish bishxash-ę́ę] > pro_obj >
yił adeesdǫǫh (head = ashkii) has the same truth conditions, so (38) alone does
not decide; (42) does.
HC-3 H&P (41), lenient reading (42)
parse [ADV atʼééd hadoolghaazh-go] > DP ashkii > [DP_rel pro_girl > łééchąąʼí bishxash-ę́ę] > yiztał
B REL head = lower DP łééchąąʼí (crossed subject; higher DP null) ID-8′(a)
plain pro_girl follows ashkii; ID-5 would force pro = the boy ID-5
lenient speakers: context overrides ID-5 → pro = the girl ID-5°
→ 'the boy kicked the dog that bit her' (42) ✓ for those speakers
strict speakers: ID-5 applies → 'the dog that bit him' (43) ✓
No head other than the dog gives "the boy kicked the dog" with the girl bitten:
the kicked one is the biter, i.e. the crossed subject of bishxash. So for
lenient speakers ID-8 is contradicted, not merely avoided.
HC-4 Speas (4.53b) łį́į́ʼ dzaanééz biztał-ę́ę: both overt
B REL head = łį́į́ʼ (higher) ✓; head = dzaanééz (lower) blocked ID-8′(b)
→ same verdicts as REL-6 in §12
HC-5 W (29) yiztał=yę́ę biiłtsą́ 'The one he kicked saw him' (T106c: yi- relative
under a bi- matrix)
S2 (relative, yi-): higher = pro_k (kicker), lower = pro_j (kicked) = head ID-4′, ID-8′(a)
S1 (matrix, bi-): object DP (the seen) moves over the subject DP (the seer) ID-2
higher = the seen, lower = the seer
ID-7′ head = S2 lower → relative = S1 lower (the seer) ID-7′(b)
S1 higher (the seen) ↔ S2 higher (the kicker) → 'saw him (the kicker)' ID-7′(a)
→ 'The one he kicked saw him' ✓ as printed; Willie p. 167: with bi- the
relative is "an adjunct to the agent argument"
ID-7 (by merge site) S1 d_subj (seer) ↔ S2 d_subj (kicker): the kicker would
be the seer, and the relative (S2 d_obj) the seen → 'He saw the one he
kicked', the meaning of (28), not (29) ✗ WRONG
A spell yiztał (yi-), biiłtsą́ (bi-) as printed
HC-6 W (30) biztał=yę́ę yiyiiltsą́ 'He saw the one that kicked him'
S2 (bi-): higher = the kicked, lower = the kicker = head ID-2, ID-8′(a)
S1 (yi-): higher = seer (pro), lower = seen
ID-7′ relative = S1 lower (seen); seer = S2 higher (the kicked) ✓ as printed ID-7′
13.6 What this shows and what it does not
HYPOTHESIST106a answered: ID-8 is too strong, and none of the three structural explanations offered for it survives as stated. A freezing account (the crossed subject is frozen) and a “subject must be highest to head” account both predict that a bi- relative can never be about its crossed subject. Willie’s (15), (19), (30), (31) and H&P’s (42) show the opposite. A locality account (the moved object DP intervenes) predicts the same block whenever the moved DP is present; but under ID-1 the null object is a real DP (pro) in that position, and it does not block in (19) or (42). What remains is ID-8′(b): the block holds when both DPs are overt nouns. Whether that is syntax (overt and null DPs land differently) or interpretation (H&P’s surface principles (28), (44)) is OPEN (T107c).
HYPOTHESISConsequence for §12. REL-6 ((4.53b)) stands, now under ID-8′(b). REL-8 (the starred reading of (4.54), with the null higher DP inside the relative) is no longer excluded by head access: its head dzaanééz is the lower DP and the higher DP is pro, the configuration of H&P (38). But that starred reading (‘the horse bit the mule that it was kicked by’) has the same truth conditions as the good one (‘the horse that was kicked by the mule bit it’): in both, the horse bites the mule and the mule kicked the horse. So Speas’s star separates two structures, not two meanings, and H&P accept that structure for (38). No reading that differs in meaning is newly generated.
HYPOTHESIST106c answered for Willie’s speech: yes. In (29) the yi- relative’s subject pairs with the bi- matrix verb’s moved object, as ID-7′ predicts, and merge-site parallelism (ID-7) gives the meaning of (28) instead. This is a second datum, after (4.54), against stating parallelism over d merge sites. Willie’s own rule (27) says the same: “Voice alternations … change the relations between arguments and theta roles, but the same referents must be involved” (p. 166).
HYPOTHESIST106b narrowed, still open. Willie’s (30) shows a bi- relative as the object of a yi- verb whose null subject is the relative’s moved object (parallel). If Willie’s (27) is obligatory for a matrix with no overt noun (p. 167), the plain derivation of ‘S/he bit the horse that the mule kicked’ is blocked, and Speas’s “only” (p. 282) follows. But Platero’s (4.15b), a disjoint null matrix subject with a yi- relative, is good (§11), so for Platero’s speakers parallelism is not obligatory with yi-. Whether it is obligatory with bi- relatives is T107a.
HYPOTHESISWhat it says about Grant’s proposal. Head choice, like parallelism, follows the DP positions after the inverse movement: the default head is the lower DP, which is the d_obj complement in a yi- clause but the d_subj complement in a bi- clause. That is Willie’s “with the Inverse voice, the agent is the internal argument” restated as positions. The DP-movement half of the proposal now does work in three places (ID-3, ID-7′, ID-8′). The “d stays” half is still untested: a moving-d analysis with the same DP positions gives the same heads. Nothing here bears on whether nouns are inside or outside the verb word.
GRANT'S CONJECTURE ID-1 (light determiners taking DPs) remains Grant’s conjecture. Willie (pronominal arguments, nouns as adjuncts) and H&P (deletion and surface interpretation) state neither it nor ID-8′ in this form.
ID
Question
What would settle it
T107a
Is parallelism (ID-7′) obligatory for bi- relatives under a verb with no overt noun, as Willie’s (27) is for her speech, while optional with yi- for Platero’s speakers ((4.15b))? That would explain Speas’s “only” (p. 282, T106b).
Platero 1978/1982 originals; a bi- relative with a disjoint null matrix subject judged by Platero-type speakers.
T107b
H&P (20) (crossed subject overt, head = moved object, head noun after REL) is good, Willie (24b) (same without the noun after REL) is out. Is it the head noun, Willie’s definiteness account, or a dialect difference?
Willie on post-relative nouns with bi- relatives (her p. 158 notes a possible dialect difference); Platero 1974 IJAL. Answered 2026-11-10 in §14: the head noun, as an external head that Willie’s grammar lacks (p. 158, dialect difference); the definiteness account does not separate the two; open remainder T108a.
T107c
Why does overtness matter (ID-8′(b) vs (a))? Do overt DPs and pro land in different positions (e.g. only overt DPs move to a higher A-position), or is ID-8′ a surface-interpretation principle in H&P’s sense (50)?
Cases where an overt and a null DP differ in binding or scope inside a relative; Speas Ch. 4 on pro positions; speaker judgments.
Sources: Willie 1991, Ch. 6 (box page images, not committed), pp. 153–155, 159–168. Hale & Platero 1974 (box page images, not committed), pp. 15, 19–21. Speas 1986 pp. 281–283 as in §12. Platero 1978, 1982 cited through Speas and Willie, not opened.
14. Update 2026-11-10 (early morning): a noun after the relative marker — H&P’s (20) is a relative with an external head noun, which Willie says her speech does not have (T107b)
HYPOTHESIS In plain words: §13 left a clash. Hale & Platero’s (20), ‘The girl that was seen by the boy is laughing’, is good, but Willie’s (24b), the same kind of bi- relative meaning ‘the one who was kissed by the boy’, is out. The difference between them is the noun after the relative marker. In Hale & Platero’s analysis that noun (atʼééd ‘girl’) is the relative’s own head, sitting outside the relative clause. Willie says plainly (p. 158) that in her speech a noun after the relative verb is never part of the relative, and she calls this a possible dialect difference from Platero (1974). Willie’s definiteness explanation of (24b) would rule out (20) too, so it does not separate the two sentences. The likeliest answer is the one Willie herself names: a difference between speakers in whether a relative can have a head noun after it. As before, nothing here tells “d stays” from “d moves”.
Notation as in §§10–13: B = clause syntax (rule set B, right-headed); A = word spell-out (rule set A, the left-headed prefix string). “Higher” and “lower” DP mean positions after ID-2 (the bi- movement of the object DP over the subject DP). “External head” means a DP merged with the relative clause from outside it, after REL; “headless” means there is no such DP. Structures are schematic; only the Navajo strings, glosses and judgments are data.
14.1 Sources opened this run
SOURCED Willie 1991, Ch. 6, page images of the box copy, printed pp. 156–158 (not opened in §13), and p. 164 re-read: (7)–(9) and “the Direct/Inverse voice alternation of the main clause verb affects the theta role assigned to the external argument to which the relative is adjoined” (p. 156); (10), §6.2 “a nominal can be added to either a relative or main clause”, (11), (12) (p. 157); the statement on a following nominal, the note on Platero (1974), and the analysis of (12) (p. 158); (24) with its definiteness explanation (p. 164).
SOURCED Hale & Platero 1974, page images (PDF page = printed page + 3): head-final relatives with a following head noun, (1)–(2) (p. 9); the structure (19), constraint (18) and (20) (p. 15); parsing principle (44) (p. 21); (45)–(49) (p. 22); “not absolute” (p. 23). The box OCR text was used only to find pages; every form below is checked against the page images.
Not opened: Platero 1974 (IJAL), Platero 1978, 1982 (not on the box). Willie’s Ch. 3 on definite external arguments is cited through her p. 164 only.
14.2 Data
Glosses are the authors’ own, in Leipzig layout (INV = inverse bi-; 3.3 = third-person object and subject; REL = relativizer). Spelling as on the page images; apostrophes as U+02BC.
Ex. (page)
Navajo
Gloss
Translation and judgment as printed
W (11) p. 157
ʼatʼééd yidloh=ę́ę biztał
girl 3.laughing=REL INV3.3.kicked
‘The girl who was laughing kicked him.’
W (12) p. 157
yidloh=ę́ę ʼatʼééd biztał
3-laugh=REL girl INV 3-3-kicked
‘The girl kicked the one who was laughing.’
W (24) p. 164
ʼashkii biztsʼǫs=yę́ę yóóʼeelwod
boy INV3.3.kissed=REL 3.ran.away
a. ‘The boy, the one who kissed her, ran away.’ b. *‘The one who was kissed by the boy ran away.’
H&P (20) p. 15
Ashkii biiłtsą́-(n)ę́e atʼééd yidloh.
boy INV.saw-REL girl laugh
‘The girl that was seen by the boy is laughing.’ Source (19): [NP [S [Satʼééd ashkii biiłtsą́] REL -(n)ę́ę] [NPatʼééd]] yidloh, inner atʼééd deleted.
H&P (46) p. 22
Ashkii łééchąąʼí bishxash-ę́ę łį́į́ʼ yizloh.
boy dog bit-REL horse roped
Allowed (48): ‘The boy that was bitten by the dog roped the horse.’ Not allowed (49): ‘The boy roped the horse that was bitten by the dog.’
H&P (47) p. 22
Ashkii łééchąąʼí bishxash.
boy dog bit (no gloss line printed; words as glossed in (46))
‘The boy was bitten by the dog.’
SOURCED Willie (p. 157): “In (11) the nominal ʼatʼééd is an adjunct to the single argument of the relativized verb.” Then (pp. 157–158, the example meant is (12), though the page prints “(11)” again): “ʼatʼééd can not be an adjunct to the relativized verb since it follows it. This interpretation is counter to Platero (1974) who claims that a nominal following a relatived verb can be part of the relative clause; this may be a dialect difference.” On (12): “there are two separate adjuncts, one for each of the arguments (agent and patient) of the main clause verb. The relative is the (external) patient adjunct and the simple nominal is the (internal) agent adjunct because the main clause verb is an Inverse form” (p. 158). On (24) (p. 164): “Inverse clauses require definite external arguments … the Inverse voice requires a definite subject, which cannot be subjected to further definiteness marking by means of a restrictive relative.”
SOURCED Hale & Platero (p. 9): one Navajo relative type is “the type in which the head noun phrase follows the relative clause”. Parsing principle (44) (p. 21): “In a string Σ of the form (NP) NP V, the subsequence (NP) NP is a clausemate with V if the string Σ exists as a well-formed simple sentence of Navajo.” (46) “allows only interpretations according to which the noun phrases ashkii ‘boy’ and łééchąąʼí ‘dog’ are clause mates with the verb bishxash”, because (47) is a well-formed simple sentence (p. 22). (44) “is not absolute — it can be relaxed by some speakers” given strong context (p. 23).
14.3 Rules first
ID-1–ID-7′, ID-5° and ID-8′(a), (b) are unchanged. ID-8′(c) is split by whether there is an external head.
EXT-HEAD An overt DP right after REL can be merged as the external head SOURCED (H&P (19)–(20), p. 15;
of the relative: [DP [CP ... DP_j ... REL] DP_j], with the DP_j inside Willie p. 158 on Platero 1974)
the relative null (H&P: deleted under identity). HYPOTHESIS (restatement in
Grammar P (Platero 1974; H&P): available. DP terms)
Grammar W (Willie 1991): not available; a DP after REL belongs to the
main clause (W (12), p. 158).
In P, EXT-HEAD loses to H&P's parse (44) when the nouns before REL
already make a simple sentence with the relative verb ((46)–(49)).
ID-8′(c′) LOWER OVERT, HIGHER NULL, bi- (replaces the bi- half of (c)).
headless: the higher DP cannot head (W (24b)). SOURCED (W), untested for P
with EXT-HEAD (P): the higher DP is the relative's gap, linked to the SOURCED (H&P (20))
external head (H&P (20)).
The yi- half of (c) (W (22a)) is unchanged.
14.4 Derivations (schematic structure; only the Navajo strings are data)
EH-1 H&P (20) Ashkii biiłtsą́-(n)ę́e atʼééd yidloh, grammar P, H&P's structure (19)
B Merge [dP d_obj [DP_j girl]] with V; v; [dP d_subj [DP ashkii]] in Spec,vP ID-1
B Move DP_j → above DP ashkii; d_obj stays in VP ID-2
B REL -(n)ę́ę; gap = higher DP_j ID-8′(c′)
B Merge external head DP atʼééd_j with the relative EXT-HEAD
inner DP_j deleted (H&P: only NP1 of NP1 NP2 V may delete, (18))
[Note 2026-11-10 (§15): (18) is the constraint H&P argue against as a
condition on rules (p. 16); their analysis is raising (p. 28). The step
holds under RD-3 S2 or MATCH; only the attribution changes.]
A spell higher copy of DP_obj over DP_subj → bi-; word biiłtsą́-(n)ę́e ID-3
B matrix [DP relative + atʼééd] = subject of yidloh
→ 'The girl that was seen by the boy is laughing' ✓ as printed
EH-2 The same string in grammar W (prediction, not a printed judgment)
B atʼééd after REL is not in the relative DP; it is a matrix DP W p. 158, (12)
B relative ʼashkii biiłtsą́=ę́ę alone = the (24) configuration:
head = lower DP ashkii only (24a); higher DP as head * W (24b), ID-8′(c′)
(also excluded by Willie's definiteness account, p. 164)
B matrix yidloh has one argument; the relative and atʼééd both relate to it
→ the laugher is the boy and the girl at once: no reading
'the girl that was seen by the boy is laughing' PREDICTED (T108b)
EH-3 H&P (46) Ashkii łééchąąʼí bishxash-ę́ę łį́į́ʼ yizloh, grammar P
parse (44): ashkii łééchąąʼí bishxash = (47), a simple sentence → both nouns
are inside the relative H&P (44)
B REL both DPs overt → higher DP ashkii heads ID-8′(b)
B matrix łį́į́ʼ = object of yizloh → (48) 'The boy that was bitten by the dog
roped the horse' ✓ as printed
✗ (49) ashkii in the main clause, łį́į́ʼ as external head of
[łį́į́ʼ_j(null) łééchąąʼí bishxash-ę́ę]: the EH-1 configuration (higher
DP null, lower overt, external head). Excluded by (44), since (44)
puts ashkii inside the relative ✓ not allowed, as printed
14.5 What this shows and what it does not
HYPOTHESIST107b answered as far as these sources go: it is the head noun after REL, and Willie’s own words make that a dialect difference. The question offered three options. (i) The head noun: yes, but only for grammar P. In H&P’s structure (19) the noun after REL is the relative’s external head, and the null higher DP inside is its gap (EH-1). Willie states that in her speech a noun after the relative verb is never part of the relative (p. 158), and her (12) shows such a noun taking a main-clause role instead. So the two sentences are not a minimal pair inside one grammar. (ii) Willie’s definiteness account: it does not separate them. It excludes a restrictive bi- relative on the definite higher DP (p. 164), and H&P’s (20) is exactly that, translated as a restrictive relative (H&P do not discuss restrictiveness). Read strictly, it predicts (20) out too. (iii) Dialect: the only option that fits both sources, and the one Willie names herself.
HYPOTHESISWhat stays open. We still do not know whether the external head is what makes (20) good for grammar-P speakers, or whether those speakers also accept the headless (24b) type, which would mean that Willie’s definiteness restriction is part of the dialect difference too. No headless bi- relative with a null higher DP as head and one overt noun is printed in the H&P pages opened (T108a).
HYPOTHESISEXT-HEAD is limited in grammar P. In (46) the noun after REL (łį́į́ʼ) could in principle be an external head, giving (49), which is the EH-1 configuration. H&P exclude (49), and their reason is the string parse (44), not anything about the inverse movement (EH-3). So with external heads in the picture, head choice in bi- relatives depends on ID-8′ and on (44) together. (48) is also one more datum for ID-8′(b): both nouns overt, the higher DP (ashkii) heads.
HYPOTHESISWhat it says about Grant’s proposal. In EH-1 the d heads inside the relative take only the inner DPs; the external head DP is not the complement of any d in the relative. The derivation works the same whether d_obj stays in the VP or moves with its DP, so “d stays” is still untested. The DP-movement half again does the work: what the external head links to is the higher DP after ID-2, the moved object. Nothing here bears on whether nouns are inside or outside the verb word, since the external head is outside the relative clause in every analysis.
GRANT'S CONJECTURE ID-1 (light determiners taking DPs) remains Grant’s conjecture. Willie (pronominal arguments, nominals as adjuncts, the enclitic as head) and H&P (deletion plus surface parsing) state neither it nor EXT-HEAD in this form.
OPEN T107a (obligatory parallelism with bi-) and T107c (why overtness matters) are not advanced by these pages and stay open.
ID
Question
What would settle it
T108a
Do grammar-P speakers (Platero 1974; H&P) accept a headless bi- relative headed by the null higher DP with one overt noun, the (24b) type? If not, the external head is what licenses H&P (20); if so, Willie’s definiteness restriction is part of the dialect difference too.
Platero 1974 IJAL and Platero 1978, 1982 originals; speaker judgments on (24b)-type strings from H&P-type speakers. Sharpened 2026-11-10 (afternoon) in §15 (RD-5): if they reject the (24b) type, Willie’s analysis of a noun after REL (S1) cannot derive (20) for them either.
T108b
EH-2 predicts that speakers of grammar W have no reading ‘the girl that was seen by the boy is laughing’ for H&P’s (20) string. Is that right, and what, if anything, does the string mean for them?
Willie-type speaker judgments; any printed relative-plus-following-noun example with an intransitive main verb in Willie 1991 (not in pp. 153–168).
T108c
In grammar P, how is the external head DP linked to the null DP inside: by movement of the inner DP (a raising analysis) or by deletion under identity (matching, H&P’s analysis)? Can (44) effects such as (49) tell the two apart?
Reconstruction or idiom data inside externally headed relatives; Platero 1974. Answered 2026-11-10 (afternoon) in §15 for the second question: no, (44) reads only the surface string (H&P p. 24). Correction: matching by deletion is Platero 1973’s analysis, which H&P reject in favour of raising (p. 28); raising vs matching reduces to T34a on the relatives topic.
Sources: Willie 1991, Ch. 6 (box page images, not committed), pp. 156–158, 164. Hale & Platero 1974 (box page images, not committed), pp. 9, 15, 21–23. Platero 1974, 1978, 1982 cited through Willie, H&P and Speas, not opened.
15. Update 2026-11-10 (afternoon): raising or matching for a noun after the relative marker — Hale & Platero’s parse (44) cannot tell them apart, and their own answer is raising, not deletion (T108c)
HYPOTHESIS In plain words: §14 asked how the noun after the relative marker in Hale & Platero’s (20) is tied to the missing participant inside the relative clause. Either the noun itself moves out of the relative (raising), or a second copy of it stays inside and is deleted (matching), and §14 asked whether H&P’s parse rule (44) could tell the two apart. H&P answer that directly: rules like (44) read only the finished string, so they “do not care how a particular surface string is derived”, by deletion or by movement (p. 24). So (44) cannot decide. §14 also called deletion “H&P’s analysis”, which was wrong. The deletion analysis is Platero’s 1973 one, which H&P argue against, and they end by saying the raising hypothesis wins “decisively” (p. 28). What their evidence actually shows is movement rather than deletion, not raising rather than a modern matching analysis. That question was already open on the relatives topic (T34a). One new point came out of re-running the derivations with light determiners. If the noun after the relative is just a separate noun of the main clause (Willie’s analysis), then (20) needs exactly the configuration that Willie stars in (24b). So T108a (do Hale & Platero’s speakers accept the (24b) type?) also decides between Willie’s analysis and a raising or matching one for those speakers. In every option the verb’s argument prefixes stay on the relative verb, and what leaves the relative is something inside the DP, never d. This is consistent with “d stays”, but it does not test it.
Notation as in §§10–14: B = clause syntax (rule set B, right-headed); A = word spell-out (rule set A, the left-headed prefix string); the word boundary is the switch from B to A. “Higher” and “lower” DP mean positions after ID-2 (the bi- movement of the object DP over the subject DP). Drel is the relative marker -ę́ę/-yę́ę/-ígíí, treated as a determiner (H&P p. 27). Op is a silent copy of the relativized DP’s D-layer. S1, S2 and MATCH are the three analyses of a noun after REL already set out on the relatives topic (relative clause as CP, §6). Structures are schematic. Only the Navajo strings, glosses and judgments are data.
15.1 Sources opened this run
SOURCED Hale & Platero 1974, page images of the box copy (PDF page = printed page + 3): (1)–(2) (p. 9); (3)–(5) and “this second alternative is preferred” (p. 10); (9)–(12), the raising alternative (p. 12); p. 13 (optional raising; Chomsky’s and Platero’s island argument; (13), obligatory deletion in relatives); (21)–(27) and “this reasoning is fallacious … not really a condition on rules but rather a condition on the interpretation” (p. 16); (50) (p. 23); “do not care how a particular surface string is derived” and (51)–(52) (p. 24); (53)–(54) and “incontrovertable evidence” (p. 25); (55) and optional raising (p. 26); the determiner analysis, “two versions of raising”, (57) (p. 27); the conclusion (p. 28). The box OCR text was used only to find passages (pp. 14–15 there).
Reused, not re-read: H&P (19)–(20) (p. 15) and (44)–(49) (pp. 21–22) as read for §14; Willie 1991 (12), (24) and p. 158 as in §14. On the relatives topic, the site’s readings of the same H&P pages (Hale & Platero on relatives, §5) and the RAISE-D / MATCH test matrix (relative clause as CP, §5; RC9) were built on Platero 1973 and are taken over as they stand.
Not opened: Platero 1974 (IJAL), 1978, 1982; Kaufman 1972 (cited by H&P as 1973 in the text and 1972 in the references, p. 28). None is on the box.
15.2 Data
Glosses are H&P’s, in Leipzig layout (“.” for their “:”). H&P do not segment verbs, and neither do we. Spelling as on the page images; apostrophes as U+02BC. For (57) H&P print translations only; the gloss line there is ours, built from their own glosses of REL and the translation.
Ex. (page)
Navajo
Gloss
Translation as printed
H&P (1) p. 9
Tłʼéédą́ą́ʼ yáłtiʼ-ę́ę hastiin ałhosh.
last.night speak-REL man sleep
‘The man who spoke last night is sleeping.’ (noun after REL)
H&P (4) p. 10
Tłʼéédą́ą́ʼ hastiin yáłtiʼ-ę́ę ałhosh.
last.night man speak-REL sleep
Same meaning; “there is no head noun phrase in the surface structure”, and this form “is preferred” (p. 10).
H&P (20) p. 15
Ashkii biiłtsą́-(n)ę́e atʼééd yidloh.
boy INV.saw-REL girl laugh
‘The girl that was seen by the boy is laughing.’ (as in §14)
H&P (46) p. 22
Ashkii łééchąąʼí bishxash-ę́ę łį́į́ʼ yizloh.
boy dog bit-REL horse roped
(48) allowed; (49) ‘The boy roped the horse that was bitten by the dog’ not allowed (as in §14)
H&P (57i) p. 27
Łį́į́ʼ-ę́ę dilwoʼ.
horse-REL be.fast
‘The aforementioned horse is fast.’
H&P (57ii) p. 27
Łį́į́ʼ-ígíí dilwoʼ.
horse-REL be.fast
‘The (one which is a) horse is fast.’
SOURCED H&P’s argument, in their words. Platero (1973) derived both (1) and (4) by deleting one of two identical noun phrases (pp. 9–10, rules (3), (5)). Raising instead moves “the lower noun phrase into head position” by an “optional raising rule” (12) (p. 12). Against deletion: relativization obeys Ross’s islands and pronominalization does not, and deletion is obligatory in relatives but optional elsewhere ((13)) (p. 13). The shared constraint (18) is “not really a condition on rules but rather a condition on the interpretation of Navajo sentences with, so to speak, missing noun phrases” (p. 16). Then (50) (p. 23): (i) the rules that remove a noun phrase “apply without constraint”; (ii) principles “like (28) and (44)” apply “at the surface structure level”; (iii) a derivation is ill formed if they give it a reading at odds with its deep structure. These principles “do not care how a particular surface string is derived. That is to say, their application is totally independent of whether a particular syntactic position is vacated by means of a deletion rule or by means of a movement rule” (p. 24). For raising: Kaufman’s rightward raising of spatial enclitics ((51)–(55), pp. 24–26), and the claim that REL is “a determiner emanating from the embedded noun phrase” with “two versions of raising — one in which only the determiner raises and another in which both the noun phrase and the determiner raise” (p. 27). Conclusion: “the struggle between the two competing conceptions of Navajo relativization is won decisively by the raising hypothesis”, with “the exact formulation of the raising rule; the question as to whether it is a copying rule or not” left open (p. 28).
SOURCED Print note: p. 13 says the optional rule covers “(1) — in which the rule would not have applied” and “(4) — in which the rule would have applied”. That is the reverse of the trees (10)–(11) and of p. 26 (“sentences like (1) — with raising”). We follow the trees, as the relatives topic already does.
15.3 Rules first
ID-1–ID-7′, ID-5°, ID-8′(a), (b), (c′) and EXT-HEAD (§14) are unchanged. EXT-HEAD is now read as a cover term for whichever of RD-3’s options grammar P uses.
RD-1 REL IS A DETERMINER. DP_rel = [DP [CP Op_j [C′ TP C_rel]] D_rel], SOURCED (D_rel as a determiner:
D_rel = -ę́ę / -yę́ę / -ígíí, head-final, enclitic on the last word H&P p. 27, (57))
of TP. HYPOTHESIS (CP layer; = REL-CP on
the relatives topic)
RD-2 WHAT LEAVES THE RELATIVE. The relativized DP_j is the complement of HYPOTHESIS
some d_x inside TP (ID-1). Op_j, a silent copy of DP_j's D-layer,
moves to Spec,CP from DP_j's highest A-position, i.e. after ID-2.
d_x is not part of what moves: it is spelled in the relative verb. SOURCED (the prefix is there:
[= RAISE-D / OP-ρ on the relatives topic, restated over d and DP] bi- in H&P (20), (46))
RD-3 A NOUN AFTER REL (three options, taken over from the relatives topic):
S1 the noun is a separate DP of the main clause, linked to the same
main-clause argument as DP_rel (Willie's analysis, W p. 158, (12))
S2 the noun is inside DP_rel: it raises from DP_j to Spec,CP, and the
remnant TP moves to Spec,DP, above D_rel (remnant raising; H&P's
"noun phrase and determiner raise", p. 27, as restated on the
relatives topic)
MATCH an outside noun next to [CP ... D_rel], with a twin inside DP_j that
is not pronounced (Platero 1973's two-copy structure)
RD-4 SURFACE BLINDNESS. (28) and (44) read the finished string and do not SOURCED (H&P (50), pp. 23–24)
see how a position was emptied. S1, S2 and MATCH give the same string
for the same sentence, so (28) and (44) treat them alike.
RD-5 S1 NEEDS THE (24b) CONFIGURATION. Under S1 the relative in (20) is HYPOTHESIS (derived from S1 + ID-8′)
headless, with one overt noun (the lower DP ashkii) and the higher DP
null as its head: exactly W (24b). So S1 derives (20) only in a grammar
without W's (24b) block. S2 and MATCH put atʼééd inside DP_rel, which is
the EXT-HEAD half of ID-8′(c′).
15.4 Derivations (schematic structure; only the Navajo strings are data)
RX-1 H&P (20) Ashkii biiłtsą́-(n)ę́e atʼééd yidloh, first steps (all options)
B Merge [dP d_obj [DP_j ... D]] with V; v; [dP d_subj [DP ashkii]] in Spec,vP ID-1
B Move DP_j → above DP ashkii; d_obj stays in VP ID-2
B Merge C_rel; Op_j (D-layer copy of DP_j) → Spec,CP, from the higher position RD-2
B Merge D_rel -(n)ę́ę over CP RD-1
A spell highest copy of DP_obj over DP_subj → bi-; d_obj and d_subj are in the
relative verb; D_rel leans on it: biiłtsą́-(n)ę́e ID-3, RD-1
Then three ways to place atʼééd:
RX-1a S1 (atʼééd is a main-clause DP)
B relative = [ashkii biiłtsą́-(n)ę́e]: one overt noun (the lower DP), head =
the null higher DP = W (24b) configuration
B matrix DP_rel and DP atʼééd both linked to yidloh's one argument RD-3 S1
→ 'the girl that was seen by the boy is laughing'
✓ only if the grammar lacks the (24b) block; in grammar W, * (EH-2) RD-5
RX-1b S2 (atʼééd raised inside DP_rel)
B DP_j = [DP atʼééd D]: the noun raises from DP_j to Spec,CP; the remnant
TP moves to Spec,DP; order TP–D_rel–noun: biiłtsą́-(n)ę́e atʼééd RD-3 S2
→ ✓ as printed, without the (24b) configuration (the noun is inside)
RX-1c MATCH (outside atʼééd, unpronounced twin in DP_j)
B DP_j = [DP atʼééd D], not pronounced; outside atʼééd merged after D_rel RD-3 MATCH
→ ✓ as printed, likewise without the (24b) configuration
In all three: d_obj and d_subj stay in the relative verb (bi- is spelled there).
What leaves the relative is Op (RX-1), the noun (S2), or nothing (S1, MATCH).
RX-2 H&P (46) Ashkii łééchąąʼí bishxash-ę́ę łį́į́ʼ yizloh, (48) vs (49)
surface the string is the same under S1, S2 and MATCH
(44) ashkii łééchąąʼí bishxash = (47), a simple sentence → both nouns are
read as clausemates of bishxash H&P (44)
(50iii) (49) puts ashkii in the main clause: its reading differs from what (44)
assigns → ill formed, whichever rule emptied the inner position RD-4
→ (49) is excluded under all three; (48) ✓ under all three (ID-8′(b))
So the (44) effect in (49) does not separate raising from matching. T108c (b): no
RX-3 H&P (4) Tłʼéédą́ą́ʼ hastiin yáłtiʼ-ę́ę ałhosh vs (1) Tłʼéédą́ą́ʼ yáłtiʼ-ę́ę hastiin ałhosh
B Merge [dP d_subj [DP_j hastiin D]] in Spec,vP of 'speak' ID-1
B Op_j → Spec,CP; D_rel -ę́ę RD-1, RD-2
(4) hastiin stays in DP_j → noun inside, nothing between REL and ałhosh RD-2 only
(1) hastiin after REL: S1, S2 or MATCH (RD-3); no Navajo datum here chooses
'speak' has no object here, so this pair says nothing about the yi-/bi- choice
or about where d is.
15.5 What this shows and what it does not
HYPOTHESIST108c answered for its second half, narrowed for its first. Second half: no, (44) effects such as (49) cannot tell raising from matching. H&P say so explicitly (p. 24), and RX-2 shows why: the string is the same and (44) reads only the string. First half: H&P’s own answer is raising, but their arguments (islands, obligatory deletion, enclitic raising) show that the relative tie is movement and not anaphoric deletion. They do not exclude a matching analysis in which an operator moves inside the relative. That was already the site’s verdict on the relatives topic (relative clause as CP, §5), where the deciding test is reconstruction with a noun pronounced after REL (T34a). No such sentence is printed in the H&P pages. So T108c reduces to T34a.
HYPOTHESISCorrection to §14. §14 called deletion under identity “H&P’s analysis” (T108c row) and used H&P’s constraint (18) as a deletion condition in EH-1. H&P state (18) only to argue against it as a condition on rules (p. 16). They replace it with the interpretive principle (28) and the general scheme (50), and they conclude for raising (p. 28). EH-1’s steps and verdict are unchanged under RD-3 S2 or MATCH; only the attribution was wrong. Notes are added at EH-1 and in the T108c row.
HYPOTHESISNew narrowing of T108a (RD-5). Willie’s analysis of a noun after REL (S1) can derive H&P’s (20) only through the (24b) configuration, which she stars. So T108a now decides two things at once. If grammar-P speakers reject the (24b) type, S1 is not available to them for (20), and the noun after REL must be inside the relative’s DP (S2 or MATCH). This refines the claim on the relatives topic that S1 “works on all printed data”: it works only if grammar P lacks Willie’s (24b) block. If they accept the (24b) type, S1 is enough, and the dialect difference is Willie’s definiteness restriction, not the external head.
HYPOTHESISWhat it says about Grant’s proposal. In RD-2 and all three options of RD-3, the argument prefixes of the relative verb stay in that verb (bi- in (20), (46)). What leaves the relative is something inside the DP that d takes: its D-layer (Op) or its noun. So relativization is a second kind of DP movement, after ID-2, that leaves d behind, and RD-2 can be stated without mentioning d. That fits “d stays”. A moving-d analysis can match it, though: it only has to restrict d’s movement to A-movement inside the clause. So this is not a test. One caution. H&P’s “only the determiner raises” (p. 27) refers to the relative marker -ę́ę/-ígíí, a D inside the DP. It is not a verb prefix, so it is not evidence that d moves. In fact the relatives give a sourced overt D inside the DP complement of d: (57) shows the same -ę́ę/-ígíí on plain nouns. Nothing here bears on whether nouns are inside or outside the verb word.
GRANT'S CONJECTURE ID-1 (light determiners taking DPs) remains Grant’s conjecture. H&P (deletion versus raising, surface interpretation) and Willie (pronominal arguments, nominals as adjuncts) state neither it nor RD-2 in this form.
OPEN T107a (obligatory parallelism with bi-), T107c (why overtness matters), T108a and T108b are not answered by these pages. T108a is sharpened by RD-5 (see its row in §14). New: T109a, T109b.
ID
Question
What would settle it
T109a
RD-2 says Op leaves DP_j from its position after ID-2. Is there any datum where extraction from the lower (merge) position would give a different result, for example a bi- relative inside an island, or the parallelism readings of §12 recomputed with Op from the merge position?
Platero 1973 island data re-read with bi- relatives (relatives topic); Speas 1986 (4.53)–(4.55).
T109b
If RD-1 is right and the DP complement of d can carry an overt D (-ę́ę/-ígíí, (57)), can such a D-marked noun stand inside an ordinary transitive clause next to the verb, and does it change the yi-/bi- choice (ID-3) or NP-PRO (ID-5)?
Printed simple sentences with N-ę́ę or N-ígíí arguments of a 3.3 verb (Young & Morgan; Willie Ch. 3).
Sources: Hale & Platero 1974 (box page images, not committed), pp. 9–10, 12–13, 16, 23–28, with pp. 15, 21–22 as read for §14. Willie 1991 pp. 158, 164 as in §14. Platero 1973 through the relatives topic. Platero 1974, 1978, 1982 and Kaufman 1972 not opened.
Restated on the unified structure (2026-11-15). Nouns are restated as arguments merged in their theta positions (VP, Spec,vP) and, for the subject, raised to Spec,TP by the working default; a lone noun is linked by the H4″ default. Adjunction and pronominal-argument analyses below are labelled contrast; where nouns sit for Condition C is to be re-run on the unified spine. Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). Flat-S and other pre-VPISH notation below is labelled contrast. Status on the spine: see unified structure, §2 and §4, first rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.
A short paper in which Ken Hale argues that Navajo is configurational after all. A null subject does block coreference with a noun inside its object or inside a clause under a postposition, as Condition C predicts. Coreference is allowed only where Platero's parsing rule reads two nouns as shared by two verbs. This page reads every Navajo sentence in the paper against Grant's two phenomena. It supplies the equal-rank Condition C case asked for in T22b. Whether it counts against the feature-bundle hypothesis depends on ten assumptions (A1–A10), which the new step-by-step derivation page lists and tests. Written 2026-10-15, 12:46pm run (PT). Corrected 1:07pm: the first version overstated the result (“FB-5 is wrong”, “D6 withdrawn”); the wording below is now conditional.
Ground rules (Grant's). Hale, Platero, Speas, Willie and Jelinek are each treated as possibly partly right. Only data that contradict the feature-bundle hypothesis (FB, Not quite a word) count against it, and here the question is whether they do. Dowty's (1991) proto-roles are the baseline, and labels such as “agent” name a participant's role in the action. Grant does not speak Navajo. Every Navajo form below is copied from a page image of the paper, with its example number and printed page, and Hale's own glosses are kept as printed. None is invented.
Bottom line
HYPOTHESISFB-5's Condition C exemption is in tension with (10) and (11a), conditional on assumptions A1–A10. FB-5 said a null bundle can never trigger Condition C on a noun, because the noun is linked to its slot rather than merged under the bundle. Hale's (10) and (11a,b) report that it does: in (10) the null subject of ‘give’ cannot be ‘my father’, who is named inside the theme phrase ‘my father's very horse’; in (11a,b) the null subject cannot be ‘your younger brother’, named inside a clause under the postposition ‘about it’. Both participants are people, so rank (FB-4) plays no part, and Condition C is the obvious candidate to exclude these readings (pp. 172–173). That is the equal-rank test asked for in T22b. But the inference from these judgments to “the nouns sit under the subject's reach” needs ten assumptions (audit), none independently established for Navajo; A3 (“an adjunct is outside the subject's c-command domain”) is contradicted by English data, A5 (stable judgments) by Speas's (4.119), and (11b) needs A10, which this site's own psych-verb analysis does not support. The 12:46pm wording “FB-5 is wrong” and “the D6 verdict is withdrawn” was too strong. D6 is restored to compatible, undecided.
HYPOTHESISWhat is not touched. FB's core, FB-1 to FB-4 (the arguments are role-bearing bundles spelled as prefixes, ranked by voice), is untouched: Hale's data are about where nouns sit, not about what bears the role. If the assumptions A1–A10 hold, FB would need to add that a linked noun sits inside the projection of the head whose bundle it links to, within reach of that head's external bundle (FB-3c LINK-IN, §6), replace the exemption with Condition C over unlinked coreference (FB-5′), and restrict the parallel-linking rule to Hale's own condition, a transitive embedded verb (FB-6′). These are conditional additions, not forced revisions. With them all nine of the paper's data groups fit (§9: 0 cut against FB's core); without them, FB-5's exemption is in tension with (10) and (11a) only.
HYPOTHESISThe cost, if the assumptions hold. With LINK-IN, FB and Speas's configurational analysis put the nouns in the same places, and differ only on which member of the noun–prefix pair carries the role (T24f). The adjunct version of the Pronominal Argument Hypothesis (T-WEAK on the thesis page) is then hit in one placement only, nouns adjoined above the subject, as in Hale's tree (9); nouns adjoined low inside the clause (D3) give the same Condition C result as nouns in argument positions. Hale's (5) adds an extraction argument, itself conditional on an unopened analysis (§4).
OPENA conflict between sources. Hale's (11a) and Speas's (4.119) (1986:319–320) have the same shape: one noun, an intransitive verb inside a nominalized clause, and a postposition ‘about it’ under a verb of telling. Hale reports coreference as impossible and Speas reports it as possible. Neither Hale's parsing rule nor FB-6′ allows Speas's reading. This is T24a, and it is why A5 (stable judgments) is not secured. The derivations are on the derivation page.
1. Source and caveats
SOURCED Hale, Ken 1999. “Conflicting truths.” In M. Darnell, E. Moravcsik, F. Newmeyer, M. Noonan & K. Wheatley (eds.), Functionalism and Formalism in Linguistics, Vol. I: General Papers, pp. 167–175. Amsterdam: John Benjamins. The bibliographic details come from Hale's CV (MIT); the scan prints only the page numbers. The copy is MIT's scan (ling-phil.mit.edu/papers/hale/papers/hale082.pdf, 9 pages, image only, no text layer). All nine pages were read on page images (150 dpi). The Navajo lines of (1a–b), (2), (3a), (5), (7), (10), (11a,b) and (12) were read again in 300 dpi crops, with the key words magnified; (1c–d) and (3b) were read at 150 dpi only. The paper says the data come from “work done nearly twenty years ago by Paul Platero, Ellavina Tsosie Perkins, and others” (pp. 167–168).
Fidelity notes. (7) glosses the matrix verb “3.FEM.3.brand”, but the translation is ‘will brand’. FEM is probably a misprint for FUT, but it is kept as printed. (5b) prints yiiłtsá(n)-í-da with the n in parentheses. (7) prints the matrix verb yídiʼdoołił with no tone mark on the second i, and (10) prints yeiníłtį́. (1) and (3) print the bare stems yiiłtsą́, siztał, yiztał, biztał. The macron-less “3y”/“3b” glosses are Hale's, and n. 2 explains them (§5). The trees (8)–(9) use English words, which are reproduced in prose only.
2. The data SOURCED
Hale's glosses are given as printed: P = perfective, REL = relativizer, PRN = the dislocated pronominal element, DA = the negative-polarity enclitic, YBr = younger brother, Fa = father, Mo = mother, ns = non-singular, anim = animate object class.
(1) a. (Shí) dzaanééz yiiłtsą́. (I) mule P.1SG.see ‘I saw the mule.’
b. Dzaanééz (shí) siztał. mule (me) 1SG.P.3.kick ‘The mule kicked me.’
c. (Shí) yiiłtsą́. (I) P.1SG.see ‘I saw it.’
d. (Shí) siztał. (me) 1SG.P.3.kick ‘It kicked me.’ (p. 168)
(2) a. Łį́į’ dzaanééz yiztał. horse mule 3y.P.3.kick ‘The horse kicked the mule.’ (direct)
b. Dzaanééz łį́į’ biztał. mule horse 3b.P.3.kick ‘The horse kicked the mule.’ (inverse)
(3) a. Dzaanééz yiztał. mule 3y.P.3.kick ‘It kicked the mule.’ not: ‘The mule kicked it.’
b. Łį́į’ biztał. horse 3b.P.3.kick ‘The horse kicked it.’ not: ‘It kicked the horse.’ (p. 169)
(4) (NP1) NP2 X AGR_OBJ V If AGR_OBJ = yi-, NP2 = Object
If AGR_OBJ = bi-, NP2 = Subject (p. 169)
(5) a. Doo háí-da bi-łį́’ yiiłtsą́ą-da. NEG who-DA 3-horse P.1SG.see-DA
‘I didn't see anyone's horse.’
b. Doo ___ bi-łį́’ yiiłtsá(n)-í-da. NEG ___ 3-horse P.1SG.see-PRN-DA
‘I didn't see anyone's horse.’ (p. 170)
(7) [Jį́į́dą́ą́’ shi-zhé’é łį́į’ nayiisnii’-ę́ę] yídi’doołił.
day.past 1SG-father horse 3.P.3.buy-REL 3.FEM.3.brand
‘My father_i will brand the horse he_i bought (earlier) today.’ (p. 171)
(10) Si-tsilí [ts’ídá shi-zhé’é bi-łį́’-ígíí] yeiníłtį́.
1SG-YBr very 1SG-Fa 3-horse-REL 3.to.3.P.3.give/anim
‘He_i gave my younger brother the very horse of my father_j.’ (i ≠ j) (p. 172)
(11) a. [Yiską́ągo ni-tsilí nih-aa doogáł-ígíí] y-ee shi-ł hoolne’.
tomorrow 2SG-YBr 1ns-to FUT.3.go-REL 3-of 1SG-with 3.P.tell
‘(S)he_i told me about your younger brother_j coming to us tomorrow.’ (i ≠ j) (p. 172)
b. [Yiską́ągo ni-tsilí nih-aa doogáł-ígíí] b-aa bi-ł hózhǫ́.
tomorrow 2SG-YBr 1ns-to FUT.3.go-REL 3-about 3-with A.P.good
‘(S)he_i is happy about your younger brother_j coming to us tomorrow.’ (i ≠ j) (p. 173)
(12) [Yiską́ągo shi-zhé’é shi-má y-íká-’adoolwoł-ígíí] y-ee yi-ł hoolne’.
tomorrow 1SG-Fa 1SG-Mo 3-for-FUT.3.run-REL 3-of 3-with P.3.tell
‘My father_i told my mother (about the fact) he_i would help her tomorrow.’ (p. 173)
(13) X of (4) may be a transitive verb. It follows that (NP1) NP2 will be interpreted
as linearly parallel arguments of X and V. (p. 173)
3. Hale's argument SOURCED
The parsing rule is rigid. A single noun before a transitive verb gets the relation it would have if a co-argument preceded it, (3). This holds “quite rigidly” (p. 169). The variable X in (4) “cannot be a nominal argument, overt or nonovert”, and the rule “cannot in general be overridden in the absence of extraordinary contextual motivation … it is essentially inviolable” (pp. 169–170).
Nouns are not adjuncts. Hale grants that the Pronominal Argument view (Jelinek 1984; Baker 1996) “is an attractive idea and it has been fruitful in the study of Navajo” (p. 170). But “there is reason to believe … that overt nominal expressions in Navajo are not adjuncts, at least not in the sense of the Polysynthesis Parameter of Baker (1996)”. His first reason is (5b). There the negative-polarity element is moved out of the object phrase ‘anyone's horse’ to a position after the verb (Hale & Platero 1996, unpublished MIT ms.), which would violate the Condition on Extraction Domains if that phrase were an adjunct (p. 170). 8:46pm update: Jelinek 1984 (Warlpiri) and Baker's later “Agreement, Dislocation…” footnote on Navajo are now opened for attachment height — Where do Jelinek and Baker attach the nouns? (T25c partly answered). Baker 1996 itself remains unopened.
Condition C is real. (7) was once an argument for the adjunct view. A null matrix subject that c-commanded the relative clause would make (7) a “straightforward Condition C violation” (p. 171, tree (8)). If nouns are adjuncts, nothing c-commands them (tree (9)). But Navajo, “unlike Mohawk”, does not allow coreference in (10), and the same prohibition holds in the factive nominalizations (11a,b). Hale writes that “the matrix subject (pro) c-commands all overt nominals (and hence, all R-expressions)” (p. 172), and that “(10) and (11) show that Condition C is obeyed in Navajo” (p. 173). He now weighs the evidence “in favor of Navajo configurationality” (p. 172), with Speas 1990.
The reconciliation. (7) and (12) are possible because of Platero's Interpretation of Grammatical Relations (Platero 1978, 1982). If X in (4) is a transitive verb, the nouns are read as shared, parallel arguments of X and V, (13). “All of this can be reconciled if we assume simply that Condition C is overriden by the force of (4)” (p. 173). “When the parallel interpretation permitted by (4) is not possible or relevant in a given construction, Condition C is in full force in Navajo … the conflict is resolved by overriding Condition C, not by declaring the sentences ungrammatical” (p. 174). For a different formulation of the principle he points to Speas 1990:221–237.
4. Extraction, (5): a second argument against outside nouns (conditional on an unopened analysis)
SOURCED Hale's claim: (5b) is derived by movement out of the noun phrase bi-łį́ʼ (his (6), … doo [DP __ N] … PRN-da), and movement out of an adjunct would violate the CED (p. 170). The full argument is in Hale & Platero 1996, which is unpublished and not opened, so this page takes the movement analysis on Hale's word (T24d).
HYPOTHESIS Under FB the possessor háí-da links to the possessor bundle bi- (FB-1 covers that slot; Speas 1986:279 grants that possessive bi- “may be + or − argument”). The phrase bi-łį́ʼ links to the internal bundle of ‘see’. If linked phrases were adjoined outside the verb's projection, (5b) would be a CED violation. So (5b) requires LINK-IN as well. This bears on D7 (extraction) on the thesis page: it bears against outside linking if the movement analysis in Hale & Platero 1996 (not opened, T25h) and the CED applying to an adjoined nominal both hold, and it is COMPATIBLE with FB plus LINK-IN.
5. What this means for each version
Version
Effect of Hale 1999
T-STRONG (nouns are never arguments; roles alone decide links)
Already failed. (3) is one more case: a single noun links by position, and both readings of Dzaanééz yiztał are equally plausible, so the roles cannot choose. Hale calls the rule “essentially inviolable”.
T-MID / FB with FB-5 as stated (bundles are arguments; linked nouns are outside c-command)
In tension with (10) and (11a), which show disjointness with no rank confound, and with (5b) (CED). Conditional on A1, A2, A4, A5, A7, A8, A9 (and A10 for (11b)) and, for (5b), on an unopened analysis. Compatible with Speas's (4.119). Undecided.
Compatible: 0 of 9 data groups cut against it. If the assumptions hold, it is structurally close to Speas (T24f). The additions are conditional, not required by the evidence.
T-WEAK (Pronominal Argument Hypothesis: nominals are adjuncts)
Hit, conditionally, by (10), (11a,b) and (5b), which is Hale's own conclusion: nouns adjoined above the subject are excluded if A1–A10 hold. Nouns adjoined low inside the clause are not excluded (D3). Willie's and Jelinek's analyses keep the arguments in the prefixes, which FB keeps.
Speas (nouns in θ-positions; prefixes as non-θ members of a chain)
Supported on placement. It still needs a device for (7) and (12): ATB in 1986, the Parallel Function Constraint in 1990, and for Hale an override by (13).
SOURCED n. 2 (p. 174): Hale glosses yi- as an object prefix by custom (Young & Morgan 1980:G171), but Hu Matthews suggested that it is really a subject prefix, “and the idea is developed in Uyechi (1996)”. Hale finds “much merit” in it. HYPOTHESIS FB-4 is stated over the two bundles' ranks, not over which one the prefix spells, so it is neutral on this. FB-4's clause “bi- marks external ≠ ρA” would have to be restated if yi-/bi- agree with the subject. That is T24e.
6. Rules (conditional additions)
HYPOTHESIS The three rules below are what FB would add if assumptions A1–A10 of the audit hold. They are not required by the evidence as it stands.
FB-3c LINK-IN (conditional) A noun, possessive phrase or nominalized clause linked to a bundle of head H
(V or a grammatical P) is merged inside H's projection, c-commanded by the external
bundle of the clause that contains H. [placement only; the noun still bears no ρ]
FB-5′ (replaces FB-5's exemption) β is φ-level (a variable; Déchaine & Wiltschko 2002).
An R-expression N may not be coindexed with a bundle that c-commands it
unless N is linked to that bundle. [Condition C over unlinked coreference]
FB-6′ (replaces FB-6 for overt nouns; Platero's IGR as given by Hale (13))
In (NP1) NP2 X V, if X is transitive, the overt nouns link to the same-position
bundles of both X and V (external↔external, internal↔internal; FB-4 applies in each).
Otherwise a noun links only to bundles of its own head (FB-3a).
Null bundles with no overt noun stay free in reference (as before).
Rule
Status
Basis
LINK-IN
HYPOTHESIS conditional
Needed for (10), (11a) (Condition C) and (5b) (CED) only if A1–A10 hold and the unopened (5b) analysis is right (pp. 170–173)
FB-5′
HYPOTHESIS
Condition C as Hale states it, with linking playing the part of his “override”. This is a restatement, not a deeper explanation: it predicts what Hale's (4)+(13) predict
FB-6′ (transitive X)
SOURCED condition, HYPOTHESIS statement over bundles
Hale (13), from Platero 1978, 1982. Why X must be transitive is not explained (T24b)
FB-3a(b) as a hard rule
SOURCED
Hale's (4) is “essentially inviolable” (p. 170). This matches H&P 1974 (28). Contrast FB-3a(a′), the lenient default (H&P 1974 (44))
7. Rule set A vs rule set B
HYPOTHESIS Hale's (4) is a clause-level, left-to-right principle (rule set B) that reads one feature, yi- vs bi-, off the inside of the verb word, which rule set A built. The word is where the two meet. B does not see the prefix string's internal order, only the voice value FB-4 fixed. (13) extends the same B-level reading across two verb words in a row. In (5b) the dislocated -í (PRN) and the enclitic -da follow the finished verb word at its right edge, which is B's domain, as with the enclitics on the H&P 1974 page. This is not stated by Hale.
8. VPISH-aware derivations
Conventions as on the Speas page: Σ = role-schema step; Merge/Agree/Move at clause level; A = word-internal linearization (the prefix string); B = clause level (nouns, clauses, enclitics). V-P = VPISH raising of the external bundle to Spec,M/S inside the word. Proto-role labels are ours (Dowty 1991 baseline). Verb words are spelled out whole, because Hale does not segment them.
CT1. (3a) Dzaanééz yiztał: one noun, linked by position
Σ Σ(√KICK) = {b1[ρ: causation, movement], b2[ρ: affected]} HYPOTHESIS (labels)
FB-4 no noun for b1; both 3rd person; direct form, b1 external HYPOTHESIS
Merge b2 in Comp,V ; b1 in Spec,v ; V-P: b1 → Spec,M/S VPISH
Agree b2 ↔ AGR_OBJ: yi- (direct) SOURCED (form)
A spell-out yiztał SOURCED (p. 169)
B FB-3a(b) = Hale (4): noun next to a yi- verb → internal b2 SOURCED (rule)
dzaanééz links to b2 (merged inside VP by LINK-IN)
Result ‘It kicked the mule’, never ‘The mule kicked it’. Both are plausible,
so position decides, not role fit. [against T-STRONG; compatible with FB]
CT2. (10): Condition C with two people
Σ Σ(√GIVE/anim) = {g1[ρ: vol, causation], g2[ρ: recipient], g3[ρ: affected, moved]}
Merge g3 in Comp,V ; g2 above it ; g1 in Spec,v ; V-P: g1 → Spec,M/S
A spell-out yeiníłtį́ SOURCED (p. 172)
B si-tsilí links to g2 ; [ts’ídá shi-zhé’é bi-łį́’-ígíí] links to g3
inside DP: shi-zhé’é links to the possessor bundle bi- (FB-1), nowhere else
LINK-IN: the DP sits inside VP, c-commanded by g1
FB-5′ index g1 = shi-zhé’é? g1 c-commands the noun and is not its link → excluded
Result ‘He_i gave … the horse of my father_j’, i ≠ j. SOURCED (judgment).
FB-5 as first stated (noun outside, no Condition C) allows i = j: that is the tension
with (10), conditional on A1–A10 (see the derivation page, D1–D2).
CT3. (11a): the same under a postposition
Inner clause (go): Σ {m1[ρ: movement]} ; Merge m1 Spec,v ; A doogáł ; B ni-tsilí → m1
nih-aa: P bundle 1ns ; REL -ígíí closes the clause (a D, H&P 1974:27)
P: y-ee 3-of: P bundle p1 ; [clause-ígíí] links to p1 ; LINK-IN: clause inside PP
Matrix (tell): Σ {t1[ρ: vol, sentient], t2[ρ: addressee via shi-ł]} ; Merge t1 Spec,v
A hoolne’ ; PP adjoined inside VP, so t1 c-commands ni-tsilí
FB-6′ X = doogáł is intransitive → no parallel link of ni-tsilí to t1
FB-5′ t1 = ni-tsilí excluded
Result ‘(S)he_i told me about your younger brother_j coming’, i ≠ j. SOURCED.
(11b) is NOT quite the same: b-aa ‘about it’ and hózhǫ́, but the ‘(S)he’ is the 3 of bi-ł
‘with (him/her)’, inside a postposition (correction 1:07pm; derivation page, D6, A10).
CT4. (7): parallel linking licenses coreference
Inner (buy): Σ {k1[ρ: vol, causation], k2[ρ: affected]} ; FB-4 → direct
Merge, V-P ; A nayiisnii’ ; REL -ę́ę
B shi-zhé’é → k1, łį́į’ → k2 (Hale (4), yi-)
Matrix (brand): Σ {d1[ρ: vol, causation], d2[ρ: affected]} ; A yídi’doołił
FB-6′ X = nayiisnii’ is transitive → the nouns also link to d1, d2 (Hale (13))
shi-zhé’é is linked to d1, so FB-5′ does not apply to that pair
LINK-IN [REL clause] inside matrix VP; Jį́į́dą́ą́’ is inside the relative
Result ‘My father_i will brand the horse he_i bought’. SOURCED (judgment).
Hale: Condition C overridden by (4). FB: linking, not binding. Same prediction.
CT5. (12): the parallel reaches a postposition's bundle
Inner (help): shi-zhé’é → external, shi-má → the P bundle y-íká ‘for her’ ; A ’adoolwoł
P/matrix: y-ee ‘of it’ takes the clause ; yi-ł ‘with her’ ; A hoolne’
FB-6′ the inner verb is transitive in Hale's sense, so (13) applies:
shi-zhé’é links to the teller, and the second noun is read as the matrix yi-ł bundle
Result ‘My father_i told my mother he_i would help her’. SOURCED.
HYPOTHESIS: (13) treats a P bundle (yi-ł) as a parallel slot, as FB-1 allows.
OPEN: Hale counts y-íká-’adoolwoł as transitive; whether X's “object” may itself
sit in a P (T24b).
CT6. (5b): extraction from a linked phrase
Merge [DP háí-da bi-łį́’] inside VP (LINK-IN), linked to the internal bundle of ‘see’
Move háí-da → right edge, spelled PRN -í (Hale & Platero 1996, via Hale p. 170)
A yiiłtsą́ (verb word)
B -í-da after the verb word: yiiłtsá(n)-í-da SOURCED (form)
Result fine. If the DP were an adjunct outside VP, the move would break the CED.
9. Datum-by-datum table
Hale datum (page)
Hale's conclusion
FB reading
Verdict for FB
(1a–d) optional pronouns, variable order (p. 168)
Pronominals optional, “preferably absent”
Bundles complete without nouns (FB-2)
COMPATIBLE
(2a,b) direct/inverse with two animals (p. 169)
yi- with SOV, bi- with OSV
FB-4 (equal rank: both voices) + FB-3a(b)
COMPATIBLE
(3a,b) one noun (p. 169)
(4), “essentially inviolable”
FB-3a(b), hard
COMPATIBLE (against T-STRONG)
(5)/(6) NPI extraction (p. 170)
Nouns are not adjuncts (CED)
Needs LINK-IN
Bears against outside linking, conditional on an unopened analysis (T25h); COMPATIBLE with LINK-IN
(7) relative, coreference ✓ (p. 171)
Condition C overridden by (4)/(13)
FB-6′ parallel link
COMPATIBLE (was counted as support via FB-5)
(10) possessive phrase, i ≠ j (p. 172)
Condition C obeyed
FB-5′ + LINK-IN
In tension with FB-5's exemption, conditional on A1, A2, A4, A5, A8, A9; COMPATIBLE with LINK-IN
(11a,b) factive under P, i ≠ j (pp. 172–173)
Condition C obeyed
FB-5′ + LINK-IN; FB-6′ blocked (X intransitive)
(11a): in tension with FB-5's exemption, conditional on A1, A2, A4, A5, A7, A8, A9. (11b): not independent, needs A10. COMPATIBLE with LINK-IN
(12) factive, two nouns, coreference ✓ (p. 173)
(13) applies
FB-6′, reaching a P bundle
COMPATIBLE
n. 2: yi- as a subject prefix? (p. 174)
“much merit” (Uyechi 1996)
FB-4 neutral; the bi- clause needs restating
OPEN (T24e)
Tally (corrected 1:07pm): over 9 groups, 0 cut against FB's core (FB-1–4). (10) and (11a) are in tension with FB-5's exemption, conditional on the assumptions; (11b) adds nothing independent; (5b) bears on outside linking, conditional on an unopened analysis. All 9 are compatible with the conditional additions, and the exemption-based FB-5 is undecided against Speas's (4.119). Over the paper as a whole, Hale's “configurational” conclusion and FB-revised assign the same structures. They differ only in where the role sits.
Knock-on effects for the Speas 1986 table (there). (4.15a) moves from SUPPORTS to COMPATIBLE, because it can be derived by FB-6′, as Hale derives (7). (4.119)–(4.121) move from SUPPORTS to OPEN: they conflict with Hale's (11a), and neither FB-6′ nor (13) allows them (T24a). The Speas tally is 0 against, 1 supports, 16 compatible, 1 open; the Condition C row (D6) is compatible, undecided, not withdrawn.
10. What each source gets right, and what it misses
Source
Gets right (FB keeps)
Misses or leaves open
Hale 1999
(10) is a clean case: one verb, two people, no postposition. If its judgment is stable, the best available explanation is Condition C with the noun under the subject's reach. The parsing rule (4) is rigid and does real work. A bare “adjunct” is not assumed to exempt a noun
Why (13) needs a transitive X; how the override is stated in the grammar (“overriding” is a description); no account of Speas's (4.119)
Jelinek 1984 / Willie 1991 (adjunct nouns)
The prefixes are the arguments (FB-1). 8:46pm, revised 9:20pm: Jelinek 1984 tree (28) is pre-VPISH notation for Warlpiri (flat S: V, AUX, CPP); it gives no verdict on A3 and is not a datum about Navajo (page; restatement in VPISH/TP/CP is T26g)
Adjunct placement above the subject is in tension with (10), (11a), and (5b), conditional on A1–A10; low adjunction is not excluded. Baker's later n. 16 contemplates PP/vP adjunction for Navajo (against A3); Baker 1996 unopened. Later restatements say nouns are adjuncts “to the clause” without drawing them (T26g)
Speas 1986/1990
Nouns inside the clause; a parallelism device
Puts the role on the noun or the chain. No datum here forces that over the bundle
FB (this site)
The core survives
FB-5's exemption is in tension with (10) and (11a) if A1–A10 hold, and is undecided against (4.119). The first version of this page overstated that (corrected 1:07pm)
11. Open questions T24a–f
#
Question
What would decide it
T24a
Hale (11a) i ≠ j vs Speas (4.119) i = j in the same shape (one noun, intransitive inner verb, ‘about it’, a verb of telling). Is the difference mood (optative wish …anaa’ vs plain perfective), the time adverb (‘last year’ vs ‘tomorrow’), the possessor's person (‘my’ vs ‘your’), or the speaker?
Speaker judgments on minimal pairs; Speas 1990:221–237
T24b
Why must X in (13) be transitive? Is it X's own yi-/bi- that licenses the parallel parse? What if X's object is 1st/2nd person, or sits in a postposition as in (12)?
Platero 1978 (MIT diss.), 1982 (IJAL 48:286–305)
T24c
Is the disjointness in (10)/(11) Condition C, or a fourth-person effect (would a 4th-person matrix subject allow coreference)? Hale does not discuss it.
Willie 1991 4th-person data; speaker judgments
T24d
The CED argument in full: Hale & Platero 1996, Negative Polarity in Navajo (MIT ms.). Is the PRN position reached by movement?
The manuscript; the Hale & Platero NPI excerpt already cited on the thesis page (earlier labelled Hale & Perkins)
T24e
Is yi- a subject prefix (Matthews p.c.; Uyechi 1996)? Restate FB-4's bi- clause accordingly.
Uyechi 1996 (in Jelinek et al. eds., pp. 123–135)
T24f
With LINK-IN, what datum separates “ρ on the bundle, noun linked” (FB) from “ρ on the noun or chain” (Speas)? Candidates: clauses with no nouns that are fully interpreted (FB-2), quantifier scope (D8), and a noun whose sort clashes with the role its position assigns.
Restated on the unified structure (2026-11-15). Nouns are restated as arguments merged in their theta positions (VP, Spec,vP) and, for the subject, raised to Spec,TP by the working default; a lone noun is linked by the H4″ default. Adjunction and pronominal-argument analyses below are labelled contrast; where nouns sit for Condition C is to be re-run on the unified spine. Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). Flat-S and other pre-VPISH notation below is labelled contrast. Derivations below in which the verb or subject does not raise are labelled contrast; the working default raises V→v→T and the subject to Spec,TP. Status on the spine: see unified structure, §2 and §4, first rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.
2026-11-11 audit pass 2. Wording that used VPISH as the name of a spine, tree, skeleton or set of slots named a structure after the hypothesis. VPISH is a hypothesis name, never a node; the structure is a vP/TP spine with the subject first-merged in left Spec,vP. Only the wording changed.
Grant asked for Hale's (1999) argument to be derived at the level of detail of the syntax pages: explicit trees, Merge and Move, c-command at every step, and VPISH. This page does that for Hale's (10) and (11a,b) and for Speas's contrary (4.119), under four views of where the nouns sit: outside the clause (as an adjunct above the subject), inside the clause (adjoined within the verb's projection, or filling the argument slot), and Grant's new STRATA idea, in which the noun enters at a later layer. Then every extra assumption Hale's argument needs is listed and tested. Written 2026-10-15, 1:07pm run (PT).
Ground rules (Grant's). Hale, Speas, Willie, Jelinek and Baker are each treated as possibly partly right; no single account is the judge. A source that does not state a hypothesis is not evidence against it; only data that contradict it count. Grant does not speak Navajo. Every Navajo form below is copied from a page image (Hale 1999, pp. 172–173; Speas 1986, pp. 319–320) with its example number, and the glosses are kept as printed. Nothing Navajo is invented. Where a derivation needs a structural label (g1, N3, v′), that label is ours and is marked as such.
Update 2026-10-16 (4:46am): re-run in a VPISH/TP/CP spine (T27i). The four views below use Hale's S-based trees (8), (9), which are pre-VPISH notation (Grant's stance). Condition C by site restates view (a-S) as named sites (CP, QualP, M/SP, vP, PP, a left-edge specifier) and re-derives (10), (11a,b) and (4.119) with derivational c-command only. Result: no site predicts both Hale and Speas. A noun adjoined to vP is outside the reach of a subject that stays in Spec,v (Hale 2003), so “adjunct” and “outside the reach” come apart in both directions. (11b) needs A10 at every site. D6 stays compatible, undecided. (4.119)'s verb is printed hólne, without a final glottal stop; corrected below.
Bottom line, in plain words
HYPOTHESISHale's argument is valid but conditional. If the nouns sit in a position the hidden subject can “see” (c-command), Condition C predicts exactly what Hale reports in (10) and (11): the subject cannot be the same person as the noun. If the nouns are adjoined above the subject, as in Hale's own tree (9), Condition C does not apply and coreference should be fine. The derivations in §4 confirm both. So the argument does rule out one specific placement, nouns hanging above the subject, provided ten further assumptions (A1 to A10, §5) all hold. None of the ten is independently established for Navajo. One (A3, “an adjunct is out of the subject's reach”) is contradicted by English data we opened. One (A5, the judgments are stable) is contradicted by Speas's (4.119), a sentence of the same shape with the opposite judgment. One (A10) fails for (11b) on this site's own analysis of psych verbs.
HYPOTHESISWhat this does to the feature-bundle hypothesis (FB). The earlier page wording (“FB-5 is wrong”, “D6 withdrawn”) overstated what Hale's paper shows, and is corrected here and on /conflicting-truths/. FB-1 to FB-4 are untouched. FB-5's exemption (a linked noun is outside the bundle's reach) is in tension with (10) and (11a), but only conditional on A1, A2, A4, A5, A7, A8, A9 (and A10 for (11b)). Whether that tension counts against the broader claim that nouns are linked rather than merged also needs A3. The Condition C verdict (D6) goes back to compatible, undecided: Hale's two-people cases pull one way, Speas's (4.119) pulls the other, and each needs assumptions nobody has verified.
GRANT'S CONJECTURESTRATA (roles first, event structure second, the noun last) is a family of derivations, not one prediction. Its slot-filling version predicts the same as “inside the clause”; its “enters outside the syntax” version predicts the same as “outside”; one in-between version differs only in a corner (§7). Published layered models (Distributed Morphology, Borer, Ramchand) support the first version, where the noun's shell is in the tree from the start.
Condition C: a pronoun cannot refer to a full noun phrase (a name, “my father”, “the horse”) that it c-commands. Example (English, only to show the idea): in “He said John left”, “he” cannot be John. In “His mother said John left”, “his” can be John. In the second sentence “his” sits inside the subject phrase and does not “see” John. Opened English cases with the same logic: *“Zelda sent him1 back all Dr. Levin's1 flowers”, and *“He1 sent Zelda back all the flowers which she had bought for Dr. Levin1” (Reinhart 1976, via Abe, “Late Merge and Phase Theory”, ex. (2a), (5)).
1.2 c-command in one picture
S
/ \
He_i VP He c-commands everything inside its sister VP
/ \ (V and NP, and anything inside NP).
V NP So He_i and a name inside NP cannot be the same person.
\
John_j "His mother" would sit inside the left box, not above VP.
In plain terms: A c-commands B if B is A's sister or sits inside A's sister. “At least as high, on the same branch.” This page uses the step-by-step version of the idea (§3): when something is added to the tree, it c-commands everything already inside the thing it is added to.
1.3 Why a hidden subject prefix should count, and why it might not
Why it should. On the Pronominal Argument view (Jelinek; Willie; Hale's (16)) the person prefix on the verb is the subject argument. An argument in the subject slot is higher than the object slot, pronounced or not. On the VPISH analysis used on this site the subject is merged inside the verb phrase, in the specifier of vP, and still c-commands the whole VP. Hale draws a free-standing null pronoun “Proi” at that spot in trees (8) and (9) (pp. 171–172).
Why it might not. (i) The prefix is inside a word. Whether a piece inside a word can c-command phrases outside the word depends on how c-command is defined and on whether the prefix counts as a syntactic object at all (Speas treats prefixes as infixes visible to syntax; this is A2). (ii) If the prefix is only agreement, it has no index of its own and cannot bind anything. (iii) Even if the subject slot is there, the question is whether the noun is within its reach, and that is exactly what the “outside” and “inside” views disagree about (A3).
1.4 Hale's argument in four lines
P1 If the nouns are adjuncts, they sit outside the c-command domain of the matrix subject. (p. 173)
P2 So under that view coreference between the subject and a noun inside them should be OK.
P3 But in (10) and (11a,b) it is not OK (i ≠ j). (pp. 172–173)
C So the nouns are not outside; Navajo is configurational; Condition C is obeyed. (pp. 172–174)
(Where coreference is OK, (7) and (12), Platero's parallel parse (13) overrides Condition C.)
The conclusion needs P1 and P3 to be true and needs Condition C to be what produces P3. Each of those hides assumptions. §5 lists them.
2. The data SOURCED
(10) Si-tsilí [ts’ídá shi-zhé’é bi-łį́’-ígíí] yeiníłtį́. (Hale 1999:172)
1SG-YBr very 1SG-Fa 3-horse-REL 3.to.3.P.3.give/anim
‘He_i gave my younger brother the very horse of my father_j.’ (i ≠ j)
(11a) [Yiską́ągo ni-tsilí nih-aa doogáł-ígíí] y-ee shi-ł hoolne’. (Hale 1999:172)
tomorrow 2SG-YBr 1ns-to FUT.3.go-REL 3-of 1SG-with 3.P.tell
‘(S)he_i told me about your younger brother_j coming to us tomorrow.’ (i ≠ j)
(11b) [Yiską́ągo ni-tsilí nih-aa doogáł-ígíí] b-aa bi-ł hózhǫ́. (Hale 1999:173)
tomorrow 2SG-YBr 1ns-to FUT.3.go-REL 3-about 3-with A.P.good
‘(S)he_i is happy about your younger brother_j coming to us tomorrow.’ (i ≠ j)
(4.119) Kóhoot'éédą́ą́' shínaaí Naakaii Bikéyahgóó naayáhą́ą yee nihił hólne …anaa' (Speas 1986:319)
last year my-brother Mexico-to went-NOM about-it us tell(OPT) would-that
‘I wish my brother would tell us about his going to Mexico last year’
(4.121) Speas's parse: ‘I wish he_i would tell us about my brother_i's going to Mexico last year.’ (p. 320)
(7) [Jį́į́dą́ą́’ shi-zhé’é łį́į’ nayiisnii’-ę́ę] yídi’doołił. (Hale 1999:171)
‘My father_i will brand the horse he_i bought (earlier) today.’ (coreference OK)
(12) [Yiską́ągo shi-zhé’é shi-má y-íká-’adoolwoł-ígíí] y-ee yi-ł hoolne’. (Hale 1999:173)
‘My father_i told my mother (about the fact) he_i would help her tomorrow.’ (coreference OK)
SOURCED Hale's own words on (10): “(10) cannot mean that my father gave his (i.e., my father's) horse to my younger brother. This is to be expected if Navajo is configurational, of course, since the matrix subject (pro) c-commands all overt nominals (and hence, all R-expressions)” (p. 172). On (11): “if Navajo were nonconfigurational, the prohibited coreference reading should in fact be possible, since the factive complements would be adjuncts and therefore outside the c-command domain of the matrix subject (pro)” (p. 173). On the override: “All of this can be reconciled if we assume simply that Condition C is overriden by the force of (4)” (p. 173); “Condition C, to the extent that it is a true principle of grammar, is evidently true of Navajo” (p. 174). Speas on (4.119): the trip is modified by ‘last year’, so ‘my brother’ “is located within the embedded clause”, and it “must be construed as the subject of this sentence, or at least be construed as coreferent with the subject” (pp. 319–320).
SOURCED Spelling and gloss notes (kept as printed). (11a) glosses y-ee as ‘3-of’ and translates ‘about’; (11b) glosses b-aa as ‘3-about’. (4.119): the nasal hooks in naayáhą́ą are faint and the first letters of the last word are blotted; Speas glosses the clause-final element NOM, where Hale glosses -ígíí REL. Hale's judgments come from “work done nearly twenty years ago by Paul Platero, Ellavina Tsosie Perkins, and others” (pp. 167–168). The source of (4.119) is not stated on the page read.
3. Conventions
Derivational c-command.SOURCED (3) in Abe (following Epstein 1999): “α c-commands β only if β is a term of γ at the stage where α is merged with γ, where a term of γ is γ itself or any syntactic object dominated by γ.” Every step table below therefore lists what α is merged with and what c-command that creates. For Hale's trees (8), (9) the older tree definition gives the same answers.
vP/TP skeleton (VPISH base positions) (the site's convention, derivations). HYPOTHESIS Roles come from Σ (the verb's role schema). The external bundle g1 is merged in Spec,vP inside the verb phrase (VPISH); it can raise inside the word to Spec,M/S (“V-P”), or stay (“V-H”, in situ). Hale does not segment the verb words, so neither do we. Bundles g1, g2, g3 and nouns N1, N2, N3 are our labels. “g → X” means g c-commands X; “⇄” means sisters.
The four views.
Label
Where the nouns sit
Whose view
(a-S) OUTSIDE
Prefixes are the arguments; nouns are adjoined to S, above the subject (Hale's tree (9))
Hale's picture of the adjunct view (Jelinek 1984; Baker 1996 as Hale reads it)
(a-V) ADJOIN-IN
Prefixes are the arguments; each noun is adjoined to the projection of the head whose bundle it links to, before the subject is merged (= FB-3c LINK-IN)
This site's FB, revised 12:46pm
(b) ARG-IN
Each noun is merged in the argument position; the prefix is a clitic or chain tail
Speas 1986/1990 (nouns in θ-positions), Hale's own reading of his data
(c) STRATA
Layer 1 bundles; Layer 2 event-structure roles; Layer 3 noun enters. Three sub-versions in §7
GRANT'S CONJECTURE
4. Derivations
D1. (10), view (b) ARG-IN: the noun fills the argument position
Σ Σ(√GIVE/anim) = {g1 giver (external), g2 recipient, g3 theme} HYPOTHESIS (labels ours)
N1 = si-tsilí N2 = [ts’ídá shi-zhé’é bi-łį́’-ígíí] N3 = shi-zhé’é (inside N2)
S1 Merge V with N2 (theme slot, g3) γ = V
c-command: V ⇄ N2. Inside N2: N3 → bi-łį́’ (possessor over the possessed noun)
S2 Merge N1 (recipient slot, g2) with [V N2]
new: N1 → V, N2, N3 (order of N1 and N2 does not matter for C: neither holds a pronoun)
S3 Merge g1 (external bundle, Spec,vP) with v′ = [v [N1 [V N2]]]
new: g1 → v′, N1, V, N2, N3 ← the relation that matters: g1 → N3
S4 Move V-P: g1 raises to Spec,M/S inside the word; spell-out yeiníłtį́ (not segmented; Hale p. 172)
g1 re-merges with M′ ⊇ all of the above, so g1 → N3 still holds
S5 Index N3 = j ; g1 = i
S6 C g1 c-commands N3 ⇒ i ≠ j
Result: i ≠ j. Matches Hale (10).
D2. (10), view (a-S) OUTSIDE: nouns adjoined to S above the subject (Hale's (9))
S1–S4 Bundles only. g3 in Comp,V; g2 above it; g1 in Spec,vP; V-P raising; spell-out yeiníłtį́
Terms present: V, v, M, g1, g2, g3. N1, N2, N3 are not yet in the derivation.
c-command: g1 → g2, g3, V
S5 Adjoin N2 and N1 to S, the finished clause γ = S
new: N2 → g1, g2, g3, V ; N3 → g1, g2, g3 (a name may c-command a pronoun: no violation)
NOT created: g1 → N3. When g1 was merged, N3 was not a term of γ (Abe (3)).
Tree check, Hale's (9): the first node over g1 does not dominate the adjoined N2.
S6 Link N1 ↔ g2 ; N2 ↔ g3 ; N3 ↔ possessor bi- by feature match (FB-3). No c-command is involved.
S7 Index N3 = j ; g1 = i
S8 C nothing c-commands N3 ⇒ i = j allowed, i ≠ j allowed
Result: i = j should be possible. This is the reading Hale reports as impossible.
(a-S) is contradicted by (10) only if A1, A2, A4, A5, A8 and A9 hold (§5).
D3. (10), view (a-V) ADJOIN-IN (FB-3c LINK-IN)
S1 Merge bundles g3, g2 into VP2 (no nouns yet)
S2 Adjoin N2 to VP2, then N1 to the result, cyclically, before the subject is merged
new: N2 → g2, g3, V (and N3 is now a term of VP2′)
S3 Merge g1 with v′ ⊇ VP2′ new: g1 → N1, N2, N3
S4 Move V-P (as D1)
S5 C g1 → N3 ⇒ i ≠ j
Result: the same as D1. For Condition C, adjoining a noun inside the verb's projection is
indistinguishable from merging it in the argument slot.
HYPOTHESISWhat D1–D3 show. Condition C separates only two placements: above the subject (a-S) versus anywhere under the subject's reach (a-V, b). It does not separate “noun is an argument” from “noun is an adjunct low in the clause”. “Adjunct” and “outside the subject's reach” are different properties (that is A3, §5).
D4. (11a), views (b) and (a-V): a clause under ‘about it’
Σ Inner (go): Σ(√GO) = {m1 mover}. Matrix (tell): {t1 teller (external), t2 addressee, p1 the topic bundle of y-ee}
N = ni-tsilí (index j). Clause C1 = [yiską́ągo ni-tsilí nih-aa doogáł-ígíí]. Labels ours.
INNER I1 Merge N with the inner v′ in Spec,vP as the subject of ‘go’ (view b); m1 links to N; PP nih-aa inside
I2 Move V-P raises m1 inside the inner word; spell-out doogáł
I3 Merge -ígíí (D; H&P 1974) closes the clause: C1 is a DP/relative nominal. N is inside C1.
MATRIX M1 Merge P y-ee with C1 (view b: C1 in the complement of P; p1 is the prefix y- on the P)
new: P ⇄ C1; terms of PP1: P, C1, N, inner clause ...
M2 Merge PP2 shi-ł ‘with me’ with PP1 (addressee), inside the matrix VP
M3 Merge t1 (Spec,vP) with v′ ⊇ PP1, PP2: new t1 → PP1, C1, N ← the relation that matters
M4 Move V-P: t1 raises inside the matrix word; spell-out hoolne’
M5 C t1 c-commands N ⇒ i ≠ j
(a-V) C1 adjoined to PP1 or VP before M3: the same, t1 → N.
Result: i ≠ j. Matches Hale (11a).
D5. (11a), view (a-S) OUTSIDE
INNER as I1–I3, with N adjoined to the inner clause (N links to m1)
MATRIX M1 Merge bundles t1, t2, p1 and the P words y-ee, shi-ł; spell-out hoolne’. Terms: t1, t2, p1, V, v …
M2 Adjoin C1 to S (above t1), linked to p1 γ = S
new: C1 → t1 ; N → t1. NOT created: t1 → N (N was not a term when t1 merged)
M3 C nothing c-commands N ⇒ i = j allowed
Result: i = j should be possible, against Hale (11a). Contradicted only given A1, A2, A4, A5, A7, A8, A9.
D6. (11b): the experiencer is inside a postposition (a new complication)
Σ Matrix: Σ(√GOOD, stative) = {z the thing that is good (the clause), e the experiencer}
The printed gloss is “3-about 3-with A.P.good”: the pronoun that Hale translates as the ‘(S)he’ is the
3 of bi-ł ‘with (him/her)’, the object of a postposition. Not the verb's own subject prefix.
E1 Merge P b-aa with C1 (the ‘about it’ clause) PP1 = [b-aa C1]
E2 Merge P bi-ł with its pronominal bundle e PP2 = [bi-ł e] (e is inside PP2)
E3 Merge both PPs into VP; verb hózhǫ́
c-command: e → (only the P bi-ł). The first node above e is PP2, which does not contain PP1.
So e ↛ C1, e ↛ N, unless e is raised to a subject position, or the VP is a shell in which PP2 is above PP1.
E4 C if e ↛ N: i = j allowed. if e → N (A10): i ≠ j.
Result: (11b) gives i ≠ j only with an extra assumption (A10): the experiencer PP c-commands the clause.
SOURCED Willie (1991:47–48, 105–106) says the Experiencer of psych verbs is “the object of a grammatical postposition” and “comparable to a ‘Dative Subject’”, and, for the ‘know’-type psych verbs, that “these constructions permit adjoined subordinate clauses” (p. 49) (Who likes it?). HYPOTHESIS On this site's own VPISH analysis of psych verbs, nothing raises to subject: the Experiencer stays in its PP. Under that analysis (11b) is not a clean Condition C test, so it should not be counted as a second, independent datum alongside (11a). The earlier page CT3 said (11b) “is the same” as (11a); that was too quick.
D7. (4.119), Speas's contrary judgment
Σ Inner (go): {m1}. Matrix (tell, optative): {t1 teller, t2 addressee 'us', p1 topic of yee}
N = shínaaí ‘my brother’. Clause C1 = [kóhoot'éédą́ą́' shínaaí Naakaii Bikéyahgóó naayáhą́ą] (Speas p. 319)
INNER ‘last year’ modifies the trip, so N is inside C1 (Speas p. 319): N is the subject of ‘went’ (m1 links to N)
MATRIX as D4: PP1 = [yee C1]; PP2 = nihił ‘with us’; t1 external
View (b)/(a-V): t1 → PP1, C1, N ⇒ C applies ⇒ i ≠ j
View (a-S): C1 above t1 ⇒ t1 ↛ N ⇒ i = j allowed
Speas's judgment: i = j is the reading actually used (4.121).
HYPOTHESISWhich views match which judgments.
View
(10)
(11a)
(11b)
(4.119)
(a-S) OUTSIDE (above the subject)
free ✗
free ✗
free ✗
free ✓
(a-V) ADJOIN-IN
i ≠ j ✓
i ≠ j ✓
free ✗ unless A10
i ≠ j ✗
(b) ARG-IN
i ≠ j ✓
i ≠ j ✓
free ✗ unless A10
i ≠ j ✗
Reported judgment
i ≠ j (Hale)
i ≠ j (Hale)
i ≠ j (Hale)
i = j (Speas)
No view matches all four. The inside views match Hale and need a device for Speas; Speas's own device is across-the-board (ATB) coindexing tied to a “unified representation” of adjunction constructions (§4.7.6.3, pp. 319–320). The outside view matches Speas and needs a device for Hale. This is T24a.
5. The assumption audit
For each assumption: what Hale says, whether anything other than these same data supports it for Navajo, where it could fail, and what follows if it does. “If it fails” never means the data disappear. (10) and (11a) are reported facts. It means Hale's inference to where the nouns sit is blocked, and the disjointness would need another explanation. That makes the argument an inference to the best explanation, not a refutation.
#
Assumption
Hale's words (page)
Independent support?
Where it could fail
If it fails
A1
Condition C, a c-command constraint, is what makes (10) and (11) disjoint; nothing else does
“(10) and (11) show that Condition C is obeyed in Navajo” (p. 173); “to the extent that it is a true principle of grammar, is evidently true of Navajo” (p. 174)
Partly. Condition C is well attested across languages (not tested here). In Navajo the only support is (10) and (11) themselves, so it cannot also be used to choose among placements without circularity. Speas treats (4.15d) as a Condition C case, but rank also excludes it (Speas page)
Hale's n. 3 grants that such conditions can be overridden by pragmatics (citing Grimshaw & Rosen 1990, not opened); Mohawk shows no such effect (p. 172); fourth-person or obviation effects are untested (T24c)
The disjointness has another source; nothing about noun placement follows
A2
The binder is a syntactic object in the subject position that is visible to binding: a null pro, or the prefix bundle g1
Trees (8), (9) put a free-standing “Proi” as sister of VP (pp. 171–172); “the matrix subject (pro) c-commands all overt nominals” (p. 172)
Partly. Hale's PAT (16) and Willie call the prefix a full argument; Speas treats prefixes as infixes visible to syntax (pp. 200–202). Nobody tests whether a word-internal prefix binds outside the word
If the prefix is agreement with no index, or if only a D-level pronoun can be a Condition C binder and the bundle is φ-level (Déchaine & Wiltschko 2002, a variable)
No binder, so no Condition C; the data still need an explanation
A3
An adjunct noun is outside the subject's c-command domain (equivalently, FB-5: “linked, not merged” means no c-command)
“the factive complements would be adjuncts and therefore outside the c-command domain of the matrix subject” (p. 173)
No as a blanket claim. Only S-adjunction, as in Hale's (9), gives this. 8:46pm, revised 9:20pm: Jelinek 1984 tree (28) is pre-VPISH notation for Warlpiri (flat S: V, AUX, CPP; no VP, no Spec,vP), so it gives no verdict on A3: her notation cannot be translated directly (page). Later restatements in structures with a low external argument say nouns are adjuncts “to the clause” (Hale 2003:51; Ormston 1993 on Baker) but draw no tree for the nouns (T26g, progress). Baker's later n. 16 leaves Navajo open and contemplates PP/vP adjunction (against A3); Baker 1996 still unopened. Opened English counter-evidence: *“He1 sent Zelda back all the flowers which she had bought for Dr. Levin1” (Abe, ex. (5)). 2026-10-16 (T27i): restated site by site, A3 holds in 5 of 10 non-argument configurations and survives only as a timing condition, A3′: a noun is outside the subject's reach iff it enters after the subject's last Merge (Condition C by site §6.1)
Adjuncts can adjoin to VP or PP, inside the subject's reach (D3)
Hale's data no longer separate adjunct from argument; only “above the subject” is excluded. This is the weakest premise and the one the FB-5 verdict turns on
A4
No other factor excludes the i = j reading: no competing way to say ‘his own’, no pragmatics, no effect of tsʼídá ‘very’ in (10)
None stated; n. 3 allows pragmatic overriding (p. 174)
Untested. No controls or minimal pairs in the paper
A dedicated form for ‘his own horse’, if one exists, could block the plain reading (we have no data on this)
The exclusion is pragmatic, not structural
A5
The judgments are stable across speakers and contexts
Data from “work done nearly twenty years ago by Paul Platero, Ellavina Tsosie Perkins, and others” (pp. 167–168)
Unverified. No sample, contexts or minimal pairs given. Speas's (4.119), same shape, has the opposite judgment (T24a)
Speakers differ, or mood, tense or the possessor's person matter (§6)
The empirical base shrinks to a tendency in some constructions
A6
The override (4)+(13) applies exactly to (7) and (12) and never to (10), (11): it needs a transitive inner verb X
(13): X “may be a transitive verb”; nouns “can, and preferably will” be read as shared (p. 173). (4) is “essentially inviolable” (p. 170)
No. A stipulation from Platero 1978, 1982 (not opened). The line is thin: (12)'s inner verb y-íká-ʼadoolwoł (‘3-for-FUT.3.run’) has a postposition inside the verb word and counts as transitive, while (11a)'s inner verb doogáł has a separate PP nih-aa and counts as intransitive
If transitivity is not the right test, or the override is a soft preference (“preferably”), it may reach (11) too, and (4.119)
Condition C is not categorical in Navajo; A1 weakens
A7
The nominalized clause or PP sits in, or reconstructs into, the matrix VP: not left-dislocated above the subject
Tree (8) puts the relative clause inside VP. For (11) Hale gives no tree and states the c-command (pp. 172–173)
Untested. The bracketed clause is sentence-initial in (11), and a null subject has no linear position, so order cannot show where it sits
If it is above the subject, Condition C needs reconstruction. Whether reconstruction applies for Condition C is itself contested in English (surveys cited in abstracts: Stockwell et al. find it, Bruening & Al Khalaf do not; not opened beyond abstracts)
No c-command, no Condition C in (11)
A8
The nouns are argument-type R-expressions (DPs), not predicate-like nominals
Not discussed; the nouns are treated as R-expressions (p. 172)
Plausible, untested. Déchaine & Wiltschko's pro-NP is a predicate, not an R-expression
If nominals in a pronominal-argument language are predicate-like modifiers of the slot
No R-expression, no Condition C
A9
Condition C is checked on the finished representation, with the noun in the position tested
Implicit (trees (8), (9))
Standard but not Navajo-specific. Condition C as an “everywhere” condition is Lebeaux's idea (via Stepanov, NELS 30, snippet). Late merge changes the outcome (Abe (3), (4))
If a noun enters after the check (STRATA, §7)
No violation even with the noun below the subject
A10
(11b) only: the experiencer pronoun in bi-ł ‘with (him/her)’ c-commands the ‘about it’ clause
“the matrix subject (pro)” (p. 172); the gloss is “3-about 3-with A.P.good”, with the pronoun inside a postposition
No. This site's psych-verb analysis (Willie 1991) keeps the Experiencer inside its PP (page)
A PP-internal pronoun does not c-command a sister PP in the simplest structure (D6)
(11b) drops out as independent evidence
HYPOTHESISCount. Ten assumptions. None is independently established for Navajo. Three are specifically weakened by evidence we opened or already have: A3 (English adjunct data; Baker later n. 16 contemplates PP/vP for Navajo; Jelinek 1984's flat-S tree is pre-VPISH notation and gives no verdict), A5 (Speas (4.119)), A10 (the site's psych-verb analysis). A6 is a stipulation. A1, A4, A8 and A9 are plausible but untested for Navajo.
What Hale's argument gets right. (i) The inference itself is sound. (ii) (10) is a clean case: one verb, two people, a possessor inside the theme phrase, no postposition, no parallel verb, and its structure does not depend on A6, A7 or A10. If its judgment is stable, the best available explanation is Condition C with the noun under the subject's reach. That is a real result, and it is the one FB-5 has to answer. (iii) The paper correctly identifies that adjunct status alone must not be assumed to exempt a noun, which is what A3 says in the other direction.
6. Speas's (4.119) against Hale's (11a): what differs
Feature
Hale (11a)
Speas (4.119)
Judgment
i ≠ j
i = j is the reading used (4.121)
Same in both
one noun; an intransitive inner verb (‘go’) in a nominalized clause; a postposition ‘about it’ under a verb of telling; a null matrix subject
Mood of the matrix verb
perfective, ‘told’ (3.P.tell)
optative, ‘would tell’, with …anaa’ ‘would-that’ (a wish)
Time
‘tomorrow’, future (FUT.3.go)
‘last year’, past (went-NOM)
Person of the possessor on the noun
2SG, ‘your younger brother’
1SG, ‘my brother’
Clause-final element, as glossed
-ígíí REL
NOM
Extra overt material in the clause
‘to us’ as a postposition with a pronoun (nih-aa)
an overt goal noun (‘Mexico-to’)
Source of the judgment
Platero, Perkins and others (pp. 167–168)
not stated on pp. 319–320
Device to get the other reading
none: (13) needs a transitive inner verb
ATB coindexing / unified thematic representation (§4.7.6.3); Speas says Condition C is “found here too” and still lets the reading through (pp. 319–320)
OPENWhich differences could matter (none tested; no Navajo is constructed here): (1) the optative wish context, in which the speaker's perspective may change who counts as ‘he’; (2) the past versus future time of the clause; (3) a first-person possessor on the noun (‘my’) against a second-person one (‘your’): the speaker is in the picture in (4.119); (4) the different nominalizers; (5) the judgment itself, which may vary by speaker. Resolving T24a needs minimal pairs that change one of these at a time with the same speakers. Until then A5 is open, and any use of (10) and (11) or of (4.119) as the deciding case is premature.
Two sources, same shape, opposite judgments, each with its own override: this is what a field with unsettled data looks like, and it is the situation in which a single account (Hale's, Speas's or ours) should not be treated as the judge.
7. STRATA (Grant's layered conjecture)
GRANT'S CONJECTURE Argument structure is built in layers. Layer 1: theta roles or feature bundles, always applying (the pronominal prefix slots). Layer 2: other event-structure roles (Levin & Rappaport Hovav style). Layer 3: lexical-item selection: the noun enters as a later layer of the derivation. Nouns are not “outside” the clause; they come in last.
HYPOTHESISMapping to FB. L1 = the bundles of FB-1 (person, number, capacity κ, and an empty role slot). L2 = Σ assigns the role ρ and the external/internal split; voice (yi-/bi-) and V-P raising happen here or at its edge. L3 = a noun is chosen and linked by feature match (FB-3). The layering is Grant's; the identification with FB's parts is ours.
“Enters later” does not say where. Three sub-versions follow, with the c-command relations at the point where the noun enters, for (10) (the others are parallel).
L1 Skeleton, always: [vP g1 [VP g2 [V′ V g3]]] c-command at L1: g1 → g2, g3, V
L2 Σ(√GIVE/anim) fills ρ on g1, g2, g3; voice; (V-P raising of g1 is a parameter, see below)
c-command at L2: unchanged.
L3 Noun entry. State at entry: g1 already c-commands the whole VP.
S-i SLOT-FILLING. N2 is placed in the slot g3; N1 in the slot g2 (DP shells with open values already in the tree).
At entry, g1 → slot(g3) → N2 → N3.
C evaluated after L3 ⇒ g1 → N3 ⇒ i ≠ j (same as D1/D3)
C evaluated at L2, before L3 ⇒ N3 absent ⇒ no check ⇒ i = j allowed (depends on A9)
S-ii LATE MERGE inside the same phase, after g1 has merged (Lebeaux; Abe).
When g1 merged with v′, N3 was not a term, so g1 ↛ N3 (Abe (3)).
If g1 then re-merges higher (V-P raising to Spec,M/S, outside the vP phase) with M′ ⊇ N2, g1 → N3 ⇒ i ≠ j.
If g1 does not raise (V-H, in situ), or raises before the noun enters, g1 ↛ N3 ⇒ i = j allowed.
S-iii NOUN ENTERS AT THE DISCOURSE / PF LAYER, outside the syntax's binding computation.
No syntactic position under g1 ⇒ g1 ↛ N3 ⇒ i = j allowed (same as D2, view a-S)
HYPOTHESISDoes Condition C then apply to the hidden subject prefix and the noun? It depends on the sub-version. S-i: yes (if Condition C is read after Layer 3). S-iii: no. S-ii: yes if the bundle re-merges above the noun after it enters (V-P with the noun entering inside the vP phase); no if the bundle stays low or raises first. S-ii is the one place STRATA can predict something that neither (a-S) nor (b) predicts, and it needs two extra assumptions (the phase boundary and the order of raising and noun entry) that no Navajo datum fixes. The mechanism comes from English work on late-merged adjuncts, where even the empirical asymmetry is contested.
List 1 holds roots and bundles of grammatical features; syntax combines them; List 2 (Vocabulary) supplies phonological forms “for roots as well as bundles of grammatical features, unless roots come with their phonological forms from the narrow lexicon”; List 3 (Encyclopedia) lists special meanings. Whether roots are late-inserted is “orthogonal” and left open. The head that projects the agent is a boundary for special meanings: “No idioms with fixed agents” (6), (7a)
Full text
Supports separating the argument skeleton (including the agent head) from root content. Says nothing on binding
Embick & Noyer, “Distributed Morphology and the Syntax/Morphology Interface” (Oct. 2004 draft; Arizona copy)
Abstract morphemes get their phonology by Vocabulary Insertion after syntax (Late Insertion). Roots, in this chapter, “are present with all of their features throughout the derivation, and require no such insertion process” (§2.2, with fn. 9). Their Hupa (Athabaskan) example (from Golla 1970) treats subject and object person-number prefixes as spell-outs of feature bundles such as [+1 +PL +SUBJ] ↔ di-, [+PL +Obj] ↔ noh-
Full text
Precedent for Layer 1 (the prefixes as exponents of syntactic bundles), in an Athabaskan language. Not for Layer 3: in this version the noun's root is in the tree from the start
Haugen & Siddiqi, “Roots and the Derivation”, Linguistic Inquiry (doi 10.1162/ling_a_00136)
Abstract only: argue against Early Root Insertion and for Late Insertion of Roots “just like other morphemes”, on evidence from root suppletion and hyponymous direct objects in noun incorporation
Abstract only
The one DM source found that puts the root late. Even so, the root's terminal node would be in the structure, so c-command is not affected. Not read for binding
Borer 2005, Structuring Sense I–II, as described by Potts's review (Language)
Listemes are “raw material” “poured into the structural mould” (I:108, quoted); “Only when the derivation is over … can the conceptual packages be opened” (I:12, quoted); arguments are introduced by functional heads and the verb is an event modifier (II); proper-name versus common-noun reading comes from the functional structure (I, Ch. 3). Potts notes that it is hard to see how a verb could impose requirements on its arguments in this system
Review only; the books were not opened
Closest to S-i: the content comes last, but the DP shell with its open value is in the derivation. Binding is not discussed in the review
Ramchand 2008, Verb Meaning and the Lexicon, via her slides “Argument Structure for the 21st Century” and Bruening's review (J. Linguistics 46, 2010)
Verbs decompose into init, proc, res heads with arguments in their specifiers; roles come from position; “Thematic roles are not listed and memorized as part of verbal lexical entries”; nominal projections connect to the clause by “property predication”. Embedding in the event structure is respected in realization (Levin & Rappaport Hovav 2005:183, quoted in the slides). Bruening: unlike radically constructionalist models, category information stays with lexical items
Slides and review; the book was not opened
Layer 2. Nominal DPs merge in the specifiers: S-i
Speas 1990, Warlpiri (§3.4.3, pp. 167–172)
Verbal grid positions are discharged only by pronominal clitics; nominals (KPs) are modifiers linked by theta-identification
Opened in earlier runs
A two-layer arrangement already: pronouns first, nouns linked afterwards, inside a configurational tree
Late merge of adjuncts: Lebeaux 1988, Epstein 1999, via Abe, “Late Merge and Phase Theory” (CRISSP); Stepanov, NELS 30 (snippet)
Adjuncts can be merged counter-cyclically, which avoids Condition C after A-bar movement; derivational c-command (3); late merge is limited by the phase (Abe (4), (5))
Abe in full; Lebeaux, Epstein not opened
The mechanism behind S-ii. Contested empirically (abstract-level: Stockwell et al. for the asymmetry, Bruening & Al Khalaf against)
HYPOTHESISWhat the precedents support. Every published layered model we opened builds the structural skeleton first and adds root or listeme content last. In each, the structure for the noun (a DP shell or a specifier) is in the derivation from early on. That is S-i. None puts the noun and its shell outside the subject's reach, and none, as far as opened, says when Condition C is read.
Does STRATA make different predictions from the inside-the-clause view? Honestly: not on any datum we have. S-i equals inside. S-iii equals outside. S-ii differs from inside only in the corner where the bundle stays low or raises before the noun enters. What could separate STRATA from the inside view, none of it available yet: (1) a fronted noun above the subject, where late merge predicts no reconstruction for an adjunct-type noun; (2) a noun whose features clash with the slot (number, person or κ): STRATA predicts a Layer-3 matching failure, the inside view a role-assignment failure; (3) noun-less clauses, where all views agree. STRATA is still a useful idea: it states the Grant/FB intuition that the argument exists before any noun, and it shows exactly which extra timing assumption would exempt nouns from Condition C.
8. Verdict
Question
Answer
Is Hale's argument valid?
Yes. Derivations D1–D5 confirm that nouns under the subject's reach give i ≠ j and nouns above the subject give free coreference.
Does it hold?
Conditional on A1–A9 (and A10 for (11b)). It excludes one placement: nouns adjoined above the subject, as in his tree (9). It does not establish that nouns are arguments rather than adjuncts (needs A3), and it does not touch FB-1 to FB-4.
Is it independently supported?
No premise is independently established for Navajo. A3 meets counter-evidence in English; A5 meets Speas's (4.119); A10 meets the site's psych-verb analysis; A6 is a stipulation.
Does it disprove FB-5?
No. FB-5's exemption (the noun is outside the bundle's reach) is in tension with (10) and (11a), conditional on A1, A2, A4, A5, A7, A8, A9. That is an inference to the best explanation, with Speas's (4.119) pointing the other way. The earlier statements “FB-5 is wrong”, “D6 withdrawn” and “FB has to go inside” were stronger than the evidence.
Status of FB after this audit
COMPATIBLE, UNDECIDED on noun placement and Condition C (D6). LINK-IN, FB-5′ and FB-6′ are available as conditional additions: they are needed only if the A-set holds. (10) is the cleanest case and the next thing to test.
STRATA
Same predictions as inside (S-i) or outside (S-iii); a corner of its own in S-ii. Precedents support S-i.
9. Next steps (T25)
#
Step
Status
T25a
Settle the judgments: elicit (10), (11a) and (4.119) in minimal pairs that change one variable at a time (mood, time, possessor person, nominalizer), with the same speakers, and with a control for a dedicated ‘his own’ form (A4, A5)
OPEN speakers
T25b
Read Platero 1978, 1982 for why the parallel parse needs a transitive inner verb and why y-íká-ʼadoolwoł counts (A6)
OPEN
T25c
Find where Baker 1996 and Jelinek 1984 attach the nominals. If they attach below the subject, A3 fails for those authors and Hale's target moves (A3)
NOT YET ANSWERED (8:46pm partial; revised 9:20pm): Jelinek 1984 tree (28) is pre-VPISH notation, so no verdict on A3; restating the PAH in VPISH/TP/CP structure is T26g (progress); Baker 1996 still unopened; Baker later n. 16 contemplates PP/vP adjunction for Navajo (against A3). See /jelinek-baker-adjunction/
T25d
Binding tests from a prefix: can a word-internal prefix c-command out of the word, or be a binder at all? Preminger-style clitic/agreement tests (A2)
OPEN
T25e
Condition C cases with a bare subject prefix and a clear in-clause noun, no postposition, no nominalized clause, like (10), to be gathered from published sources
OPEN
T25f
Read Speas 1990 Ch. 4 (§4.3.4, pp. 221–237) for her Parallel Function Constraint and how it treats (4.119)-type sentences
OPEN not available
T25g
STRATA: find a datum that separates S-ii from inside (a fronted noun above the subject; a feature clash between noun and slot); read Borer's and Ramchand's books for binding
OPEN
T27i → T28
Re-run D1–D7 with modern attachment sites and derivational c-command
Restated on the unified structure (2026-11-15). Nouns are restated as arguments merged in their theta positions (VP, Spec,vP) and, for the subject, raised to Spec,TP by the working default; a lone noun is linked by the H4″ default. Adjunction and pronominal-argument analyses below are labelled contrast; where nouns sit for Condition C is to be re-run on the unified spine. Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). Flat-S and other pre-VPISH notation below is labelled contrast. Status on the spine: see unified structure, §2 and §4, first rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.
2026-11-11 audit pass 2. Wording that used VPISH as the name of a spine, tree, skeleton or set of slots named a structure after the hypothesis. VPISH is a hypothesis name, never a node; the structure is a vP/TP spine with the subject first-merged in left Spec,vP. Only the wording changed.
2026-11-11 audit pass 3. Status badges now use only the label set SOURCED, HYPOTHESIS, SPECULATIVE, OPEN, NONE and GRANT'S CONJECTURE; verdict or task words that had been styled as badges (for example Opened, PRE-VPISH, DONE, SUPPORTS, PARTLY ANSWERED) are now plain text or a label with a qualifier. Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference. See the pass 3 report.
Hale 1999's Condition C argument turns on assumption A3: if nouns are adjuncts, they sit outside the subject's reach. That only holds if the adjuncts hang above the subject, as in Hale's tree (9). So T25c asked: where do Jelinek 1984 and Baker actually attach the nominals? This page reports what the opened sources say. Written 2026-10-15, 8:46pm run (PT). Revised 9:20pm: Jelinek's 1984 notation turns out not to answer the question (see the notice below).
HYPOTHESISRevision, 2026-10-15, 9:20pm (Grant's point). Jelinek 1984's tree (28) is drawn in pre-VPISH notation: a flat S that is subdivided directly into V, AUX and CPP daughters, with no VP and no TP or CP. Her own text works in 1984 Government-and-Binding terms: “INFL governs the subject”, and “in a configurational language … objects appear in a VP constituent and subjects do not” (p. 44, page-checked). The VP-internal subject hypothesis came a few years later (Fukui & Speas 1986, Kuroda 1988, Koopman & Sportiche 1991, per search-result snippets; those papers were not opened). So tree (28) is a notation for Warlpiri, not a datum about Navajo, and it cannot be read off directly as a statement about whether a subject merged in Spec,vP reaches a noun. The 8:46pm finding “A3 holds for Jelinek 1984” is downgraded to: her notation cannot be translated directly; no verdict. Restating the pronominal-argument hypothesis in a VPISH/TP/CP structure is the new open task T26g (§12, with what has been opened so far).
Ground rules (Grant's). Jelinek 1984 is about Warlpiri, not Navajo. Baker's later paper leaves Navajo open. Baker 1996 The Polysynthesis Parameter itself was not opened. No Navajo Condition C data appear in either newly opened source. Every Warlpiri form below is copied from a page image of Jelinek 1984 with its example number; nothing Navajo is invented. Hale, Speas, Willie, Jelinek and Baker are each treated as possibly partly right.
SOURCEDJelinek 1984 (Warlpiri): Nominals are optional Case Particle Phrases (CPPs) that are sisters to Verb–AUX under S. Tree (28), pp. 50–51; restated p. 73: “CPPs are sisters to the verb under S.” Parameter (73b): nominals “are not verbal arguments, but are optional adjuncts to the clitic pronouns that serve as verbal arguments.” Primary L-case = adargumental adjunct; secondary L-case = adsentential adjunct (p. 51).
OPENC-command (revised 9:20pm): no verdict. Read literally, in tree (28) the subject clitic sits inside AUX and the CPP is a sister of AUX, so a first-branching-node reading gives no c-command from the clitic to the CPP. But that applies a modern c-command definition to a notation with no VP, no Spec,vP, no TP or CP, in which AUX is a constituent Jelinek treats as holding INFL and the clitics. The notation does not say whether the CPP would attach above or below the position where the subject argument is merged in a vP/TP tree. So there is no verdict on A3 for Jelinek 1984. The 8:46pm wording “A3 holds for Jelinek 1984's Warlpiri analysis” is withdrawn.
SOURCEDBaker's later footnote (n. 16, PDF p. 47): Leaves Navajo open. Two options: (i) Navajo is IE-like (agreement does not cause dislocation), as Rice assumes for Slave; (ii) Jelinek is right and “dislocated NPs can adjoin to a wider range of categories in Navajo than in Slave or Kinande, including PP and vP.” Option (ii) would put nouns inside the subject's reach → A3 fails for that Baker speculation. Baker 1996 trees still unopened.
OPENT25c verdict (revised 9:20pm): NOT ANSWERED from Jelinek 1984; partial from later sources. Jelinek 1984 gives no usable attachment height (pre-VPISH notation). Baker's later footnote does not commit to S-adjunction for Navajo and contemplates PP/vP adjunction (against A3). Hale 2003, Ormston 1993 on Baker, and Frantz's review of Baker 1996 say nouns are adjuncts “to the clause” in structures that do have a VP-internal subject, but none draws a tree for the nouns (§12, T26g).
2. Source and caveats
Source
What it is
What we used
What we did not
Jelinek, Eloise 1984. “Empty Categories, Case, and Configurationality.” NLLT 2:39–76
The original Pronominal Argument Hypothesis paper. About Warlpiri.
Box: [local working copy, not in the repository] (Carnie & Harley ms collection PDF; paper starts PDF p. 8 = printed p. 39). Text extract jel84.txt. Page images: tree-019.png (p. 50 with tree (28)), tree-020.png, j-*.png, j2-*.png, end-*.png (pp. 73–74)
No Navajo adjunction-height claim. Navajo is mentioned only via Hale, Jeanne & Platero 1977 for an NP hierarchy (around printed p. 66) — not for where adjuncts attach
Baker, Mark C. “Agreement, Dislocation, and Partial Configurationality.” Rutgers ms (~2001–2005; Utrecht workshop 2001; later in a Jelinek festschrift)
A later paper on when agreement forces dislocation
Box: [local working copy, not in the repository] (agreementdislocation.pdf, agreement.txt, footnote page pg-47.png; n. 16 begins on pg-46.png)
Baker 1996 The Polysynthesis Parameter was not opened. Only this later paper's footnote on Navajo
SOURCED Earlier this project had opened Jelinek 1984 only for the PAH statement on p. 44 (clitic pronouns as arguments; nominals as optional adjuncts). The tree (28), the Case Compatibility rules, and the Configurationality Parameter (73) are new to this run.
3. Jelinek 1984: the tree and the parameter SOURCED
3.1 The key prose (p. 50)
SOURCED Verbatim from the page image (tree-019.png, printed p. 50):
Warlpiri nominals are adjuncts to the Verb-AUX complex, which constitutes a complete finite sentence. They are governed by their case particles/postpositions, forming Case Particle Phrases that are sisters to the Verb-AUX:
3.2 Tree (28) SOURCED(older author's pre-VPISH notation; contrast only, not a datum about Navajo)
S
/|\
/ | \
/ | \····· (dashed = optional)
V AUX CPP
(+Tense/ /|\ / \
Aspect) T S O Nominal CP
Legend (as printed, p. 50):
T = Tense/Aspect/Modality
S = Subject Clitic; NOM case
O = Object Clitic; ACC case
CP = Case Particle; (ERG, ABS, DAT, LOCATIVE, etc.)
--- = Optional
HYPOTHESIS Labelling note (9:20pm): tree (28) is 1984 GB notation for Warlpiri: S branches directly into V, AUX and CPP, and the subject is a clitic inside AUX, not a specifier of a VP or vP. It is not evidence about Navajo, and “sister of Verb–AUX under S” has no direct counterpart in a VPISH/TP/CP tree.
SOURCED Restated on p. 73: “CPPs are sisters to the verb under S.” The order of CPPs is free; AUX is second-position; “any case particle phrase (CPP) may appear in the sentence initial position, whereupon the verb appears after AUX, with no fixed order with respect to any CPPs present” (p. 51).
3.3 Adargumental vs adsentential (p. 51)
SOURCED Primary L-case (ERG, ABS, DAT) “marks a nominal as an adargumental adjunct, as giving more information on the referent of a clitic verbal argument.” Secondary L-case (LOCATIVE, PERLATIVE, ALLATIVE, ELATIVE, etc.) marks an adsentential adjunct: it “cannot be coindexed with a clitic pronoun” (p. 51).
3.4 Linking Rule (31) and Case Compatibility (32)
(31) Linking Rule
A clitic pronoun may be coindexed with a nominal, providing
the L-case of the nominal and the G-case of the clitic pronoun
are compatible (assigning a distinct index to each clitic).
(32) Case Compatibility Rule
a. NOM G-case is compatible with ABS and ERG L-case.
b. ACC G-case is compatible with ABS and DAT L-case.
c. DAT G-case is compatible with DAT L-case.
SOURCED (pp. 52–53). Coindexing is not bi-unique: more than one nominal, or none, may link to a clitic.
3.5 Configurationality Parameter (73)
(73) Configurationality Parameter (Extended):
a. In a configurational language, object nominals are properly
governed by the verb.
b. In a W-type non-configurational language, nominals are not
verbal arguments, but are optional adjuncts to the clitic pronouns
that serve as verbal arguments.
SOURCED (p. 73). “Whereas grammatical relations are defined configurationally in (73a), there is no asymmetry between subject and object in (73b)” (p. 74).
3.6 Warlpiri examples (keep forms as printed)
(26) *ngajulu-rlu ka-ngku-rna nyuntu-Ø nya-nyi (p. 50; starred)
I-ERG PRES-2sgACC-1sgNOM you-ABS see-NONPAST
(Clitic order is fixed: SUBJECT before OBJECT; reversing it is out.)
(35) ngaju-Ø ka-rna wangka-mi (p. 54; Hale 1983:18)
I-ABS PRES-1sgNOM speak-NONPAST
‘I am speaking.’
(37) ngajulu-rlu ka-rna-ngku nyuntu-Ø nya-nyi (p. 54; Hale 1983:18)
I-ERG PRES-1sgNOM-2sgACC you-ABS see-NONPAST
‘I see you.’
SOURCED Third-person ABS is often ZERO (printed Ø / unmarked). These are Warlpiri only. No Navajo parallel is invented here.
4. Does the subject c-command the adjunct? Literal reading only; no verdict OPEN
Jelinek does not discuss Condition C or c-command from the subject clitic onto a CPP. The question is ours, asked because Hale's A3 needs an answer.
Tree (28), zoomed on c-command (LITERAL reading of a pre-VPISH notation; not a verdict):
S
/|\
/ | \
V AUX CPP Under first-branching-node c-command:
/|\ • AUX and CPP are sisters under S.
T S O • S (subject clitic) is a daughter of AUX.
• So S's first branching node is AUX.
• CPP is not dominated by AUX.
• Therefore S does NOT c-command CPP.
Contrast Hale 1999 tree (9) picture of the adjunct view
(OUTSIDE / a-S on the derivation page):
S
/ \
Pro S' Pro does not c-command adjuncts on the higher S.
/ \
Adj VP
HYPOTHESISDowngraded 9:20pm: her notation cannot be translated directly; no verdict on A3. The literal reading above is what the 1984 diagram says if one applies a first-branching-node definition to it. It is not the same as saying the CPP is adjoined above a Spec,vP subject. In tree (28) there is no Spec,vP, no VP and no TP for the CPP to be above or below. Hale's (9), Baker's option (ii) and this site's LINK-IN are all drawn in structures that Jelinek 1984 does not have.
OPEN Alternative definitions of c-command (e.g. m-command from the maximal projection containing the clitic) could change even the literal answer (T26e). The more basic problem is the missing structure: that is T26g.
5. Baker's later footnote on Navajo SOURCED
SOURCED Baker 1996 itself was not opened. What we have is a later Rutgers manuscript, “Agreement, Dislocation, and Partial Configurationality” (Utrecht workshop 2001; later collected in a Jelinek festschrift). Footnote 16 begins on PDF p. 46 and continues on p. 47. The Navajo paragraph (verified on pg-47.png):
I leave open the question of how these ideas apply to Navajo, in which agreement is obligatory. The obvious possibility is that Navajo is an IE-like language, with agreement not causing dislocation, as Rice assumes. But it could also be that Jelinek is right to analyze Navajo as a pronominal argument language, and the difference is that (for some reason) dislocated NPs can adjoin to a wider range of categories in Navajo than in Slave or Kinande, including PP and vP.
Baker option
What it says
Effect on A3
(i) IE-like
Agreement does not cause dislocation (as Rice assumes for Slave)
Nouns need not be adjuncts at all; A3's premise (adjunct status) may not apply
(ii) Jelinek + wider adjunction
Dislocated NPs adjoin to a wider range, “including PP and vP”
A3 fails for that speculation: PP/vP adjuncts sit inside the subject's reach (cf. D3 on the derivation page)
SOURCED Baker does not choose between (i) and (ii). He does not draw a Navajo tree. He does not say Navajo adjuncts are S-sisters. The PP/vP option is what he contemplates if Jelinek is right for Navajo.
6. Effect on Hale 1999 A3 and FB-5
Claim
Status after this page
A3 for Jelinek 1984 (Warlpiri)
OPENNo verdict (revised 9:20pm). Pre-VPISH notation; the literal reading of tree (28) gives no c-command from the clitic to the CPP, but that is not a statement about attachment relative to a Spec,vP subject
A3 for Baker 1996
OPEN Baker 1996 not opened
A3 for Baker later n. 16 on Navajo
SOURCED speculation leaves it open; option (ii) would falsify A3
Hale's target (adjuncts above the subject)
Still Hale's own picture (tree (9)). Whether any PAH author draws nouns that high in a VPISH/TP/CP structure is T26g. Not necessarily the right target for Baker on Navajo
FB-5 (linked noun outside the bundle's reach)
Unchanged: in tension with Hale (10)/(11a) only if A1–A10 hold. Jelinek 1984 gives no support for the “outside” half of A3 (no verdict); Baker n. 16 does not settle Navajo
Condition C data in these sources
None for Navajo. Warlpiri examples only
HYPOTHESIS Plain words: Jelinek's Warlpiri tree is a 1984 diagram with no VP or TP, so it cannot tell us whether the nouns hang above or below the subject in a modern tree. Baker's later footnote says that for Navajo, if the nouns are adjuncts at all, they might hang lower (on PP or vP), which would put them back inside the subject's reach. Later restatements say the nouns are adjuncts “to the clause” without drawing them (§12). Until the PAH is restated in a VPISH/TP/CP tree and Baker 1996 is opened, A3 stays the weakest premise of the audit.
7. Rule set A vs B note
HYPOTHESIS Jelinek's own contrast maps onto this site's A/B split:
AInside AUX / the verb word: clitic order is fixed (SUBJECT before OBJECT; *(26)). That is left/fixed morphosyntax — word-internal linearization.
BClause-level CPPs: order is free (any CPP may be initial; no fixed order among CPPs). That is clause-level syntax.
SOURCED Jelinek: “the clitic pronouns … do have a fixed order: SUBJECT (i.e. NOM) must appear before OBJECT (i.e. ACC)” (p. 50); “since nominals are mere adjuncts, there is nothing to require that they have a fixed order” (p. 50). Footnote 10: she no longer calls the fixed clitic order “configurational.”
8. Derivations JB1–JB4
Schematic Warlpiri, following tree (28) literally. Labels g1/g2 and CPPi are ours. The ordering remark “the subject clitic is merged before any CPP” is ours: tree (28) itself has no VP and no subject position outside AUX. No Navajo forms. These derivations show the 1984 notation, not a verdict on A3.
JB1. Bare Verb–AUX: no CPP (Jelinek's “complete finite sentence”)
Σ Σ(√SEE) = {g1 seer (EXT), g2 seen} HYPOTHESIS (labels ours)
S1 A Merge V (+Tense/Aspect)
S2 A Merge AUX = [T S=g1 O=g2] (fixed order; clitics = arguments)
c-command inside AUX: g1 does not yet see any CPP (none present)
S3 A Merge V with AUX under S → Verb–AUX complex
Result: a complete finite sentence with no nominals (Jelinek p. 50).
Spell-out schematic: nya-nyi ka-rna-ngku ‘I see you’ (cf. (37) without CPPs)
JB2. Adargumental CPP as sister under S (tree (28), literal reading; no verdict on A3)
Continue from JB1:
S4 B Merge CPPERG = [ngajulu + ERG] as sister of Verb–AUX under S
(optional; dashed line in (28))
Linking Rule (31): ERG L-case compatible with NOM G-case of g1 (32a)
Coindex CPPERG with g1
S5 B (optional) Merge CPPABS = [nyuntu + ABS] under S
Linking: ABS compatible with ACC of g2 (32b); coindex with g2
C-command at S4/S5:
g1 sits inside AUX; CPP is sister of AUX under S
→ g1 does NOT c-command CPP on this literal reading HYPOTHESIS (no verdict on A3: pre-VPISH notation)
Spell-out (37): ngajulu-rlu ka-rna-ngku nyuntu-Ø nya-nyi
Ruleset: A builds AUX-internal order; B adjoins and orders CPPs freely.
JB3. Adsentential (secondary L-case) CPP
S4′ B Merge CPPLOC (secondary L-case) as sister under S
Linking Rule: secondary L-case is NOT compatible with any G-case
→ not coindexed with a clitic (p. 51: adsentential adjunct)
C-command: same literal reading as JB2 (no verdict on A3).
Result: a sentence adjunct, not an argument adjunct.
JB4. Baker option (ii) sketched: CPP adjoined to vP (against A3)
This is NOT Jelinek's tree. It is Baker n. 16's speculation for Navajo
IF Jelinek is right AND adjunction sites include vP.
Σ {g1 EXT, g2 INT} HYPOTHESIS
S1 A Build vP with g1 in Spec,vP (VPISH: EXT argument first)
S2 B Adjoin CPP to vP (Baker's “including … vP”)
Now g1 (Spec,vP) c-commands into its sister v′, and — depending
on adjunction geometry — may c-command the CPP
→ A3 FAILS for this placement (noun inside subject's reach)
S3 Spell-out: Navajo forms NOT invented here; Baker draws no Navajo tree.
Status: SOURCED as Baker's open option; OPEN whether Navajo uses it.
9. Datum table
Datum
Source
Effect on A3 / FB
Tree (28): CPP sister of V–AUX under S
Jelinek 1984:50 (page image)
No verdict on A3 (pre-VPISH notation; literal reading only)
“Warlpiri nominals are adjuncts to the Verb-AUX complex…”
Jelinek 1984:50
Supports PAH / FB-1 (prefixes = arguments); placement is S-level
“CPPs are sisters to the verb under S.”
Jelinek 1984:73
Same as tree (28); no verdict on A3
(73b) nominals = optional adjuncts to clitic pronouns
Jelinek 1984:73
Supports T-WEAK / FB-1; does not by itself fix adjunction height
Two adjunct types, both under S in (28); neither is an argument
Linking Rule (31), Case Compatibility (32)
Jelinek 1984:52–53
Compatible with FB linking-by-match; Warlpiri case, not Navajo φ
*(26) fixed clitic order vs free CPP order
Jelinek 1984:50
Supports A/B split; no bearing on A3 height
(35), (37) Warlpiri with/without overt CPPs
Jelinek 1984:54
Nouns optional; no Condition C data
Navajo NP hierarchy via Hale, Jeanne & Platero 1977
Jelinek 1984 ~p. 66
Not an adjunction-height claim; ignored for A3
Baker n. 16: Navajo left open; options IE-like or PP/vP adjunction
Baker later paper, PDF p. 47
Option (ii) against A3; no commitment; Baker 1996 unopened
No Navajo Condition C pair in either source
both
D6 / FB-5 unchanged by this run
10. What each source gets right, and what it misses
Source
Gets right
Misses or leaves open
Jelinek 1984
Clear tree: CPPs are S-sisters of Verb–AUX. Clitics are the arguments. Free CPP order vs fixed clitic order. Linking by case compatibility
Warlpiri, not Navajo. No Condition C. No claim about Navajo attachment height (only an NP hierarchy cite). Does not discuss whether the subject clitic c-commands a CPP. Pre-VPISH notation (no VP, no Spec,vP, no TP/CP): cannot be read as a statement about attachment relative to a VPISH subject
Baker later n. 16
Honestly leaves Navajo open. Names the live choice: IE-like agreement vs wider adjunction (PP, vP)
Does not pick. Draws no Navajo tree. Baker 1996 still needed for the polysynthesis trees Hale cites
Hale 1999 tree (9)
Pictures the adjunct view as adjunction to S, above the subject; a clear, testable picture
Resembles Jelinek's flat S only loosely (her notation has no subject position outside AUX). Treats this picture as “the” adjunct view; Baker's PP/vP option shows adjuncts need not be outside
This site (FB / A3 audit)
A3 flagged as the weakest premise; English adjunct data already against a blanket “adjunct = outside”
Still needs Baker 1996; still needs a Navajo attachment-height test (T26)
11. Open questions T26a–g
#
Question
What would decide it
T26a
Where does Baker 1996 attach dislocated NPs (Mohawk / polysynthesis trees)? Does any tree put them under the subject's reach?
Open Baker 1996, esp. the chapter on the position of NPs
T26b
Where do Willie & Jelinek 2000 attach Navajo topic/focus adjuncts? Partly answered 9:50pm: in trees (43)–(48), pp. 267–269, they are adjoined to S above IP, with the clitics inside IP/V′. Open: translating S and IP into CP/TP and checking the subject clitic against v/M/S
W&J 2000 page images around the topic/focus tree
T26c
Are Navajo postpositional objects low adjuncts (Baker option (ii): adjoin to PP), arguments of P (Hale 2003 ms.:51), or both depending on grammatical vs lexical P?
Hale 2003; Willie GPP/LPP; a Condition C or binding test from a subject prefix into a PP object
T26d
Can a Navajo subject prefix c-command out of the verb word onto a clause-level noun? (Same as T25d; Preminger-style clitic vs agreement.)
Binding / Condition C minimal pairs; Preminger 2014
T26e
Under m-command (or “c-command from the AUX node”), does Jelinek's subject clitic see the CPP after all? If yes, A3 would fail even for the literal reading of tree (28).
A stated definition; any Warlpiri Condition C literature
T26f
Does any opened Navajo source put an adjunct below Spec,vP (or below the subject prefix) while still calling it an adjunct?
Willie 1991 Ch. 3; Speas 1990 §4.4; Baker 1996
T26g
OPEN(new, 9:20pm) Restate the pronominal-argument hypothesis in a VPISH/TP/CP structure, and re-ask where the nouns attach (above Spec,vP, at vP, at PP, in a left-peripheral topic/focus position). Check what Baker 1996 and later work do (the polysynthesis-parameter structure) and whether there is a minimalist reanalysis of Jelinek. Progress and sources opened: §12
Open Baker 1996 (esp. the chapters on NP position and adjunction sites); Willie & Jelinek 2000 in full; draw each author's tree for a transitive clause with an overt subject noun and an object noun, with the subject argument in Spec,v or Spec,V
12. T26g: restating the PAH in VPISH / TP / CP structure (progress; the task stays open)
OPEN Task: restate the pronominal-argument hypothesis in a structure with a VP-internal subject and TP/CP, and ask again where the nouns attach. What follows is what was opened on 2026-10-15 (9:20pm run). It does not complete the restatement, and it is not a verdict on A3.
Source
Access
What it gives
What it does not give
Hale, Ken 2003 (ms.). “On the significance of Eloise Jelinek's Pronominal Argument Hypothesis” (published in Carnie, Harley & Willie (eds.), Formal Approaches to Function in Grammar, pp. 11–43)
Pre-publication manuscript, 72 pp., opened in full (text extract; box [local working copy, not in the repository]). Cited by manuscript page
SOURCED A Merge-based, minimalist-era restatement of the PAH for Navajo. Spine Qual > M/S > v > V, head-initial (pp. 31–32). Pronominal arguments are D, maximal projections, in thematic positions: the object of a labile verb in the specifier of V, below v; an unergative's pronoun in the specifier of v itself, “external to the projection of the lexical verb” (pp. 27–29, (22)–(23)). The subject morpheme combines with Mode as the M/S portmanteau; object raising to Spec,M is considered and rejected; word order within the verb is phonology, not head movement (pp. 11–12, 29–31). Nouns: “full DP arguments construed with the verb are adjuncts to the clause, linked to pronouns in the thematic positions defined at Merge” (p. 51); overt noun phrases are “adjoined to the full verb word (or verbal clause) in accordance with the principles developed in Willie and Jelinek (2000)” (p. 27)
No tree for overt nouns, no statement of whether the subject in Spec,v c-commands them, no Condition C. Adjunct heights are drawn only for preverbs: high, adjoined to the projection of the highest functional head (one that “properly includes the subject”), or low, adjoined to vP (pp. 16–18). “To the clause” in this spine suggests a high position, but Hale does not draw it
Willie, MaryAnn & Eloise Jelinek 2000, “Navajo as a discourse configurational language”
Opened as page images in earlier runs (pp. 264–278); trees (43)–(44) p. 267 and (47)–(48) p. 269 re-read 9:50pm (box [local working copy, not in the repository], PDF 147–148). Correction: the 9:20pm version of this row said “abstract only”, which was wrong
SOURCED Navajo trees in IP/V′/S notation: S dominates a FOCUS Adjunct (and, in (47), a TOPIC Adjunct) and IP; IP dominates 3ACC and V′; V′ dominates 3NOM and the V stem; (48) adds S′ with COMP for the two-clause case. So the overt nouns are adjoined to S, above the IP that contains the pronominal clitics. The abstract (found by search) ties the order to Rizzi's left periphery
No TP, CP or vP; the subject clitic 3NOM sits inside V′ and the object clitic 3ACC higher, in IP, so “above the subject” means above IP. This is transitional notation (audit), so it is evidence about W&J's picture, not a modern spine
Baker 1996, The Polysynthesis Parameter
Not opened. Read through (i) D. G. Frantz's review in J. Linguistics (text opened; the extract has year and page numbers stripped) and (ii) Ormston 1993, McGill MA thesis on Mohawk (opened; [local working copy, not in the repository]), which adopts Baker's structure
SOURCED(secondhand) Frantz: Baker finds the NPs “in an adjunct position, as proposed by Jelinek”; evidence includes “NPs as object do not appear to be c-commanded by subjects, for the subject-object asymmetry with regard to Condition C of the Binding Theory is not present”; the agent is the specifier of a higher null-headed VP shell (Larson-style), with UTAH. Ormston (pp. 14–15): the agent is in the specifier of a higher VP shell with a null V, an AspP between, IP/MP above, and “the full NPs … are adjoined to the clause”, following Baker's pp. 172, 204–205
Which node counts as “the clause”, and whether it is above the VP-shell subject, is not shown in either secondhand source; Frantz says Ch. 4 treats “the adjunction sites of the NPs” without saying where. Ormston's tree is a student's use of Baker, not Baker's
Baker, “Agreement, Dislocation, and Partial Configurationality”, trees (16) and (19), pp. 14–17 (Kinande)
Re-read 9:50pm (box [local working copy, not in the repository]); printed pages = PDF pages
SOURCED A TP/vP/CP restatement of the PAH idea for Kinande: subject in Spec,vP, proi with Agri+T under TP; a dislocated subject NP adjoined to TP (below Focus); a dislocated object NP adjoined to CP (pp. 14–17)
Kinande, not Navajo. It shows the “outside” attachment drawn at TP and CP in a vP/TP/CP spine, so A3-type attachment has a modern formulation for that language. For Navajo Baker gives only n. 16
Baker, “Agreement, Dislocation, and Partial Configurationality” n. 16
Opened 8:46pm (§5)
SOURCED For Navajo, dislocated NPs might adjoin “including PP and vP”
No Navajo tree
HYPOTHESISWhat this adds. (1) A minimalist-era restatement of the PAH for Navajo exists (Hale 2003), and it does have an external argument merged low (Spec,v or Spec,V) and functional heads above it, so the VPISH question can be asked of it. (2) Both it and Baker's structure (via Ormston) say overt nouns are adjuncts “to the clause”. If “clause” means the top of the spine, nouns are above the subject, which is the premise Hale 1999 needs (A3). (3) But no opened source draws that tree for Navajo nouns, none defines “clause” against the spine, Baker's own footnote for Navajo contemplates PP and vP, and the Navajo-specific source (Willie & Jelinek 2000) was seen at abstract level only. So A3 is stated by the PAH restatements as clause-level attachment, but not derived from a tree we can inspect: still no verdict. (4) A point of contact, review-level only: Baker uses the absence of a subject–object Condition C asymmetry as one diagnostic for adjunct NPs in Mohawk; Hale 1999 reports Condition C effects in Navajo and concludes the opposite. The two share the premise that adjunct nouns are outside the subject's reach (A3).
HYPOTHESISRestatement sketch for T26g (ours; to be checked against Baker 1996 and W&J 2000; 9:50pm: W&J (47) puts nouns above IP and Baker's Kinande trees put them at TP or CP, but neither is a Navajo spine). The question to ask of each author, in a spine [CP [TP/M/S [vP/vP-shell Subj-argument [VP …]]]]: is the noun (a) adjoined to CP/TP (above the subject argument), (b) adjoined to vP or VP (below TP, in reach of a Spec,v subject), (c) adjoined to PP, or (d) in a left-peripheral topic or focus specifier? Hale 1999's A3 holds for (a) and (d) and fails for (b) and (c). Nothing opened so far tells us which of these any PAH author actually chooses for Navajo.
Older verb-centred descriptions: not sourced
OPEN Grant's remark that older Navajo descriptions were verb-centred is not sourced here. What was found and opened is bibliographic only: Young & Morgan 1943, The Navaho Language (Phoenix Indian School / U.S. Indian Service); Sapir & Hoijer 1942, Navaho Texts; Hoijer 1945 (cited for Apachean tree terms) — in reference lists in ERIC ED102820 and ED101549 and in the front matter of Young & Morgan 1987. None of those works was opened, and no opened source characterises them as verb-centred. The nearest opened statements are late and are not evidence about 1940s descriptions: a publisher blurb on a book-listing page for Young 2000, The Navajo Verb (“The verb is the most important and the most complex part of Navajo grammar”), and, at search-snippet level only, McDonough & Willie: “The Navajo verb is a fully inflected form that stands as a complete proposition”. No claim is made about WWII-era sources. To source it: open Young & Morgan 1943 or Sapir & Hoijer 1942/1967 and quote the passage.
Sources: Jelinek 1984 (page images, printed pp. 50–54, 73–74; box [local working copy, not in the repository]); Baker, “Agreement, Dislocation, and Partial Configurationality” (Rutgers ms; PDF pp. 46–47, n. 16; box [local working copy, not in the repository]). Cross-links: Hale derivation (A3, T25c), Hale 1999, feature bundles, theta-args thesis, nominal adjunction, sources. Added 9:20pm: Jelinek 1984 p. 44 (INFL/VP passage); Hale 2003 ms. (full); Frantz review of Baker 1996 (J. Linguistics); Ormston 1993 (McGill MA). Not opened: Baker 1996 The Polysynthesis Parameter itself; Willie & Jelinek 2000 (abstract only); Fukui & Speas 1986, Kuroda 1988, Koopman & Sportiche 1991 (snippets only); Young & Morgan 1943; Sapir & Hoijer 1942; Hoijer 1945; Willie & Jelinek 2000 re-check for attachment height; Warlpiri Condition C literature.
Restated on the unified structure (2026-11-15). Nouns are restated as arguments merged in their theta positions (VP, Spec,vP) and, for the subject, raised to Spec,TP by the working default; a lone noun is linked by the H4″ default. Adjunction and pronominal-argument analyses below are labelled contrast; where nouns sit for Condition C is to be re-run on the unified spine. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Flat-S and other pre-VPISH notation below is labelled contrast. Derivations below in which the verb or subject does not raise are labelled contrast; the working default raises V→v→T and the subject to Spec,TP. Status on the spine: see unified structure, §2 and §4, first rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.
2026-11-11 audit pass 1. The Qual/M/S and DC/A3′/REACH rules below are historical hypotheses, not axioms or complete current vP/TP derivations. Replacing a source’s S labels alone does not establish a noun attachment site or movement path. The primary comparison starts the subject in left Spec,vP, merges T before subject raising and tests V→v→T independently. Exact LF reconstruction and the choice of interpreted copies remain OPEN; no source S-subdivision is our standard.
2026-11-11 audit pass 2. Wording that used VPISH as the name of a spine, tree, skeleton or set of slots named a structure after the hypothesis. VPISH is a hypothesis name, never a node; the structure is a vP/TP spine with the subject first-merged in left Spec,vP. Only the wording changed.
2026-11-11 audit pass 3. Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference. See the pass 3 report.
2026-11-13 audit pass 7. An earlier dated note on this page treats V→v→T and subject raising as Grant’s conjecture or a comparator to test against PF/template rivals. Since 2026-11-13 they are the working default in tensed clauses: VPISH frames where the arguments start so that syntax and semantics line up; it is not a word-order account. PF/template accounts are kept only as labelled contrast, and the open question is what else has to move. The earlier note is kept as written. See the 2026-11-13 note and the pass 7 report.
The framework audit (F1) found that this site's Condition C derivations borrowed one piece of older notation: Hale 1999's tree (9), where the nouns hang on S above a subject that is itself a daughter of S. This page replaces “S” with named positions in a modern spine (CP, the Qualifier and Mode/Subject heads of Hale 2003, vP, VP, PP, and a left-edge topic or focus position), uses only the step-by-step definition of c-command, and re-derives Hale's (10), (11a), (11b) and Speas's (4.119) at every site. Written 2026-10-16, 4:46am run (PT). Task T27i.
Ground rules (Grant's). Hale, Speas, Jelinek and Baker are each treated as possibly partly right. Silence in a source counts against nothing; only contradicting data count. Grant does not speak Navajo; every Navajo form below is copied from a page image, re-checked this run (§2). Hale 1999's trees (8), (9) are S-based, pre-VPISH notation (Grant's stance; class (b) on the audit). They are restated here and are not used as evidence about where Navajo nouns attach. Labels used: SOURCEDHYPOTHESISGRANT'S CONJECTUREOPEN.
Bottom line, in plain words
HYPOTHESISThe question that decides these sentences is not “is the noun an adjunct?” but “is the noun already in the tree when the hidden subject takes its final position?” If it is, the subject c-commands it and Condition C forbids coreference: that is what Hale reports for (10) and (11a). If the noun is attached after that point, coreference is free: that is what Speas reports for (4.119). Of the eleven site-and-subject combinations tested (§5), six predict Hale's judgment for (11a) (five of them also for (10); the sixth, a postposition site, does not apply to (10)), five predict Speas's, and none predicts both. (11b) is predicted by no site at all unless one extra assumption (A10) is added.
HYPOTHESISTwo surprises. (1) Attaching a noun to vP, the low verb phrase, does not by itself put it in the subject's reach. In Hale 2003's own structure the subject pronoun is merged in Spec,v and does not move (pp. 28–30). If that pronoun is the binder (and not the Mode/Subject head it combines with), a noun added to vP after it is outside the subject's reach, exactly like a noun on CP. It falls inside only if the subject later moves up past it (the site's V-P option). (2) Hale 2003 puts overt nouns “adjoined to the full verb word (or verbal clause)” (p. 27). Read with his Spec,v subject and step-by-step c-command, that placement predicts free coreference in his own 1999 sentences (10) and (11a). Hale 2003 says nothing about Condition C on the pages read, so this is a tension we derive, not a claim of his.
HYPOTHESISWhat survives. Assumption A3 (“an adjunct is outside the subject's reach”) holds in 5 of the 10 non-argument configurations (adjunct or left-edge) and fails in the other 5. It survives only in a restated form, A3′: a noun is outside the subject's reach if and only if it enters the derivation after the subject's last Merge. The Condition C verdict D6 stays compatible, undecided. Its content is sharper, though: Hale's (10) and (11a) no longer bear on “argument or adjunct” at all. They bear on two open facts about Navajo: how high the nouns attach, and whether the subject pronoun moves (V-H vs V-P).
GRANT'S CONJECTURESTRATA (roles first, event structure next, the noun last) gets a precise reading under step-by-step c-command. “The noun comes in later” predicts Speas's judgment if “later” means after the subject's final Merge. It predicts Hale's if the subject still moves after the noun is in (§6.4).
SPINE (Hale 2003 pp. 31–32, head-initial extended projection; CP added above, ours)
[CP C [QualP Qual [M/SP M/S [vP Subj [v′ v [VP … V ]]]]]] Qual optional
DC Derivational c-command (Abe, ex. (3), after Epstein 1999):
“α c-commands β only if β is a term of γ at the stage where α is merged with γ,
where a term of γ is γ itself or any syntactic object dominated by γ.”
Convention (ours, as on /hale-derivation/ §3): the converse also holds, so Merge(α, γ)
makes α c-command exactly the terms of γ at that stage. Move = re-Merge and creates
new c-command in the same way. Nothing else creates c-command.
ADJ Adjunction is Merge. Segments do not matter under DC: only the order of Merge does.
(Hale 2003 pp. 32–33: an adjunct is dominated by a segment of vP, “not the maximal
projection itself”. DC makes that distinction irrelevant for binding.)
COND-C An R-expression N (a name or a noun like ‘my father’) may not be coindexed with a
binder B that c-commands it.
REACH (lemma, follows from DC; HYPOTHESIS ours)
B c-commands N ⇔ some Merge or re-Merge of B takes a γ of which N is already a term.
So N is in B's reach iff N enters before B's LAST (re-)Merge, inside the γ B merges with.
A3′ (restated A3) N is outside B's reach iff N enters after B's last (re-)Merge,
or enters inside a constituent B never merges with.
Binders.SOURCED In Hale 2003 “the subject of the transitive is a pronominal argument in the specifier of v” (p. 28); arguments are “licensed in their thematic positions. Hence, there is no motivation for syntactic movement which would reposition pronominal arguments” (pp. 29–30); and, for the embedded clause of a causative, “the external argument (subject, surface causee object) … is a specifier, and not inherently a part of M/S” (n. 11, p. 68). Portmanteau formation requires “that Mode locally c-command and govern the pronoun” (n. 11, p. 68). Hale's diagrams nonetheless spell the subject inside the M/S portmanteau (tree (8), p. 11; “the subject (in M/S)”, p. 16). So which object is the binder is itself open (assumption A2). HYPOTHESIS Three binder settings follow:
Binder
Where the subject bundle g1 is at its last Merge
Whose
B1 stay (V-H)
Spec,v. One Merge only, with v′.
Hale 2003 (pp. 28–30, n. 11); the site's V-H
B2 raise (V-P)
Spec,M/S. g1 re-merges with M/S′, a γ that contains vP and anything adjoined to vP.
The site's V-P option (derivations); Baker's Kinande pro in Spec,TP (tree (16), pp. 14–15)
B3 the M/S head
The portmanteau head M/S itself, if it counts as the binder (assumption A2). M/S merges with vP.
A2 variant; gives the same answers as B2 for every site at or below vP, and cannot reach above M/S′
Rule-set tags (the site's convention, headedness, derivations): A word-internal linearization (functional heads before their complements; inside the verb word), B clause-level linearization and linking (nouns, postpositional phrases, particles), Σ role-schema step. The verb word is the changeover point. Note: the run brief for this task gave the letters the other way round (A = clause, B = prefixes). This page keeps the letters used on every other page of the site, so the tags stay comparable.
2. The data, re-checked on the page images this run SOURCED
(10) Si-tsilí [ts’ídá shi-zhé’é bi-łį́’-ígíí] yeiníłtį́. (Hale 1999:172)
1SG-YBr very 1SG-Fa 3-horse-REL 3.to.3.P.3.give/anim
‘He_i gave my younger brother the very horse of my father_j.’
Hale: Navajo “does not permit a reading of sentences like (10) in which i = j” (p. 172)
(11a) [Yiską́ągo ni-tsilí nih-aa doogáł-ígíí] y-ee shi-ł hoolne’. (Hale 1999:172)
tomorrow 2SG-YBr 1ns-to FUT.3.go-REL 3-of 1SG-with 3.P.tell
‘(S)he_i told me about your younger brother_j coming to us tomorrow.’ (i ≠ j)
(11b) [Yiską́ągo ni-tsilí nih-aa doogáł-ígíí] b-aa bi-ł hózhǫ́. (Hale 1999:173)
tomorrow 2SG-YBr 1ns-to FUT.3.go-REL 3-about 3-with A.P.good
‘(S)he_i is happy about your younger brother_j coming to us tomorrow.’ (i ≠ j)
(4.119) Kóhoot'éédą́ą́' shínaaí Naakaii Bikéyahgóó naayáhą́ą yee nihił hólne …anaa'
last.year my-brother Mexico-to went-NOM about-it us tell(OPT) would-that
‘I wish my brother would tell us about his going to Mexico last year’ (Speas 1986:319)
(4.121) ‘I wish he_i would tell us about my brother_i's going to Mexico last year.’ (p. 320)
SOURCED Hale on (11a,b): “if Navajo were nonconfigurational, the prohibited coreference reading should in fact be possible, since the factive complements would be adjuncts and therefore outside the c-command domain of the matrix subject (pro)” (p. 173). Speas on (4.119): ‘my brother’ “is located within the embedded clause” (p. 319) and “must be construed as the subject of this sentence, or at least, be construed as coreferent with the subject, so that the apparent violation of condition C which was encountered in the relative clause case is found here too” (p. 320).
Fidelity notes (this run). (i) Hale's page numbers re-checked on the page images: (10) and (11a) are on p. 172, (11b) on p. 173. (ii) Correction: the matrix verb of (4.119) is printed hólne, with no final glottal stop (Speas p. 319, checked at 300 dpi). Earlier pages (Hale derivation, Speas 1986) printed hólneʼ; they are corrected. (iii) Speas's glosses are lower-case in the original (“went-nom”, “tell(opt.)”); they are given here in Leipzig capitals. ‘last year’ is printed as two words. The first letters of the last word are blotted, so it stays “…anaa'”. (iv) Hale's gloss “A.P.good” is kept as printed and not segmented; the person Hale translates as ‘(S)he’ in (11b) is the 3 of bi-ł ‘3-with’.
3. The attachment sites, and who proposes each
Site
What it means in the spine
Where it comes from (opened)
ARG argument position
The noun is merged in the slot itself (Comp,V, or the complement of a P). Not an adjunct. Listed as the baseline
Speas 1986 (nouns in θ-positions); view (b) on Hale derivation
PP
Adjoined to the postpositional phrase whose pronoun it links to. Inherits the height of that PP (here assumed inside VP: HYPOTHESIS)
SOURCED Baker n. 16: in Navajo, dislocated NPs might adjoin “to a wider range of categories … including PP and vP” (PDF p. 47)
VP
Adjoined inside the verb phrase, before the subject is merged (= FB-3c LINK-IN)
SOURCED Baker n. 16 (above); Hale 2003's low site for preverbs, “adjunction to the verbal projection vP” (p. 17)
M/SP (the TP-like layer)
Adjoined to the Mode/Subject projection: below a subject that raises there later, or above it
SOURCED Hale 2003's high site for preverbs: adjoined to a projection of Qual or M/S, “a projection which properly includes the subject” (p. 16). Baker's Kinande dislocated subject is adjoined to TP above proi in Spec,TP (tree (16), pp. 14–15)
QualP / “the full verb word”
Adjoined to the highest head of the verb word
SOURCED Hale 2003: “the high adjunction site is the projection defined by the highest functional head present in the particular verb word” (p. 16); overt nouns “adjoined to the full verb word (or verbal clause)” (p. 27); full DPs are “adjuncts to the clause” (p. 51)
CP
Adjoined to the top of the clause
SOURCED Baker's Kinande dislocated object, adjoined to CP (tree (19), pp. 16–17)
SPEC-L left-edge topic/focus
The specifier of a topic or focus head above M/SP; either first-merged there (base) or moved there from ARG, PP or VP (moved)
Willie & Jelinek 2000 trees (43)–(48), pp. 267–269: FOCUS/TOPIC adjuncts over IP (S/IP/V′ notation, transitional; restated here as a specifier, ours). Baker's Focus head above TP (pp. 16–17)
SOURCED Baker states the Kinande limits on these sites as (21) “An NP can be adjoined to YP only if YP contains the head of the chain of the verb that theta-marks the pronoun bound by NP” and (22) “An NP adjoined to YP is licensed by predication only if NP is coindexed with the highest nominal in YP …” (p. 19, box copy re-read this run). He “leave[s] open the question of how these ideas apply to Navajo” (n. 16).
4. Derivations
Each derivation is written once as a common skeleton, R1–R8, and then the noun is entered at each site. The step at which it enters is the only thing that changes. Labels g1, g2, g3, t1, e, N, C1 are ours. Hale does not segment the verb words, so neither do we.
4.1 (10) ‘Hei gave my younger brother the very horse of my fatherj’
Σ Σ(√GIVE/anim) = {g1 giver (external), g2 recipient (the “3.to” of the gloss), g3 theme} Σ
Roles in the action (Dowty): g1 causes and acts with volition; g3 is moved/transferred.
Nouns: N1 = si-tsilí ‘my younger brother’ → links to g2.
N2 = [ts’ídá shi-zhé’é bi-łį́’-ígíí] ‘the very horse of my father’ → links to g3.
N3 = shi-zhé’é ‘my father’, inside N2, the possessor (bi- ‘3’ on ‘horse’). N3 = j.
SKELETON
R1 Merge V with g3 (Comp,V) g3 ⇄ V Σ
R2 Merge g2 into VP (position inside VP schematic) g2 → g3, V
R3 Merge v with VP v → VP, g2, g3, V A v⌢VP
R4 Merge g1 with v′ (Spec,v; VPISH first Merge) g1 → v′ and every term of v′ Σ
R5 Merge M/S with vP; Agree M/S ⟶ g1 (portmanteau; Hale 2003 n. 11) M/S → vP A M/S⌢vP
R6 Move (B2 only) g1 re-merges with M/S′ (Spec,M/S) g1 → every term of M/S′
R7 Merge Qual (if present) with M/SP A Qual⌢M/SP
R8 Merge C (null here) with the clause B
SO Spell-out: the word yeiníłtį́ (Hale p. 172), linearized inside the word by A;
the nouns precede the word, linearized by B. Link: N1 ↔ g2, N2 ↔ g3 by φ-match. B
ENTRY OF N2 (with N3 inside it), site by site
ARG N2 replaces g3 at R1. At R4, N3 is a term of v′ ⇒ g1 → N3. i ≠ j ✓ Hale
VP N2 adjoins to VP after R2. At R4, N3 is a term of v′ ⇒ g1 → N3. i ≠ j ✓
PP not applicable: N2 is the theme, not the object of a postposition. n/a
vP, B1 N2 adjoins to vP after R4. g1's only Merge was R4; N3 not yet a term ⇒ g1 ↛ N3. free ✗
vP, B2 N2 adjoins to vP after R4. At R6, g1 re-merges with M/S′ ⊇ vP ⊇ N3 ⇒ g1 → N3. i ≠ j ✓
M/SP-below, B2 N2 adjoins to the M/S projection after R5, before R6. At R6, g1 → N3. i ≠ j ✓
M/SP-above N2 adjoins after R6 (B2), or anywhere above vP (B1). g1 ↛ N3. free ✗
QualP N2 adjoins after R7 (“full verb word”, Hale 2003 p. 27). g1 ↛ N3. free ✗
CP N2 adjoins after R8. g1 ↛ N3. free ✗
SPEC-L base N2 first-merged in Spec,Top/Foc above M/SP. g1 ↛ N3. free ✗
SPEC-L moved N2 merged at ARG or VP, then moved up. The lower copy is c-commanded (R4).
If Condition C reads the lower copy (A7): i ≠ j ✓. If not: free ✗.
4.2 (11a) ‘(S)hei told me about your younger brotherj coming to us tomorrow’
Σ Matrix Σ(tell) = {t1 teller (external), t2 addressee (shi-ł ‘1SG-with’), p1 (the 3 of y-ee ‘3-of’)} Σ
C1 = [yiską́ągo ni-tsilí nih-aa doogáł-ígíí]; N = ni-tsilí ‘your younger brother’ = j,
the goer of doogáł (FUT.3.go), inside C1 wherever C1 itself is attached.
INNER Build C1: m1 (goer) in the inner verb word doogáł; N links to m1; nih-aa ‘1ns-to’ inside;
-ígíí (REL) closes C1 (Hale & Platero 1974 treat it as a determiner). A inside the word, B for the clause
MATRIX R1′ Merge P y-ee with p1 (PP1 = ‘about it’) R2′ Merge PP1 and PP2 (shi-ł) into VP
R3′ Merge v R4′ Merge t1 in Spec,v R5′ Merge M/S; Agree R6′ (B2) t1 re-merges in Spec,M/S
R7′ Qual (none here) R8′ C. Spell-out hoolne’ A; C1, y-ee, shi-ł before it B.
ENTRY OF C1 (N inside), site by site
ARG C1 is the complement of y-ee (at R1′). R4′: t1 → N. i ≠ j ✓ Hale
PP C1 adjoins to PP1 (Baker n. 16), PP1 inside VP. R4′: t1 → N. i ≠ j ✓
VP C1 adjoins to VP before R4′. R4′: t1 → N. i ≠ j ✓
vP, B1 C1 adjoins to vP after R4′. t1 ↛ N. free ✗
vP, B2 as above, then R6′: t1 re-merges over it. t1 → N. i ≠ j ✓
M/SP-below, B2 before R6′. t1 → N. i ≠ j ✓
M/SP-above, QualP, CP, SPEC-L base t1 ↛ N. free ✗
SPEC-L moved from ARG/PP/VP; i ≠ j only if Condition C reads the lower copy (A7).
HYPOTHESIS Hale's remark that the factive clauses “would be adjuncts and therefore outside the c-command domain” (p. 173) is true in 5 of the 10 non-argument rows of §5 (vP with B1, M/SP-above, QualP, CP, SPEC-L base) and false in the other 5 (PP, VP, vP with B2, M/SP-below, SPEC-L moved).
4.3 (11b), the earlier D6, re-run: the binder is inside a postposition
Σ Matrix Σ(√GOOD, stative): z (what is good: C1, via b-aa ‘3-about’) and e (the experiencer,
the 3 of bi-ł ‘3-with’). e is the person Hale translates ‘(S)he_i’. Σ
E1 Merge P b-aa with p (PP1, ‘about it’) E2 Merge P bi-ł with e (PP2)
DC: e's only Merge is with P bi-ł ⇒ e c-commands exactly {bi-ł}.
E3 Merge PP1, PP2 into VP; v; M/S (A.P.good); nothing raises (the site's psych-verb analysis,
after Willie 1991:47–49, 105–106; page)
RESULT, every site (ARG, PP, VP, vP, M/SP, QualP, CP, SPEC-L): e ↛ N ⇒ free ✗ Hale
WITH A10, restated for a vP/TP spine: “PP2 binds as if it were e”, and PP2 is merged above PP1.
Then the binder's last Merge is E3 (PP2 with its sister inside VP):
ARG, PP1-adjunction, VP-adjunction below PP2: e → N ⇒ i ≠ j ✓
vP, M/SP, QualP, CP, SPEC-L base: e ↛ N ⇒ free ✗
(No B2 version: on the site's analysis the experiencer never moves to Spec,M/S.)
HYPOTHESISD6 result. The re-run confirms and sharpens the 1:07pm finding. With only Merge-order c-command, a pronoun inside a postposition c-commands nothing outside it, so (11b) cannot be predicted at any attachment site without A10. With A10 it patterns like (11a) with a binder that never moves (B1). (11b) is therefore not a second, independent test of where nouns attach.
4.4 (4.119) ‘I wish my brother would tell us about his going to Mexico last year’
Σ Matrix Σ(tell, optative) = {t1 teller (external), t2 addressee (nihił ‘us’), p1 (the 3 of yee ‘about-it’)} Σ
C1 = [kóhoot'éédą́ą́' shínaaí Naakaii Bikéyahgóó naayáhą́ą]; N = shínaaí ‘my brother’,
inside C1 (‘last year’ modifies the trip, Speas p. 319). Above the ‘tell’ clause: …anaa' ‘would-that’.
SKELETON as 4.2 (R1′–R8′); spell-out hólne A; C1, yee, nihił before it B.
ENTRY OF C1 same c-command results as 4.2, site by site.
REPORTED i = j is the reading used (4.121).
So the sites that PREDICT Speas's judgment are those where t1 ↛ N:
vP with B1; M/SP-above; QualP; CP; SPEC-L base. (Also any site in the wish layer above
the ‘tell’ clause: schematic, OPEN; nothing opened puts C1 there.)
Speas's own analysis: N IS c-commanded (an “apparent violation of condition C … found here too”,
p. 320); the reading is let through by across-the-board coindexing (§4.7.6.3). That is an
override at the ARG site, not a site prediction.
5. Results table HYPOTHESIS
✓ = the derivation predicts the judgment the source reports; ✗ = it predicts the opposite. Reported: (10), (11a), (11b) i ≠ j (Hale); (4.119) i = j (Speas).
#
Site
Binder
Is the noun in the subject's reach?
(10)
(11a)
(11b)
(4.119)
1
ARG (not an adjunct)
B1, B2
yes
✓
✓
✗ (✓ with A10)
✗
2
PP (PP inside VP)
B1, B2
yes
n/a
✓
✗ (✓ with A10)
✗
3
VP (LINK-IN)
B1, B2
yes
✓
✓
✗ (✓ with A10)
✗
4
vP
B1 (Hale 2003)
no
✗
✗
✗
✓
5
vP
B2 (V-P), or B3
yes, after raising
✓
✓
✗
✗
6
M/SP, below the raised subject
B2
yes
✓
✓
✗
✗
7
M/SP (TP), above the subject's last position
B1, B2, B3
no
✗
✗
✗
✓
8
QualP, “the full verb word” (Hale 2003 pp. 16, 27)
all
no
✗
✗
✗
✓
9
CP (Baker (19))
all
no
✗
✗
✗
✓
10
SPEC-L, first-merged there
all
no
✗
✗
✗
✓
11
SPEC-L, moved from 1–3, Condition C at the lower copy (A7)
all
yes (lower copy)
✓
✓
✗ (✓ with A10)
✗
Tally. Rows predicting Hale's (10) and (11a): 1, 3, 5, 6, 11 (and row 2 for (11a); (10) has no PP site). Rows predicting Speas's (4.119): 4, 7, 8, 9, 10. No row predicts both. No row predicts (11b) without A10. The dividing line is REACH (§1): whether the noun is in the tree before the subject's last Merge. It is not the line between argument and adjunct, which falls between row 1 and rows 2–11.
HYPOTHESISWays to get both judgments, none tested. (a) The site differs by sentence: (11a)'s clause sits in rows 2/3/5/6, (4.119)'s in rows 4/7–10. Nothing opened says what would make the clause attach higher in (4.119). The candidates are the differences already listed (Hale derivation §6): the optative wish layer, past versus future time, a first- versus second-person possessor, the different nominalizers, or speaker variation. (b) The binder differs: the subject raises (B2) in (11a) and not in (4.119). (c) One judgment is an override: Speas's ATB coindexing for (4.119), or Hale's parallel parse (13), which needs a transitive inner verb and so does not reach (4.119) (A6). (d) The judgments are not stable (A5).
6. What survives
6.1 A3 (“an adjunct is outside the subject's reach”)
HYPOTHESISSurvives in 5 of the 10 non-argument configurations (rows 4, 7, 8, 9, 10) and fails in 5 (rows 2, 3, 5, 6, 11). As a claim about adjuncts it is false. As a claim about timing it is true by definition: A3′ (§1) is just REACH. Hale's p. 173 inference therefore needs one more premise: that a nonconfigurational analysis puts the nouns after the subject's last Merge. For Hale 2003 (row 8) it does, whichever binder is chosen. For Baker's Kinande objects (row 9) it does. For Baker's Navajo speculation, PP or vP (n. 16), it depends on the binder (rows 2, 4, 5).
6.2 D6, the Condition C verdict
HYPOTHESISStays COMPATIBLE, UNDECIDED. But what Hale's (10) and (11a) show has narrowed. Under derivational c-command they are compatible with nouns being linked adjuncts (rows 2, 3, 5, 6), so they say nothing against “nouns are not the verb's arguments”. What they constrain is (i) attachment height and (ii) whether the subject pronoun moves. Speas's (4.119) constrains the same two things in the opposite direction. So D6 now reduces to two open questions about Navajo (T28a, T28b), not to the argument/adjunct question.
6.3 The feature-bundle hypothesis (FB)
HYPOTHESIS FB-1 to FB-4 are untouched. FB-5's exemption (“a linked noun is outside the bundle's reach”) is true only where REACH fails (rows 4, 7–10). FB-3c LINK-IN (row 3) and a new option, LINK-vP + V-P (row 5: link the noun at vP, then let the subject bundle raise), both give Hale's judgments while keeping nouns as linked adjuncts. Row 5 is new this run. It differs from LINK-IN in one checkable way: in row 5 the noun is not in the reach of a subject that stays low. So FB plus V-H predicts Speas, FB plus V-P predicts Hale. This counts against nothing in FB. Only data contradict, and none here contradicts FB in every row.
6.4 STRATA GRANT'S CONJECTURE
HYPOTHESIS Under derivational c-command, when a noun enters is where it is, for binding purposes. Layer 1 (bundles) includes R4, the subject bundle's first Merge. If Layer 3 (the noun) comes after everything the subject does, STRATA patterns with rows 4/7–10 and predicts Speas's (4.119). If V-P raising of the bundle (R6) can follow Layer 3, STRATA patterns with row 5 and predicts Hale's (10), (11a). The slot-filling version (S-i on Hale derivation §7) is row 1. So the old sub-versions S-i, S-ii and S-iii are now rows 1, 5 and 4/7–10. The open question for STRATA is a single ordering: does Layer 3 precede or follow the bundle's last Merge? Layer 2 (event-structure roles, Rappaport Hovav style) adds no Merge of a binder in these derivations, so it does not change any row. It is compatible with all of them.
7. What each source gets right, and what it misses (on this test only)
Source
Gets right
Misses or leaves open
Hale 1999
The inference from (10), (11a) to “the nouns are in the subject's reach” is valid under REACH. (10) is the cleanest case: no postposition, no nominalized clause
Equates “adjunct” with “outside the reach” (p. 173). That holds in only half the adjunct sites. Trees (8), (9) are S-based notation, so they cannot show which modern site is meant. (11b) needs A10
Hale 2003
A Merge-based spine with a low subject and explicit high/low adjunct sites (pp. 16–18). The segment point (pp. 32–33) is exactly the right detail for this test
Places nouns at the full verb word (p. 27; row 8), which predicts free coreference in (10), (11a). The paper does not discuss Condition C, so this is our derived tension, not a refutation of it
Speas 1986
Reports a same-shape sentence with the opposite judgment, and says plainly that it is an apparent Condition C violation (p. 320), so it is not hidden by the analysis
Her device (ATB coindexing) is an override at row 1, not a site prediction. The source of the judgment is not stated on pp. 319–320
Baker (Kinande ms.)
Draws the outside sites in a modern spine (TP, CP; trees (16), (19)) and states licensing conditions (21), (22) (p. 19). Leaves Navajo open and names PP and vP (n. 16)
Kinande, not Navajo. HYPOTHESIS If (22) applied unchanged to Navajo, an object-linked noun such as N2 in (10) could not adjoin below the subject (the highest nominal in vP or M/SP is the subject), so it would have to go to CP (row 9). That predicts the wrong result for (10). Baker does not claim (22) for Navajo, so this counts against nothing
Willie & Jelinek 2000
Navajo-specific topic/focus positions over the clause (pp. 267–269)
S/IP notation; whether nouns are first-merged there or moved (rows 10 vs 11) is not shown on the pages read
Jelinek 1984
Not used: a flat-S, pre-VPISH Warlpiri tree (Grant's stance; page). Its restatement is T26g.
8. Gaps and next steps (T28a–f) OPEN
#
Task
Needs
T28a
Does the Navajo subject bundle move (B2, V-P) or stay (B1, V-H)? Hale 2003 says no movement of pronominal arguments (pp. 29–30) but leaves raising open for intransitive labile verbs (n. 11). Look for an opened diagnostic: scope of the subject relative to vP-level preverbs (Hale 2003 pp. 16–18)
Hale 2003 (opened); Speas 1990 Ch. 4 (not opened)
T28b
Which site is each noun at, sentence by sentence? Is anything in (4.119) (optative wish layer, past time, 1SG possessor, NOM vs REL) associated with higher attachment in any source?
Minimal pairs (T25a); Speas 1990 Ch. 4; Platero
T28c
Rows 10 vs 11: are left-edge nouns in Navajo first-merged or moved, and does Condition C read the lower copy (A7)?
W&J 2000 pp. 270–278 (re-read); Navajo reconstruction data
T28d
Where do postpositional phrases like y-ee ‘about it’ sit: inside VP or attached high? Row 2 depends on it
Hale 2003 p. 51 (bare P-objects in base position); Willie 1991 Ch. 3
T28e
A10 restated: can the pronoun of a postposition bind out of its PP (Willie's “Dative Subject”, pp. 47–48)? A Navajo case with an experiencer P and an R-expression in a sister clause, other than (11b)
Published Navajo data
T28f
Baker (21)–(22) for Navajo: if predication licensing holds, object-linked nouns go to CP. Check this against Navajo order data (W&J) before taking it further
Baker 1996 (not opened); W&J 2000
Sources, all opened in earlier runs; nothing new was opened. Re-checked this run: Hale 1999 pp. 171–173 (page images, box [local working copy, not in the repository]); Speas 1986 pp. 319–320 (page images at 110 and 300 dpi from [local working copy, not in the repository], PDF pp. 319–320 = printed pp. 319–320); Hale 2003 ms. pp. 4, 10–11, 16–17, 27–33, 51, 67–68 (text extract, [local working copy, not in the repository]); Baker, “Agreement, Dislocation, and Partial Configurationality” pp. 13–17, 19 and n. 16 (box [local working copy, not in the repository]). Used as recorded on earlier pages, not re-opened: Abe, “Late Merge and Phase Theory”, ex. (3) (quoted on Hale derivation §3); Willie & Jelinek 2000 pp. 267–269 (audit); Willie 1991 pp. 47–49, 105–106 (psych verbs). Cross-links: framework audit T27i, Hale derivation, Hale 1999, T26g, feature bundles, thesis D6, open questions, sources.
Rule sets A and B as spell-out statements: one ordering rule, two cycles? Hale 2003's hierarchy against an antisymmetric (Kayne-style) alternative (T27j)
2026-11-11 audit pass 2. Apostrophe fixes only. Navajo glottal stops typed as ' or ’ (for example in prefix names like ʼa- and ʼeʼe-, or words like Awééʼ) now print as ʼ (U+02BC), and a build step no longer turns the closing quote of an English translation inside a Navajo row into a glottal stop (‘…the girls.’, not ‘…the girls.ʼ). No form, gloss or translation wording changed.
2026-11-11 audit pass 3. Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference. See the pass 3 report.
2026-11-11 audit pass 4. Interlinear rows whose form and gloss do not line up one-to-one were re-checked against the source page. Each is marked † with a note under its table: SOURCED where the source prints it that way, OPEN where the source cannot confirm it. See the pass 4 report.
2026-11-13 audit pass 7. An earlier dated note on this page treats V→v→T and subject raising as Grant’s conjecture or a comparator to test against PF/template rivals. Since 2026-11-13 they are the working default in tensed clauses: VPISH frames where the arguments start so that syntax and semantics line up; it is not a word-order account. PF/template accounts are kept only as labelled contrast, and the open question is what else has to move. The earlier note is kept as written. See the 2026-11-13 note and the pass 7 report.
Restated on the unified structure (2026-11-15). Prefix order is restated as Grant’s hypothesis that headedness is fixed locally by Merge site, so the sublexical domain is linearized opposite to the phrasal domain and the prefixes are successive word-internal Merge and Agree sites (GRANT’S HYPOTHESIS, tested on the sourced template in §3.2; fit/misfit OPEN). A possible sublexical V-to-T is a separate, untested GRANT’S HYPOTHESIS (§3.3). Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). Derivations below in which the verb or subject does not raise are labelled contrast; the working default raises V→v→T and the subject to Spec,TP. Status on the spine: see unified structure, §3. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.
2026-11-11 audit pass 1. A/B, SA/SB, PL, CYC v1/v2, AS-1 and their K consequences are proposed linearization models, not axioms or independent demonstrations of headhood. Matching a string does not establish a unique boundary or LF. The Qual/M/S hierarchy is a brief source-analysis contrast; our primary comparator is V→v→T with a left Spec,vP subject raising separately. Every N1–N8 semantic/scope derivation and the rule family’s interaction with that movement remain OPEN for pass 2. CYC K3 and AS-1 K3 are distinct scoped consequences, not interchangeable rule IDs.
The framework audit (F3) found that the site's rule sets A and B were written as a head-direction parameter (“a head comes before its complement inside the word, after it outside”), which is X-bar-era notation. In Hale 2003 the hierarchy comes from Merge and the order comes from spell-out. This page restates A and B as explicit spell-out statements over Hale's hierarchy (Qualifier > Mode/Subject > v > V, with the subject in Spec,v and CP above the word), derives schematic forms and every Navajo verb form this needs from opened sources step by step, and then tests the antisymmetric alternative (Kayne 1994: one universal specifier–head–complement order, with head-final orders made by leftward movement). Kayne 1994 and Goad & Travis 2021 were both opened for this run. Written 2026-10-16, 12:46pm run (PT).
Ground rules (Grant's). Every analysis cited (Hale, Rice, Travis, Goad & Travis, Kayne, Speas) is treated as possibly partly right. Silence counts against nothing; only contradicting data count. Grant does not speak Navajo; every Navajo form below is copied from an opened source with its example number, and glosses are given in Leipzig capitals (the source's own gloss line is noted where it differs). No Navajo form, judgment or scope reading is invented: where a test needs one, it is listed as OPEN. Pre-VPISH trees are not used. Labels: SOURCEDHYPOTHESISGRANT'S CONJECTUREOPEN.
Rule-set letters. The site convention is kept: A = word-internal (the prefixal, “left-headed” verb word), B = clause-level (nouns, postpositional phrases, clause particles), Σ = role-schema step. The routine brief behind these runs once gave the letters the other way round (A = clause, B = prefixes); as on Condition C by site, this page keeps the site's letters. Notation. The Syntax · Semantics · Interface bot's notation conventions have not arrived (its BRIEF.md, copied 2026-10-15 9:55pm, has no notation section), so the site's provisional standard Minimalist notation stays: copies in angle brackets, Spec,X for specifiers, ⌢ for “immediately or eventually precedes”.
Bottom line, in plain words
HYPOTHESISA and B can be stated as one ordering rule applied twice. The rule is: inside a unit, everything that is not a head (pronoun prefixes, preverbs, nouns, whole phrases) comes first, ordered from highest to lowest in the tree; then the heads, also from highest to lowest. Apply it once to the verb word. Then apply it again to the clause, where the finished word counts as a single non-head. That reproduces every sourced order tested here (five Hale 2003 verb words, two clause types from the Navajo Language Academy pages, one Willie & Jelinek sentence) with no movement at all. The apparent switch from “heads first” inside the word (the conjunct prefixes Qualifier–Mode/Subject–classifier–stem) to “heads last” in the clause (ya’, -go, -ę́ę at the right edge) is not a second rule book. It comes from the word being finished, and then treated as one block, before the clause is ordered. This rule is called CYC below (§3). It also derives, rather than stipulates, Hale's split between lexical heads (V, P, N take their argument on the left) and functional heads (take their complement on the right).
GRANT'S CONJECTURE → HYPOTHESISGrant's “the word is the changeover point” survives, in a sharper form: the word is a spell-out cycle, the point at which heads stop being ordered as heads and the whole word starts being ordered as one non-head. What it costs: the cycle boundary itself is stipulated (Hale says openly that “when to stop building onto the skeleton” is unsolved, 2003 n. 12). It has two known exceptions, both flagged by Hale himself: oblique objects that jump to the far left (2003 n. 13), and “deictic subject” prefixes, which Hale treats as clause-level nouns spelled inside the word (2001:691). It also makes one sharp prediction that is still open: the question particle da’, and doo of doo … da, must be phrases (specifiers), not heads.
HYPOTHESISThe antisymmetric alternative (Kayne 1994, opened this run) gets the order of the four conjunct heads for free, with no movement, because Kayne's single order (head before complement) is exactly Hale's “head-initial” word-internal order. It also gets postpositions and clause-final particles by moving complements leftward (Kayne pp. 47–53), and Kayne even cites Hale on Navajo postposition agreement as support (p. 49). But it pays elsewhere. Every pronoun prefix and every preverb has to move above the highest conjunct head. For the object, that is exactly the raising Hale considered and rejected (2003:11–12). Hale's three trees need one to four extra silent heads each, because Kayne allows only one specifier or adjunct per phrase (pp. 21–22, 27). And a low preverb like yisdá in Hale's (13), which on Hale's reading does not scope over the subject (p. 19), has to move over the subject and then be interpreted back in its lower position.
HYPOTHESISDoes antisymmetry get the word as changeover for free? No. In Kayne's theory the same linear-order axiom applies below the word as above it (§4.5). The heads that pull their complement to the left (V, P, N, clause-final particles) include V, which is inside the word, so the boundary between “pulls” and “doesn't pull” does not follow the word. In that version the word is only a unit of pronunciation. The antisymmetric variant of Travis and Goad & Travis (who state that they hold Kayne's view, 2021 p. 6) does treat the word as special. It gets that only by adding a rule that works only inside words, Iterative Root Prefixation, which is rule set A in a new form. So the word is a real boundary in every option that fits the template, but in none of them does it come for free.
OPENWhat decides between them: which prefix is “higher” inside the word. Antisymmetry says the leftmost prefixes are highest, so the object prefix sits above the subject. Hale and CYC say the object is below the subject. Rice's generalization that the subject has scope over the object (known here only secondhand) fits Hale and CYC; antisymmetry has to treat it as reconstruction. Separately, Hale's hierarchy and Goad & Travis's put the Qualifier and Mode/Subject in opposite orders, yet both derive the same string. Scope judgments on Qualifier meanings would decide. Neither kind of judgment is in any opened source.
HIERARCHY (Merge; SOURCED Hale 2003 ms. pp. 10–11, 16–17, 26–33, 37–41; C above the word: ours, HYPOTHESIS)
[CP C [QualP (Adj)* Qual [M/SP M/S [vP (Adj) Subj [v′ v [VP (Spec) [V′ Obj V ]]]]]]]
– “Qual selects M/S, and M/S selects the verb (taken to be v+V here). These elements are therefore
in a head-complement structural relation which, in Navajo at least, is head-initial” (p. 32).
– “the structure projected by the lexical category V is right-headed, while the structures projected
by the higher nuclear elements (v, M, etc) is left-headed” (n. 4, p. 63).
– transitive subject = pronoun in Spec,v (p. 28); unaccusative argument = Spec,V ((22), p. 28);
unergative argument = Spec,v ((23), p. 28); object = complement of V ((8), p. 11).
– adjuncts (preverbs, It/Rev, Pl) adjoin to the projection of the highest nuclear head (pp. 16–17, 39)
or, “sometimes at least”, to vP (p. 17; (14) p. 19; (21) p. 26).
– “There is no head-movement in the traditional sense” (p. 30); arguments “licensed in their
thematic positions” (pp. 29–30).
VPISH Subj is merged in Spec,v (inside the verb phrase). Its φ is spelled out on M/S (the portmanteau),
which needs M/S to “locally c-command and govern the pronoun” (n. 11, pp. 67–68).
V-H (stays, Hale's default) and V-P (raises; the site's option) as on /derivations/.
PLANES (Hale p. 33: “purely expository”; Bobaljik, reported in n. 13, suggests they may be real)
NCL (nuclear) = the heads Qual, M/S, v, V
NNCL (nonnuclear) = pronoun arguments (Obj, oblique Obj) and adjuncts (Pv, It/Rev, Pl)
W (word domain; HYPOTHESIS) the domain spelled out when the highest nuclear head (Qual, else M/S)
has been merged: that head's maximal projection with the NNCL material inside it.
Hale's term is “verb word”, with the caveat that “verbal phonological phrase” may be better
(n. 6, p. 65); where the word stops is open (n. 12, pp. 68–69).
SOURCEDWhy the word boundary is not fixed in Hale. “The main problem would be to determine when to stop building onto the skeleton, in order to permit some preverbs, for example, to appear as separate words”. His example is the preverb ka ‘sick’, which “can be separated from the rest of the verb by the polar interrogative enclitic” (Hale 2003 n. 12, pp. 68–69, citing Paul Platero p.c.). The two Navajo forms in that note are printed with a broken font on the page image (p. 69), so they are not reproduced here.
Update 2026-10-20, 12:46pm. The two forms can now be read with a key from the manuscript itself (the prose glyph “¡” = á, since prose yisd¡ on p. 15 is table (10)'s legible yisdá; checked against the NLA's naniná and -ísh): ka naash[á] ‘I am sick’, ka-[á]sh nanin[á] ‘are you sick?’ (restored characters bracketed). See where the yes/no enclitic -ísh can go §2. On that page -ísh becomes a test for where Hale's skeleton stopped (EN-5): in this form, ka is outside it.
2. Rule sets A and B as spell-out statements
The old statement (refined conjecture (i′), headedness §4): LIN(H, C) = H⌢C iff H is functional and inside the word; C⌢H iff H is lexical (V, P, N) or functional outside the word. The audit's objection: that is a head-direction parameter with a domain index, not a statement about how a Merge-built structure is pronounced.
2.1 Rule set A: the verb word (Hale's spell-out, restated)
SA1 SKELETON The stem V brings a bipartite skeleton CVC+CVC whose right half the stem fills
(Hale p. 31, (24)).
SA2 NUCLEAR The left half (the “receptor”) is filled first by the nuclear heads, bottom-up,
FILL “successive cyclically”: v → coda; M/S → nucleus (neutral vowel / aspect) and onset
(pp. 33–35; (26)–(27); p. 39 for (31)).
SA3 NONNUCLEAR If the receptor is still unfilled, nonnuclear material fills it, “working from right
FILL to left, or upward in the structure” (p. 34; (28) p. 36).
SA4 PREFIX Left-over material, nuclear first, then nonnuclear, is “prefixed to the now replete
skeleton, proceeding from right to left, until no material is left” (p. 34; pp. 39–40).
SA5 WORD-UP The result is mapped to one prosodic word; Hale stays “essentially silent” on how
(pp. 37, 40; n. 12).
SOURCED SA1–SA5 are Hale's procedure in his words. HYPOTHESIS The restatement as five numbered steps, and the following lemma, are ours.
PL (planar linearization; HYPOTHESIS, follows from SA1–SA4 for every form in §4)
For x, y in W:
(a) if x ∈ NNCL and y ∈ NCL, then x ⌢ y;
(b) if x, y are in the same plane and x asymmetrically c-commands y, then x ⌢ y.
Why PL follows. SA2 fills the receptor bottom-up from the right edge (v as coda, then M/S to its left), so the higher nuclear head ends up further left. SA4 prefixes leftovers upward at the left edge, so again higher means further left. Because SA4 does nuclear before nonnuclear, every nonnuclear element ends up left of every nuclear one. SA3 is the only step that puts nonnuclear material inside the receptor, and it does so only when the nuclear heads are exhausted, so it never puts an NNCL element to the right of an NCL one. Example: y fills the onset in (28), still left of ł.
2.2 Rule set B: the clause
SB1 DOMAIN Everything outside W: nouns, postpositional phrases (PPs), clause particles, enclitics.
SB2 NONHEADS Nouns and PPs precede the verb word; the verb word precedes the clause-final particles
FIRST ya’ (tag), -go (subordinator), -ę́ę (past relative) (NLA pages; headedness §2.1).
SB3 HEIGHT Among the nouns, the higher one comes first: topic noun, then the other noun, then the
verb (W&J 2000:273, (61a)), if the topic is the higher adjunct (site assumption).
SB4 LEXICAL V, P and N take their pronoun argument on the left: the object prefix before the stem,
nee ‘2-with’, bee ‘3-with’, shi-zhé’é ‘1SG-father’ (Hale p. 32: “an arrangement which
extends to postpositional and nominal (possessive) projections as well”).
SB5 ENCLITICS -ísh / -sh attach after a phrase (first phrase: neutral question; later phrase: focus;
NLA yes/no (4), (8), (10)–(11)). Placement is phonological (second position or focus),
not ordered by SB2–SB3.
Update 2026-10-20, 12:46pm (T41f). SB5 is sharpened as EN-0 to EN-6 on where the yes/no enclitic -ísh can go. The host is a spelled-out unit (the verb word W, a postposition word, a noun, a free preverb, or a phrase), and never a piece of W (EN-1′, a prediction). The neutral position is the first constituent of C's complement, which also derives *Daʼísh (EN-2). One correction: placement is not independent of SB2. Because the head comes last, a phrase's right edge is its last word's right edge, and that gives NLA (10a)/(10b) one string with two readings (EN-4). A second point concerns the 2026-10-17 note in §3 (D1 = the MVS): NLA (6) Neésh hólǫ́? puts the enclitic between the PP and the verb word, inside the MVS. So the enclitic's seal is W (B2), not the MVS (B3) (OPEN which reading of D1 is right; T43e).
HYPOTHESISObservation. SB2 + SB3 are PL with “NNCL” read as “non-head” and “NCL” read as “head”. SB4 is PL(a) applied to V, P and N. So the clause rules and the word rules are the same statement over different units. That is the step to CYC.
3. One rule, two cycles (CYC)
Update 2026-10-17 (clause spine §7).HYPOTHESIS Where D1 stops: Willie & Jelinek 2000:270 call the verb plus any preceding postpositional phrases a “noninterruptible constituent, the Maximal Verb Complex … the domain of the voice alternation”, and their (50b) dibé łį́į́ʼ bitah yíghááh puts inverse bi- on a separate postposition word. So D1 is taken to be the MVS (QualP with its VP-internal PPs as already-spelled non-heads), not the prosodic word. SB1 below (“everything outside W: nouns, postpositional phrases …”) should accordingly be read with grammatical PPs as D1 non-heads, not D2. The clause level D2 gains silent Foc/Top heads whose specifiers hold the preverbal and outer nouns. CYC itself is unchanged.
Update 2026-10-21, 4:46am (T43e).HYPOTHESIS The note above is revised in one respect. The PPs do belong with the verb, and CYC still orders them before W inside the MVS. But the MVS is not a sealed unit: the yes/no ending can stand between a postposition word and the verb (NLA (6) Neésh hólǫ́?; … dził bąąhgóósh chʼétʼaʼ?), and Willie (pp. 24, 64, 224) and Hale 1997 (PDF p. 31: “a phonological matter, not a syntactic one”) treat attached vs separate postpositions as pronunciation over one syntax. So the sealed spell-out units are W and each PP, and D1 in the sense of “sealed word-level unit” is W (rule PC-1 on attached or separate postpositions §4). Writing a PP onto W is a post-seal step (PC-3), and only the case-realizing causative P goes inside W (PC-4).
Update 2026-10-22, 4:46am (T49).HYPOTHESIS With D1 = W (PC-1), the PP and W are both sealed units ordered in D2, where W is a non-head (K3). A grammatical PP is V's complement, so it and V are sisters, and C2 says nothing about their order. The derivations' “C1: PP ⌢ W” treated W as a head in the very domain where K3 treats it as a non-head. The fix is C1′: in D2 a sealed unit precedes the unit that spells out the head it is an argument of (SB4/K2 restated for sealed units); otherwise C2 orders sealed units. The rule-set summary and the other disagreements found are on the rule-book check (RB-1 to RB-5).
Update 2026-10-22, 12:46pm (T49h): CYC v2.HYPOTHESIS The rule text is now rewritten to match the notes above: D1 = W (sealed); each PP, possessed N and relative is its own sub-domain; D2 = the clause; and C1′ is part of the rule, widened from “argument of” to “inside a projection of” so that it also orders an adjoined beneficiary PP (Willie (85)/(86)). The MVS is no longer a domain: it is the set of PP sites inside QualP. Every derivation on this page (N1–N8) gives the same string under v2. The re-run and the reasons are on postpositions by verb §6–7. The v1 text is kept below v2 for the record.
CYC v2 (HYPOTHESIS, ours; 2026-10-22, 12:46pm)
For each spell-out domain D, let H(D) = the heads of D's spine and N(D) = everything else in D.
Any unit sealed in an earlier domain counts as ONE element of N(D).
C1 every x ∈ N(D) precedes every y ∈ H(D);
C1′ for sealed units x, y ∈ N(D): if x is first-merged inside a projection of a head that y
spells out (as its complement, specifier or adjunct), then x ⌢ y;
C2 otherwise, within N(D): if x asymmetrically c-commands y, then x ⌢ y;
C3 within H(D): if x asymmetrically c-commands y, then x ⌢ y.
Domains:
D1 = W: Qual, M/S, v, V with the pronoun slots and preverbs merged among them; sealed.
sub = each PP (P head), each possessed N (N head), each relative (-ę́ę head); sealed before D2.
D2 = the clause: C-type particles and silent Foc, Top, C as heads; W, PPs, nouns, sealed
clauses as non-heads.
The MVS is not a domain: it is the set of PP sites inside QualP, ordered in D2 by C1′.
K1–K4 as below. K5 a PP sited inside QualP precedes W (C1′). K6 a Foc/Top noun precedes every
QualP-internal PP and W (C2). K7 a clause-side PP and a noun are ordered by height alone (C2).
So: rule set A = CYC on D1 and the sub-domains; rule set B = CYC on D2. The changeover = the sealing of W.
CYC v1 (superseded 2026-10-22, 12:46pm; kept for the record):
CYC (HYPOTHESIS, ours)
For each spell-out domain D, let H(D) = the heads of D's spine and N(D) = everything else in D
(pronouns, adjuncts, phrases, and any already-spelled-out domain, which counts as ONE non-head).
C1 every x ∈ N(D) precedes every y ∈ H(D);
C2 within N(D): if x asymmetrically c-commands y, then x ⌢ y;
C3 within H(D): if x asymmetrically c-commands y, then x ⌢ y.
Domains: D1 = W (the verb word). D2 = the clause (C-type particles as heads; W, nouns and PPs as
non-heads). A PP, a possessed noun, or a relative clause is its own domain (P, N or -ę́ę
as head).
COROLLARIES (all HYPOTHESIS)
K1 Qual ⌢ M/S ⌢ v ⌢ V from C3 in D1: Hale's “head-initial” with no parameter.
K2 Obj ⌢ V; pronoun ⌢ P; possessor ⌢ N from C1: Hale's “lexical heads are right-headed” (n. 4)
with no separate setting: a pronoun complement is a non-head, so it precedes every head.
K3 [verb word] ⌢ ya’ / -go / -ę́ę from C1 in D2: the finished word is a non-head there.
K4 C3 never compares a head inside W with a head outside it. The only thing that changes at the word
is which elements count as heads: Qual, M/S, v, V in D1; nothing of W in D2.
So: rule set A = CYC on D1; rule set B = CYC on D2. The changeover = the D1/D2 boundary.
HYPOTHESISRelation to antisymmetry. C2 and C3 are Kayne's axiom (asymmetric c-command maps onto precedence; Kayne 1994 (3), pp. 5–6, and §4.1, p. 33) restricted to one plane at a time. C1 is the one thing Kayne's axiom lacks: it puts a lower non-head before a higher head. So CYC is “Kayne's axiom within each plane, plus non-heads first”. That is the smallest departure from antisymmetry that reproduces Hale's spell-out.
What would falsify CYC.
(f1) A head of W after a non-head of W in the string. None occurs in the forms in §4. Hale's receptor phonology fills slots; it does not reorder them.
(f2) Non-heads of W out of height order. SOURCED Two known cases. (i) Oblique objects “assume the highest prefixal position when not in competition with a preverb”, even preceding the Iterative (Hale 2003 n. 13, pp. 69–71, citing Young & Morgan 1987:187 and a 1987D:215 form whose characters are damaged on the page). If the oblique is lower than the iterative adjunct, C2 is violated. Hale: “not a matter entirely devoid of mystery”. (ii) Deictic subjects sit between the object and the Qualifier. Hale 2001 calls them “adjunct nominal expressions” that are “spelled out as prefixes to the conjunct sector” by an “infixation” process he restricts “to the deictic prefixes alone” (pp. 690–691). Both are exceptions to CYC, and Hale states both as exceptions to his own system.
(f3) A clause-level head before a non-head. The candidates are da’ (clause-initial only) and doo of doo … da. CYC predicts both are phrases, for example an operator in Spec,CP or an adverb. That is the option the site already prefers for da’ (derivations D6). OPEN
(f4)OPENA tension with Rice. C2 orders the disjunct adjuncts higher-first. That matches Hale's trees (11) and (31b), where the preverb is the outermost adjunct and Pl the innermost. Rice's principle, quoted secondhand, says “morphemes of greater scope appear to the right of morphemes within their scope” (Rice 2000:4, in Goad & Travis 2021 p. 2). Hale says the order among adjuncts “is determined by Rice's principles of semantic scope” (2003 p. 17), and that in the adjunct domain scope “works from right to left” (2001 p. 692). If scope = c-command, Rice's direction would reverse C2 for the disjuncts. Rice 2000 has not been opened, so this cannot be settled here (T29e).
4. Derivations
Format: each step is Merge, Agree, Move or Spell-out, tagged with the rule set it uses (A, B, Σ) and the clause of SA/SB/CYC that applies. The V-H variant (no argument movement) is shown. V-P differences are noted where the string would differ; they never change the string, because movement of a silent pronoun has no exponent (derivations D1). Role-slot steps use the schema of the proto-role schema, with Dowty's proto-role entailments as the baseline. Entailment assignments are ours and marked as such.
S1. Schematic intransitive (letters only)
Heads: V (stem), v (classifier), m (M/S). Non-head: the subject pronoun s (silent; φ on m).
1 Merge V. Σ one slot ρ1, pos = Spec,v (unergative) or Spec,V
2 Merge v with VP; Merge s in Spec,v.
3 Merge m with vP. Agree m ⟶ s: m copies s's person/number (portmanteau, Hale n. 11).
4 Spell-out D1 = mP. A H(D1) = {m, v, V}, N(D1) = {s}. C3: m ⌢ v ⌢ V. s is silent.
SA2: v = coda, m = nucleus/onset of the receptor; V fills the right half.
String: m-v-V (no movement; V-P would raise s to Spec,mP with no change to the string)
S2. Schematic full clause (letters only)
Heads in W: q (Qual), m, v, V. Non-heads in W: p (preverb, adjoined to qP), o (object pronoun).
Clause: N1 (topic noun), N2 (second noun), Prt (a clause-final particle head).
1 Merge V with o (Comp,V). 2 Merge v; Merge s in Spec,v. 3 Merge m; Agree m ⟶ s.
4 Merge q with mP. 5 Merge p, adjoined to qP.
6 Spell-out D1 = W. A N(D1) = {p, o}, H(D1) = {q, m, v, V}.
C1: p, o ⌢ q, m, v, V. C2: p ⌢ o (p c-commands o). C3: q ⌢ m ⌢ v ⌢ V.
String(W) = p-o-q-m-v-V. (SA2 fills receptor with m, v; SA4 prefixes q, then o, p)
7 Merge N2, N1 (clause level: adjoined above W, N1 higher), Merge Prt. B
8 Spell-out D2. B N(D2) = {N1, N2, [W]}, H(D2) = {Prt}.
C1: all non-heads ⌢ Prt. C2: N1 ⌢ N2 ⌢ [W] (W is the lowest non-head).
String: N1 N2 [p-o-q-m-v-V] Prt
HYPOTHESIS S2 shows the changeover. In step 6, q, m, v and V are ordered head-first. In step 8 the same material is one block, and the clause head Prt comes last. No step reverses a parameter.
N1. Local-person transitive (no voice alternation)
Hale 2003 (1), p. 5; (6) p. 8; tree (8) p. 11
ni-sh-hozh
2SG-1SG-tickle
‘I tickle you.’
Hale prints the gloss as “2s-1s-tickle”. His (6) analyses the word as Obj ni / Mode Ø / Subj sh / Cl ł / Stem ghozh; (8) puts ni as complement of V and M/S sh above vP.
Σ Σ(√GHOZH, ł) = {ρA (Spec,v), ρP (Comp,V)}, as on derivations D2. Roles in the action: ρA volitional, causing; ρP affected (our assignment, HYPOTHESIS).
Merge V ghozh with ρP (= ni). Merge v ł. Merge ρA in Spec,v.
Merge M/S (Mode Ø). Agree M/S ⟶ ρA [1SG] → portmanteau exponent sh.
Spell-out D1 A N = {ni}, H = {M/S, v, V}. C1: ni first (SA3: ni takes the receptor's onset position only if the nuclear heads leave it empty; either way it is leftmost). C3: Ø-sh ⌢ ł ⌢ ghozh. → ni-Ø-sh-ł-ghozh, surfacing as nishhozh (the ł + gh → h fusion is Hale's phonology, not derived here).
B D2 has no further material. Hale's (2) Ni ni-sh-hozh ‘I tickle YOU’ adds a contrastive pronoun as a D2 non-head before [W].
Moves: 0. HYPOTHESIS Structural relation kept: the subject in Spec,v asymmetrically c-commands ni, as in Hale (8).
N2. A low preverb that does not scope over the subject
Hale 2003 (13), p. 18 (verb of (12), from Young & Morgan 1987:772); (14) p. 19
yisdá-Ø-s-ł-tsóós
to.safety-3.OBJ-1SG.SBJ-CLF-(stem, unglossed)
surface yisdástsóós ‘I rescue it.’
Segmentation from Hale's surface template (13b): Preverb yisdá / Obj Ø / M/S sh (“assimilated to [s]”) / v ł / V tsóós. Hale gives no morpheme gloss line. The gloss is compiled from his labels: ‘to safety’ (p. 15), “third person (direct) object is phonologically nonovert if the subject is a ‘local’ … person” (p. 18). Hale gives no separate meaning for the stem tsóós, so it is left unglossed. Hale on scope: “the preverb in (13) does not have the subject in its scope. It is the coat alone that ‘comes to be in safety’” (p. 19).
Σ {ρA (Spec,v), ρP (Comp,V)}.
Merge V tsóós with Obj Ø. Merge v ł; ρA in Spec,v.
Merge Pv yisdá, adjoined to vP (Hale's (14): below M/S, so the subject is outside its scope).
Merge M/S; Agree M/S ⟶ ρA [1SG] → sh.
Spell-out D1 A N = {yisdá, Ø}, H = {M/S, v, V}. C1: yisdá before the heads, although it is lower than M/S. C2: yisdá ⌢ Ø. C3: sh ⌢ ł ⌢ tsóós. → yisdá-Ø-sh-ł-tsóós.
Moves: 0. HYPOTHESIS This is the key form for the comparison in §6. Under CYC the preverb's place in the string and its scope (under the subject) both follow from one base position.
N3. Transitive with two overt nouns and a low preverb
Hale 2003 (20b), p. 26; tree (21) p. 26; derivation (25)–(30), pp. 33–37
Mósí
jooł
na#y-Ø-ł-maas
cat
ball
around#3OBV.OBJ-IPFV.3SBJ-TR-roll
surface Mósí jooł neiłmaas. ‘The cat rolls the ball around.’
Hale's segmentation line is “mósí jooł na#y-Ø-ł-maas” (page image, p. 26). Gloss compiled from his labels: Pv na (‘around’, from the translations of (20a, b)), Obj y “the overt obviative third person pronoun (3o)” (p. 27), M/S Ø “third person zero-imperfective” (p. 27), v ł “the transitive voice element” (p. 34), V maas ‘roll (of spheroid entity)’ (p. 35). The nouns are “adjoined to the full verb word (or verbal clause)” (p. 27). (30) gives the word as neiłmaas ‘3rd (subject) rolls 3rd (object) about’ (p. 37).
Σ Labile root (Spec,V argument, (22)). Transitive: {ρA (Spec,v), ρP (Spec,V, surfacing as object)}. Roles in the action: ρP moves (it is the ball that rolls; Hale p. 26). ρA causes.
Merge V maas, ρP = y. Merge v ł, ρA in Spec,v. Merge Pv na at vP ((21)).
Merge M/S Ø; Agree M/S ⟶ ρA [3].
Spell-out D1 A N = {na, y}, H = {Ø, ł, maas}. SA2: ł = coda; neutral vowel = nucleus (27). SA3: y = onset (28). SA4: na prefixed (29a). CYC gives the same order: C1, then C2 na ⌢ y, then C3. → na-y-Ø-ł-maas → neiłmaas (Hale's phonology).
Mergejooł, Mósí in the clause (D2), linked to ρP, ρA (W&J principles, as Hale says). B
Spell-out D2 B N = {Mósí, jooł, [W]}; C2 with Mósí highest: Mósí ⌢ jooł ⌢ [neiłmaas].
Moves: 0. HYPOTHESIS The order Mósí before jooł is attributed to height (topic higher). Hale's p. 27 attributes the nouns' order to discourse principles (W&J). C2 is consistent with that only if the topic is the higher adjunct, which is a site assumption, not Hale's statement.
N4. A full verb word: three disjunct adjuncts, a qualifier and an object
Hale 2003 (31a), (31b), (32), (33), pp. 38–40 (page image p. 38)
surface Yisdánídashidoołtééł. (~ Yisdáádashidoołtééł.) ‘They will get me back to safety.’
The underlying pieces and labels are Hale's (31b) tree: Pv yisdá, Rev ní, Pl da, Qual d, M/S w, v ł, Obj sh, V tééł. The gloss is compiled from his prose: w is the gamma-perfective situation aspect “in portmanteau with 3rd person … in this future verb form” (p. 39); d is a qualifier (the future “is rendered by this same complex of elements preceded by the inceptive qualifier [d]”, Hale 2001:691); doołtééł means “3 subject will handle animate object” (p. 40). DISTR.PL follows Hale's template label “(Distributive) Plural” (ex. (10)). Surface doo < d-i-w-i by Hale's assimilation and glide deletion (p. 40), not derived here.
Σ {ρA (Spec,v), ρP = sh (Comp,V)}.
Merge V tééł with sh. Merge v ł, ρA in Spec,v. Merge M/S w; Agree M/S ⟶ ρA [3].
Merge Qual d (selects M/S, p. 32).
Merge Pl da, then Rev ní, then Pv yisdá, each adjoined to QualP (Hale: “adjoined to the maximal projection determined by the highest nuclear element”, p. 39). Pv scopes over subject and object (pp. 15–16).
Spell-out D1 A SA2: ł coda, neutral vowel nucleus, w onset. SA4: Qual d prefixed (nuclear first), then sh, da, ní, yisdá, “working leftward (upwards in the structure)” (p. 40). → (33) yisdá ní da sh diwiłtééł. CYC: N = {yisdá, ní, da, sh} by height (C2); H = {d, w, ł, tééł} by height (C3); C1 puts N first. Same string.
Moves: 0. HYPOTHESIS N4 is the strongest test of PL(a)/C1. The object sh is the lowest element in the tree, yet it precedes the Qualifier, the highest head. Height order among non-heads (C2) and head order (C3) are each respected; only C1 interleaves them.
N5. A causative: two classifiers, a low preverb and an oblique causee
Hale 2003 (34), (35), (36), p. 41 (page image)
ha#b-y-sh-ł-l-gheed
up.out#3.OBL-QUAL-1SG.SBJ-CAUS-CLF-run
surface habiishyeed ‘I run him/her/it up out.’ (idiomatically) ‘I start it (car).’
Segmentation line as printed. Labels from Hale's (35)–(36): Pv ha, Obj(obl) b (specifier of the inner vP2), Qual y, M/S sh, v1ł (outer, causative), v2l, V yeed. ‘up out’ follows Hale's gloss of ha in ha#chééh (p. 19). CAUS for v1 follows Hale's prose: “the embedding of the maximal projection of a v-phrase … as the complement of another v” (p. 41).
Σ Inner unergative: {ρ1 (Spec,v2)}. Added causer: ρC (Spec,v1). ρ1 is relinked as oblique b. This is event-structure layering (two v layers) and theta slots on the same participants, which are compatible descriptions (see plain ł-causer).
Merge V yeed; v2l; b in Spec,v2; Pv ha adjoined to vP2 ((36)).
Merge v1ł; ρC in Spec,v1. Merge M/S; Agree M/S ⟶ ρC [1SG] → sh. Merge Qual y.
Spell-out D1 A N = {ha, b} (C2: ha adjoined above b, so ha ⌢ b); H = {y, sh, ł, l, yeed} (C3). → ha-b-y-sh-ł-l-yeed → habiishyeed.
Moves: 0. SOURCED Hale's n. 13 records Jonathan Bobaljik's objection: the preverb modifies the event, so the causee “should not be in its scope”. Hale replies with Young & Morgan's position-shift principle (pp. 69–71). HYPOTHESIS CYC inherits this problem unchanged. The order ha ⌢ b encodes ha c-commanding b, which is the scope Bobaljik says is wrong.
N6. Clause level: postposition, question particle, enclitic, tag
(2) ‘Do you have sheep?’ (4) Dibéésh nee hólǫ́? ‘Do you have sheep?’ (sheep-Q 2-with exist.NI) (15) Dibé nee hólǫ́, yaʼ? ‘You have sheep, don’t you?’
As on derivations D6; glosses as printed by the NLA (“2-with”, “exist.NI”). “It is ungrammatical for da’ to appear in any other position.”
Merge P -ee with the 2nd-person pronoun. Spell-out PP domain: C1 pronoun ⌢ P → nee (SB4 / K2). B
Merge W = hólǫ́ (internal structure not given by the source; Σ one slot for what exists, as on D6). Spell-out D1 A.
Merge PP and dibé in the clause; then C[Q]. B
Spell-out D2 B. (15): H(D2) = {ya’}; C1 → dibé nee hólǫ́ yaʼ ✓. (2): if da’ ∈ H(D2), C1 predicts *dibé nee hólǫ́ daʼ, which the NLA marks ungrammatical (3c). So CYC requires da’ ∈ N(D2): a phrase, highest non-head, hence first (C2) ✓. (4): -ísh placed by SB5, outside CYC.
Moves: 0. OPEN CYC is falsified if da’ is shown to be a head (C⁰).
N7. Clause level: a relative clause inside a postpositional phrase
NLA Navajo Questions, yes/no ex. (9)
[shi-zhéʼé
tłʼóół
yizhbizh-ę́ę]
b-ee
1SG-father
rope
3-3-braid.PFV-PST†
3-with
SOURCED † Segmentation note (2026-11-11 audit pass 4): The NLA gloss line (“3-3-braid.P-past”) prints yizhbizh-ę́ę] with the gloss “3-3-braid.PFV-PST”: the gloss names more morphemes than the form shows as separate pieces (fused or unwritten morphemes), so the two lines do not match one-to-one in the source. Kept exactly as printed.
‘with [the rope my father braided]’
NLA gloss line “1-father rope 3-3-braid.P-past 3-with”, as on headedness §2.1. Leipzig rendering: P = PFV, past = PST. The verb is not segmented further.
Possessed noun: Spell-out its domain, C1 possessor ⌢ N → shi-zhéʼé. B
W yizhbizh (D1, unsegmented) A; the clause with nouns shizhéʼé > tłʼóół (C2); head -ę́ę (C1: last). B
PP: head P -ee with its 3rd-person pronoun b- (C1: b ⌢ ee). The relative clause is a non-head in the PP domain (C1: before bee). B
Moves: 0. HYPOTHESIS Three nested domains (possessed N, relative, PP), with the same rule in each.
N8. Two third persons with nouns: direct and inverse
Willie & Jelinek 2000:273, (61a) and (62b)
łį́į́ʼ
tsé
yiztał
horse
rock
3ACC-3NOM-kicked
‘The horse kicked the/a rock.’ (Direct) (62b) ʼashkii łį́į́ʼ biztał ‘The boy was kicked by the/a horse.’ (Inverse)
Glosses as printed (W&J gloss the word as a whole); the verb is not segmented here.
Σ {ρA, ρP}; voice yi-/bi- from the ranking (Dowty-based weighting; inverse).
Spell-out D1 Ayiztał / biztał (internal order not derived: the source does not segment).
B D2: the topic noun is the higher non-head (C2) → łį́į́ʼ tsé [W]; ʼashkii łį́į́ʼ [W].
Moves: 0.
5. The antisymmetric alternative
5.1 What Kayne 1994 says (opened this run)
SOURCEDThe axiom. “(3) Linear Correspondence Axiom: d(A) is a linear ordering of T”, where A is the set of pairs of nonterminals such that the first asymmetrically c-commands the second, and d maps nonterminals to the terminals they dominate (§1.2, pp. 5–6). Asymmetric c-command maps onto precedence (§4.1 (2), p. 33; argued from the asymmetry of time, §4.3).
SOURCEDOne order. “specifier-head-complement, and not the reverse, is the only order available to the subcomponents of a phrase” (§4.2, pp. 35–36). SOV “is perfectly allowable if taken to indicate a phrase marker in which the complement has raised up to some specifier position to the left of the head” (§4.1, pp. 33–35).
SOURCEDNo direction parameter. “there cannot be any directionality parameter in the standard sense”; postpositional phrases “must be derived by moving the complement of the adposition into the specifier position of that adposition (or of a higher functional head)”; “In a language where IP precedes C⁰, IP must have moved leftward into Spec,C⁰” (§5.1, pp. 47–48). “Head-final” is redefined: a head is final “if its complement comes to precede it (by moving to some higher specifier/adjoined position)”, and “in every representation, at least one head must be initial” (Preface, pp. xiv–xv).
SOURCEDOne specifier or adjunct per phrase. Adjoining more than one non-head to a non-head is impossible; specifiers are a case of adjunction; so a phrase has “at most one adjoined phrase (including a specifier)” (§3.4, pp. 21–22; §3.6, p. 27). No two clitics can adjoin to the same head (§3.3, p. 19ff). Functional heads exist partly to supply landing sites (§3.6).
SOURCEDBelow the word. “an adjoining head will invariably precede the head that it adjoins to” (§4.4, p. 38). The LCA applies to structure below the word level. A word is an adjunction structure whose head (the element adjoined to) is rightmost, which Kayne offers as the account of right-headed words (§4.5, pp. 38–42).
SOURCEDNo rightward head movement, so “in a head-final language a ‘V-T-Agr’ sequence cannot be derived by raising V rightward to T and then raising the result rightward to Agr” (§5.4, pp. 50–51). Instead, roll-up: “X[YP … Y ZP] → X[YP ZP Y t] → [YP ZP Y t] X”, which yields a sequence YX that is not a constituent, possibly the source of “pure agglutination” (§5.5 (3), pp. 52–53).
SOURCEDNavajo, by name. “Kenneth Hale (personal communication) notes that although there are languages, like Navajo, with obligatory agreement between an adposition (postposition) and its lexical complement, prepositional phrases in SVO languages never, as far as he knows, show such agreement”. Kayne's explanation: the complement of a postposition is in, or has passed through, the specifier of P (§5.3, p. 49).
SOURCEDFinal complementizers. If final C means IP in Spec,CP, three predictions follow: no that-trace effects with final complementizers (conjecture (6), p. 53); nominative anaphors possible; and wh-movement to Spec,CP generally absent in SOV languages, because Spec,CP is filled (p. 54).
Access: text of the PDF hosted on the NYU Linguistics site, read through a web fetch (a direct download from the box was refused). The copy is an OCR text with few running heads; page numbers are section-start pages from its table of contents, or the running head next to the passage where one is visible. Notes saved at [local working copy, not in the repository].
5.2 What Goad & Travis 2021 say about antisymmetry and the Navajo template (re-opened this run)
SOURCED Speas (1991b) “starts with a head-final tree … then proposes iterative lowering”. Goad & Travis object that “the structure is head-final which is a problem for those who espouse Kayne's antisymmetry hypothesis (Kayne 1994), as we do”, and that “rampant lowering” is not ideal (p. 6; Speas 1991b itself not opened).
SOURCED Harley (2011): head-final, two head movements, Local Dislocation and morphological merger; “uncharacteristically complex”, and “any order could be generated within such a powerful movement system” (p. 7; Harley 2011 not opened).
SOURCED Morphological Merger account: “the underlying structure is head-initial”, but it predicts a hard prosodic-word edge between positions 8 and 9, which their Tsilhqútʼín data do not show (pp. 8–9, 18).
SOURCED Iterative Root Prefixation (IRP): head-initial, head movement from V up, each affix prefixed “directly to the left edge of the root” (p. 11, (7)). Position 6 (the Qualifier) = v. Position 9 (classifier) = ExtArg, merged between 6 and 7 (“6.5”) and placed next to the root by prosody (pp. 4–5, 12, (8)). Disjunct prefixes are “late adjuncts spelled out in separate phases” (Travis 2010, via n. 14, p. 8, and p. 17). The verb “is not a phrase but a complex head” (n. 21, p. 13). They leave open why the pattern is rare: “ideally … this difference would follow from some particular characteristic of the left edge of the root” (p. 34).
5.3 Two antisymmetric versions, stated for Navajo
AS-1 (phrasal, no head movement; HYPOTHESIS: Kayne's axiom applied to Hale's hierarchy)
K1 LCA (Kayne (3)); asymmetric c-command → precedence.
K2 One specifier/adjunct per phrase; every extra one needs its own (silent) head F.
K3 No head movement in W (Hale p. 30): Qual, M/S, v, V stay in situ.
K4 ROLL-UP: a head marked [+R] attracts its complement into its specifier (Kayne §5.1).
[+R]: V? (see K5), P, N (possessor), clause-final particles (ya’, -go, -ę́ę).
[−R]: Qual, M/S, v; da’ (if C⁰).
K5 NR (nonhead raising): every NNCL element of W must end up asymmetrically c-commanding Qual
(else it would follow it), each in its own Spec,F, with the base height order preserved.
For the object this is Hale 2003:11–12's “raising the object to the Spec position”, which he
rejected as Case-driven; on the site it would be EPP-driven (framework audit F4).
AS-2 (complex head + IRP; Goad & Travis 2021, Travis 2010/2011 via G&T)
K1–K2 as above; head movement builds one complex head from V upward; each new affix attaches at
the left edge of the root (IRP); the result is a nested left-adjunction structure
[1 [2 [3 [ … [9+root]]]]]. Hierarchy (G&T (8)): T(7) > ExtArg(6.5 = 9) > v(6), subject agreement (8)
above, an inner inflectional domain (positions 1–5) inside vP (p. 9).
HYPOTHESISIs IRP's output LCA-compatible? Each IRP step adjoins the new affix to a different node: the root complex, then a node inside the earlier output. Kayne allows exactly this kind of nesting, one element adjoined to another adjoined element (§3.3, his (10)), and forbids only two adjunctions to the same node. So the nested structure linearizes as G&T state, by our check, not theirs. The cost is IRP itself, an attachment rule that works only inside words.
HYPOTHESISRice's scope analysis is excluded under antisymmetry. Hale renders Rice with the stem right-adjoining to Voice, and then Voice+V right-adjoining to Mode/Subj (Hale 2001:679–680; headedness (iii)). Kayne forbids right adjunction (Preface p. xiii) and rightward head movement (pp. 50–51). Rice 2000 itself is unopened, so this applies only to Hale's rendering.
6. The same forms under AS-1
Each derivation lists only the steps AS-1 needs beyond the CYC derivation in §4. Merge and Agree steps are the same. Every extra head F is silent. Strings are checked against the sourced surface order.
N1′ nishhozh (Hale (1))
LCA with no movement: M/S Ø-sh ⌢ ł ⌢ ghozh ⌢ ni (the complement follows V). ✗
Merge F1 above M/SP. Moveni → Spec,F1 (NR), crossing the subject in Spec,v.
Spell-out by LCA: ni ⌢ F1 ⌢ Ø-sh ⌢ ł ⌢ ghozh ✓.
Extra: 1 move, 1 silent head. HYPOTHESIS Consequence: ni now asymmetrically c-commands the subject.
N2′ yisdástsóós (Hale (13))
Hale's (14) has a vP with both a specifier (the subject) and an adjunct (yisdá). K2 forbids that, so Merge F0 to host yisdá instead.
LCA would give sh ⌢ yisdá. ✗ Merge F1 above M/SP. Moveyisdá → Spec,F1 (NR). (The silent object Ø's position is not observable.)
Spell-out: yisdá ⌢ sh ⌢ ł ⌢ tsóós ✓.
Interpretation: Hale's reading (subject outside the preverb's scope, p. 19) now requires reconstruction of yisdá to its base below M/S.
Extra: 1 move (2 if the silent object must also raise), 2 silent heads, plus obligatory reconstruction.
N3′ Mósí jooł neiłmaas (Hale (20b))
vP has the subject in its specifier and na adjoined: needs F0 for na (K2).
y must precede ł: Movey → Spec,F1 above v (at least). It must still follow na, so F1 is below F0. M/S is null, so how high the movement goes cannot be observed.
Clause: two nouns adjoined to the same projection are forbidden (K2), so Merge G1, G2, with Mósí in Spec,G1 above jooł in Spec,G2.
Spell-out: Mósí ⌢ jooł ⌢ na ⌢ y ⌢ Ø ⌢ ł ⌢ maas ✓.
Extra: 1–2 moves, 4 silent heads.
N4′ Yisdánídashidoołtééł (Hale (31a))
Hale's (31b) has three adjuncts on QualP, which K2 forbids: Merge F1 > F2 > F3 above QualP, hosting yisdá, ní, da.
sh (Comp,V) must precede Qual d: Merge F4 between F3 and QualP; Movesh → Spec,F4 (NR), crossing Qual, M/S, the subject and v.
Spell-out: yisdá ⌢ ní ⌢ da ⌢ sh ⌢ d ⌢ w ⌢ ł ⌢ tééł ✓ (= Hale's (33)).
Extra: 1 move, 4 silent heads. HYPOTHESIS The heads d, w, ł, tééł come out in the right order with no movement. This is what antisymmetry gets for free.
N5′ habiishyeed (Hale (34))
vP2 has b in its specifier and ha adjoined (Hale's (36)), which K2 forbids: F0 for ha.
Both must precede Qual y: Moveb → Spec,F2, then ha → Spec,F1 (F1 > F2 > QualP), crossing y, sh, the causer in Spec,v1 and ł.
Spell-out: ha ⌢ b ⌢ y ⌢ sh ⌢ ł ⌢ l ⌢ yeed ✓.
Extra: 2 moves, 3 silent heads. Bobaljik's scope problem (n. 13) is unchanged.
N6′–N8′ Clause level
N6′nee: K4 roll-up, pronoun → Spec,P, which licenses agreement (Kayne p. 49; 1 move). dibé and the PP each need a host head (K2; 2 silent heads). (15) ya’: IP → Spec,ya’ (1 move). (2) da’: a [−R] C⁰, or a phrase in a specifier. AS-1 allows both, so it is not falsified by a head da’, unlike CYC.
N7′ Roll-up three times: possessor → Spec,N; relative IP → Spec,-ę́ę; relative DP → Spec,P, licensing b- agreement (3 moves).
N8′ Two nouns: two host heads, topic higher (2 silent heads, 0 moves).
7. Results table
Form (source)
CYC string
CYC moves
AS-1 moves
AS-1 extra silent heads
AS-1 consequence
S1, S2 (schematic)
✓
0
S2: 1 (o) + roll-up of W into Spec,Prt
2–4
o above subject
N1 nishhozh (Hale (1))
✓
0
1
1
object c-commands subject
N2 yisdástsóós (Hale (13))
✓
0
1 (2)
2
preverb must reconstruct below subject (Hale p. 19)
N3 Mósí jooł neiłmaas (Hale (20b))
✓
0
1–2
4
object above v (height not observable: null M/S)
N4 Yisdánídashidoołtééł (Hale (31a))
✓
0
1
4
conjunct heads free; object crosses Qual, M/S, subject, v
N5 habiishyeed (Hale (34))
✓
0
2
3
preverb and causee cross the causer; Bobaljik's problem unchanged
N6 daʼ / -ísh / yaʼ (NLA (2), (4), (15))
✓ only if da’ is a phrase
0
2 (P, ya’)
2
da’ head or phrase both fine; P agreement explained (Kayne p. 49)
N7 relative in PP (NLA (9))
✓
0
3
0–2
b- agreement explained
N8 W&J (61a), (62b)
✓
0
0
2
none beyond host heads
Known exceptions
oblique leftmost (Hale n. 13); deictic subject infixed (Hale 2001:691)
n/a
oblique: an extra raising; deictic: a clause noun lowered into W, which needs lowering (Kayne sets lowering aside, p. 47) or a second NR target
n/a
neither option derives them without an extra statement
HYPOTHESISCount. Over the five Hale words, CYC needs 0 movements. AS-1 needs 6–8 movements and 14 silent heads. Of those heads, 8 come only from the one-specifier/adjunct limit, and the movements are all of one type (NR). The four conjunct heads come out right under both, with no movement.
8. Where the options predict differently
#
Question
CYC / Hale 2003
AS-1
AS-2 (IRP)
Data
P1
Does the object prefix c-command the subject?
No (Obj in VP, subject in Spec,v)
Yes at the surface (NR); scope Obj > Subj unless reconstructed
Agreement affixes are heads in one complex head; c-command between arguments not fixed by order
Rice: subject scopes over object (secondhand, Rice 2000:4 via G&T p. 2; Bird 2001). Fits CYC; AS-1 must reconstruct. OPEN no Navajo scope judgment opened
P2
Low preverbs (Hale (13), (14))
Base position gives both order and scope
Must move over the subject and reconstruct; movement should show locality effects
Disjuncts as separate-phase adjuncts (Travis 2010 via G&T p. 8)
OPEN a case where the preverb's movement would be blocked (an island or intervener inside W) would separate CYC from AS-1
P3
Is da’ (or doo) a head?
Must be a phrase
Either
Not addressed
OPEN (D6 diagnostics: co-occurrence with fronted háísh, embedding)
P4
Order of several clause-final particles
Lower first (C1/C3 with nested domains)
Lower first (roll-up mirror)
Not addressed
HYPOTHESISCorrection: this test, listed on headedness (v) as a falsifier for roll-up, does not separate roll-up from a head-final or CYC account. All predict the same order.
P5
Kayne's final-C correlations
No prediction
No that-trace with final C (p. 53); no wh-movement to Spec,CP (p. 54)
No prediction
W&J: “no obligatory Wh-movement in Navajo, but there may be focus (leftward) movement” (p. 272). Consistent with AS-1 if the fronting targets a focus position, not Spec,CP. That-trace: OPEN
P6
Postposition agreement
Pronoun is P's argument (SB4); noun linked
Complement passes Spec,P (Kayne p. 49)
Not addressed
Both fit bee, nee; AS-1 also explains the cross-language gap Hale reported to Kayne
P7
Prosodic edge between positions 8 and 9
None: M/S and v share the receptor syllable (SA2)
Not fixed by syntax
None (IRP), confirmed for Tsilhqútʼín (G&T pp. 18, 26–33)
G&T's data are Tsilhqútʼín, not Navajo; consistent with CYC and AS-2, against Morphological Merger
P8
Qualifier above or below Mode/Subject?
Above (Hale p. 32)
Above (to precede it with no movement)
Below: v(6) under T(7) (G&T (8), p. 12)
OPEN scope of a Qualifier meaning (inceptive, semelfactive) relative to mode/aspect, as asked on headedness §5
8.1 Does each option get the word as changeover for free?
Option
Is the word a boundary for ordering?
Free?
(i′) domain-indexed head parameter (old)
Yes, by a domain index on the parameter
No: stipulated
CYC (this page)
Yes, as a spell-out cycle; same rule in both cycles
Partly: the ordering rule is shared, but where the cycle ends is stipulated (Hale n. 12 calls it open)
AS-1 (Kayne axiom, no head movement)
No: one axiom above and below the word (Kayne §4.5); the attract/no-attract split puts V (inside the word) with P, N and final particles (outside)
No: the word is only prosodic
AS-2 (IRP, G&T)
Yes: IRP applies inside the complex head built by head movement
No: IRP is a word-only rule; its rarity is unexplained (G&T p. 34)
Rice (secondhand)
No flip: head-final throughout, with rightward stem raising
n/a; excluded under antisymmetry (Kayne pp. xiii, 50–51)
9. What each option gets right and misses
Option
Gets right
Misses or must add
Hale 2003 spell-out (SA1–SA5)
All five Hale words with no syntactic movement; low and high adjunct readings from base position (pp. 16–19); M/S and v together in one receptor syllable
The “planes” are only expository in his text (p. 33), yet the order depends on them; “when to stop” is open (n. 12); the oblique-object shift is unexplained (n. 13); deictic subjects need infixation (2001:691); Bobaljik's scope objection is unanswered
CYC (ours)
Everything Hale's procedure gets, plus the clause in the same terms; derives Hale's lexical/functional split (K2) and the apparent head-final clause (K3); gives Grant's changeover a precise meaning (the cycle boundary)
Stipulates the cycle; inherits Hale's three exceptions; predicts da’ and doo are phrases (untested); possible conflict with Rice's scope direction for disjuncts (f4); with “nouns enter in D2” it bears on Condition C (§10)
AS-1 (Kayne)
The conjunct head order with no movement and no parameter; postpositions, final particles and P-agreement by one mechanism (Kayne pp. 47–53, 49); da’ head or phrase both fine
Hale's rejected object raising, generalized to every prefix (NR); 14 silent heads for five words; reconstruction for low preverbs; no account of why the word is a word; deictic infixation needs lowering, which Kayne sets aside (p. 47)
AS-2 (IRP; Travis, G&T)
Antisymmetric and Mirror-free with one attachment rule; position 9 next to the root via prosody; supported by Tsilhqútʼín bracketing (G&T §4.3)
A word-only rule (rule set A again); a hierarchy that conflicts with Hale's on Qual vs M/S; rarity unexplained; Navajo-specific phonological evidence not given in the paper
Speas 1991b, Harley 2011 (via G&T only)
Each derives the template
Head-final bases; lowering or several post-syntactic operations (G&T pp. 6–7). Not opened: no further verdict here
10. STRATA, feature bundles, proto-roles
GRANT'S CONJECTURESTRATA (theta roles first, event-structure roles as another layer, noun selection later). HYPOTHESIS CYC gives the strata a spell-out correlate. The role slots and their pronoun bundles, and the event-structure layers (one v, or two in the causative N5), are all built and spelled out in D1. The nouns come in D2, after the word is finished. Theta slots and event layers sit side by side in D1 as compatible descriptions of the same participants (N5, step 1). This makes STRATA a natural fit for CYC, but it is an architectural match, not evidence.
HYPOTHESISA consequence to check against Condition C. If nouns enter only in D2, every noun enters after a subject pronoun that stays in Spec,v (Hale's B1). By A3′ on Condition C by site, no noun is then in the subject's reach. That predicts Speas's free coreference for Hale 1999's (10) and (11a), where Hale reports disjointness. So CYC with “nouns in D2” is compatible with Hale's judgments only if (a) the subject bundle re-merges in the clause cycle (raising out of the word, the site's V-P beyond W), or (b) nouns can be merged before W is spelled out but pronounced in D2, as with Travis's late-spelled adjuncts read in reverse. This sharpens T28a: for STRATA and CYC together, whether the subject moves is decisive.
HYPOTHESISFeature bundles (FB). D1 contains only heads and pronoun bundles, never a free noun. That fits Hale's “Navajo does not itself have noun incorporation” (2003 p. 13). Willie's incorporated body parts (possessor raising) are analysed there as deriving a complex verb, not as arguments, so on that analysis they are not counterexamples. Nothing on this page counts against FB.
Proto-roles. The Dowty-based slot choices (Σ steps) are unchanged by any linearization option. Linearization decides where a bundle is pronounced, never which participant it stands for. Navajo's ranking and topicality extras stay extensions of Dowty (inverse).
11. Gaps and next (T29a–j)
#
Task
Needs
T29a
OPEN Is da’ a head or a phrase? Is doo of doo … da? CYC requires phrases (f3).
NLA pages; texts with da’ plus a fronted háísh, or da’ in embedded questions
T29b
OPEN Scope inside the word. Does the object prefix ever scope over the subject? (AS-1 says yes at the surface; CYC/Hale say no.)
Scope or binding judgments; Rice 2000 (not opened) for the subject-over-object generalization first-hand
T29c
OPEN Is a low preverb's leftward position blocked anywhere a movement would be? That would test NR (AS-1).
Young & Morgan 1987 preverb entries; Hale 2001 on preverb attachment
T29d
OPEN Qualifier above or below Mode/Subject (Hale vs G&T): scope of inceptive or semelfactive qualifiers relative to mode/aspect.
Judgments; Young & Morgan 1987:167–199 (aspect)
T29e
OPEN Rice's scope direction for the disjunct adjuncts versus C2's higher-first order (f4).
Rice 2000 (CUP), ch. on disjunct order
T29f
OPEN The two exceptions: oblique objects leftmost (Hale n. 13) and deictic-subject infixation (Hale 2001:691). Can either follow from CYC with real planes (Bobaljik's suggestion), or from a separate cycle? Update 2026-10-18, 8:46pm: the deictic half is restated on subject on the head as head-adjunction to the highest nuclear head (SH-4, after Hale 1997 “a prefix to E” and Hale 2001 “prefixes to the Qualifier element”), so CYC C1/C3 derive it with no exception HYPOTHESIS. The oblique half is still open. Update 2026-10-19, 4:46am: Hale's handout prints a Y&M 1987:187 inchoative, (23a) Bináádaʼniilghal, with the oblique bi- before náá- and da-, and its tree (23c) merges the null-P phrase above both. With that merge site, C2 derives the order with no exception (IN-5, the ‘start to’ test ID7, HYPOTHESIS). Why it merges there, and the preverb competition, stay open (T38f).
Young & Morgan 1987:187, 215 (page images); Hale 2000 (not opened) Update 2026-10-19, 12:46pm: HMP's inchoative manuscript gives the oblique shift a cause: the inchoative's own ’ fills position IV, so the object prefix is “bumped … leftward into the disjunct prefix Position I” (n. 2), for unaccusative participants too ((10a) shináʼniidlóóh). “Certain Position I prefixes block this” (n. 4 sodiʼniiszin), which matches Hale 2003 n. 13's preverb competition. Restated as IN-5′ BUMP. HMP's head movement (V right-adjoined to M, M to E) gives C3 (HMP page §5, IN-10). Why the far left is the escape site is still OPEN (T39c).
T29g
OPEN Where does the word cycle stop? Hale's ka example (n. 12) with -ísh between preverb and verb: is the boundary B1 (#), B2 or something else, and does CYC need a cycle per disjunct (Travis 2010 via G&T)?
A clean copy of Hale 2003 n. 12's forms; Travis 2010/2011; Young & Morgan
T29h
OPEN CYC + STRATA + Condition C (§10): does the subject bundle re-merge outside the word? (Joins T28a.)
Same as T28a; plus a statement of when nouns are merged
T29i
OPEN Kayne's final-complementizer predictions for Navajo: that-trace with -go/-ę́ę clauses, and the landing site of focus fronting (P5).
OPEN Derive Willie's incorporated body-part verbs under CYC: is the noun root a non-head of W or part of a complex V?
Willie 1991 Ch. 7 page images (opened earlier)
Sources opened or re-checked for this page (2026-10-16, 12:46pm PT). Kayne 1994: newly opened (§5.1). Goad & Travis 2021 preprint pp. 2–18, 26–34 (box [local working copy, not in the repository]). Hale 2003 ms. text (box [local working copy, not in the repository]) at pp. 5, 8–19, 26–41, 63–72, with page images of pp. 18, 26, 38, 41, 69. Hale 2001 (box text) pp. 689–692. NLA question pages and W&J 2000 used as recorded on derivations and headedness. See sources.
2026-11-14 overhaul (approved by Grant for the live site, 2026-11-14). This page used to be framed as a vP/TP comparator for Willie & Jelinek’s S-adjunction trees, with silent Top and Foc heads and the older CYC/MVS rules. It is now framed by the spine (derivation): V→v→T and subject raising to Spec,TP are assumed in tensed clauses, and each element below is placed by asking what else moves. The earlier page is kept unchanged under History, including its L1–L8 derivations. No Navajo form, gloss, judgment or citation changed.
Every tensed Navajo clause is built on the same structure: the verb with its object, the agent above them, then tense, then (when present) negation and the clause-type head. The verb rises into tense, and the subject rises to the specifier of tense. This page lists where each other element sits on that structure, what sourced data place it there, and what is still open. It then restates Willie & Jelinek’s examples, earlier derived with silent topic and focus heads, on the spine.
HYPOTHESIS [CP [NegP [TP Subject [T′ [vP ⟨Subject⟩ [v′ [VP Object V⁰] v⁰]] T⁰]] Neg⁰] C⁰], heads drawn final (HS-1 on headedness; HS-2 is the live rival). V→v→T and Subject→Spec,TP are the working default.
Element
Merged
What else moves
Basis
Status
Verb word
V⁰, with v⁰ and T⁰ exponents
V→v→T (default)
Verb-final order; inside-word order is a separate spell-out question (headedness)
Default; word order inside OPEN
Agent / subject
Spec,vP
→ Spec,TP (default)
Direct order agent before patient (łį́į́ʼ dzaanééz yiztał (NX-029))
Default
Patient / object
Complement of V⁰
Nothing in the direct form; in the inverse, moves above the raised agent (H3)
A CP with =go in C⁰, adjoined at the left of the main clause at the TP level or higher (H47)
Nothing: each clause has its own spine
A main-clause noun subject follows the =go clause: yah ʼííyáa=go ʼatʼééd dah neezdá (NX-397; Willie 1991:134 (4)), translated “When he came in, the girl sat down.” So the =go clause is above the main Spec,TP (review 13); also Willie (7)–(16b), W&J (49)
H47 HYPOTHESIS, forced by NX-397 under the working default; category of =go beyond C⁰ OPEN
Topic and focus without silent heads
SOURCED W&J place two nouns as an outer “topic” adjunct and an inner “focus” immediately before the verb (2000:268): (46) ʼashkii łį́į́ʼ yiztał ‘The boy, he kicked [the horse]F’ and (47) ʼashkii łį́į́ʼ biztał ‘The boy, he was kicked by [the horse]F’.
HYPOTHESIS On the spine, these two positions need no silent Top⁰ or Foc⁰. In the direct form the outer noun is the subject in Spec,TP and the inner noun the object in VP. In the inverse, H3 has moved the patient above the raised agent, so the outer noun is the patient and the inner noun the agent, still in Spec,TP below it. W&J’s topic is the highest argument after movement and their focus is the lowest pronounced one. Whether topic and focus are also interpreted at LF (as more than the order of arguments) is OPEN; no source here tests it.
W&J’s examples on the spine
Each row assumes the working default (Merge V, v, T; V→v→T; agent Spec,vP→Spec,TP). Only what else moves is listed. The older derivations with Top/Foc and CYC are L1–L8 under History.
L4. W&J (49) yiztał=go nábidííłtsʼin ‘When hei kicked himj, hei was hit by himj.’
Each clause on its own spine; the second is inverse (H3). The =go CP is adjoined to the second clause. Willie prints the same string with the other indexing (9a), so how the silent arguments are linked across the clauses is OPEN. Review 13 found that Willie’s printed subscripts for (9a)/(9b) and her own prose about them disagree (H45b-θ vs H45b-pos); that source conflict stays OPEN (results §11).
L5. W&J (50b) dibé łį́į́ʼ bitah yíghááh ‘The horse went among the sheep.’
The inverse is on the postposition: bi- is the object of -tah, and W&J say dibé is not a constituent of the PP. Where dibé sits and whether anything moves out of the PP is OPEN.
∃e[join(e) ∧ Mover(e,horse) ∧ AMONG(e,sheep)]
L6. W&J (58b) háí-sh Jáan yiztsʼǫs, reading 2 ‘Who was it, that John kissed?’
The object question word moves to Spec,CP over the subject (H14). Reading 1 ‘Who kissed John?’ has nothing else moving. The ban on this reading in the inverse (60b) is OPEN.
Not restated: its transcription is withheld until the page image is rechecked (see §3 under History).
Open
HYPOTHESIS Where the =go clause attaches relative to a matrix noun (old L8, OPEN until 2026-11-14): Willie 1991:134 (4), NX-397, has a main-clause noun subject after the =go clause, so under the working default the =go CP is above the main Spec,TP (H47, review 13). Whether it is adjoined to TP or higher is OPEN.
OPEN The landing site and trigger of the inverse movement (H3).
OPEN Whether the polar particle da’ is C⁰ or in Spec,CP (headedness).
OPEN Whether overt nouns sit in argument positions (as here) or are adjoined and linked to silent arguments (the older view). The data on this page do not decide it.
History: the page before the 2026-11-14 rebuild
Where the verb word, the nouns, the =go clause and topic/focus sit: a vP/TP spine for Willie & Jelinek's S-trees, and “parallel processing” split into grammar and parsing (T27l)
Everything below is the page as it stood before the rebuild, kept unchanged as a record. Where it differs from the sections above, the sections above supersede it.
2026-11-11 audit pass 1. The old bracket put C on the left while calling it clause-final, and the page equated MVS, QualP and a spell-out boundary despite its own later W correction. The bracket is replaced by a conservative vP/TP comparator; the older MVS-D1/CYC rules and L1–L8 are explicitly historical HYPOTHESES / OPEN. None establishes B1, B2 or B3 uniquely. Silent Top/Foc heads and source S-subdivisions are not premises. Platero (178)’s uncertain spelling and editorial gloss are withheld, not silently “repaired”. Current evidence and verdict: movement diagnostics.
2026-11-11 audit pass 2. Wording that used VPISH as the name of a spine, tree, skeleton or set of slots named a structure after the hypothesis. VPISH is a hypothesis name, never a node; the structure is a vP/TP spine with the subject first-merged in left Spec,vP. Only the wording changed.
2026-11-11 audit pass 3. Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference. See the pass 3 report.
2026-11-11 audit pass 4. Interlinear rows whose form and gloss do not line up one-to-one were re-checked against the source page. Each is marked † with a note under its table: SOURCED where the source prints it that way, OPEN where the source cannot confirm it. See the pass 4 report.
2026-11-11 audit pass 5. The formal spine now says outright that V→v→T is GRANT’S CONJECTURE: untested here, neither proven nor refuted. (Superseded by audit pass 7, 2026-11-13: V→v→T is now the working default.) See the pass 5 report.
2026-11-13 audit pass 7. An earlier dated note on this page treats V→v→T and subject raising as Grant’s conjecture or a comparator to test against PF/template rivals. Since 2026-11-13 they are the working default in tensed clauses: VPISH frames where the arguments start so that syntax and semantics line up; it is not a word-order account. PF/template accounts are kept only as labelled contrast, and the open question is what else has to move. The earlier note is kept as written. See the 2026-11-13 note and the pass 7 report.
The framework audit (rows F7, F8, F9) found that three groups of pages rest on Willie 1991's GB machinery and on Willie & Jelinek 2000's S-adjunction trees. They are nominal adjunction (nouns “adjoined to the MVS”), wh-questions (focus) and =go chains (“parallel processing”). Task T27l asks for four things. (1) Put the MVS, the nouns, the =go clause and topic/focus into a modern spine (Hale 2003's Qual > M/S > v > V plus clause-level heads). (2) Re-read W&J 2000 pp. 267–273 for any head names. (3) Separate syntax from parsing in “parallel processing”. (4) Tie the result to the CYC rule on A/B as spell-out (word = cycle D1, clause = cycle D2) and to the four inverse options on the inverse as Agree and EPP. W&J pp. 266–273 were re-read on page images for this page. Platero's 1973 MIT thesis on the Navajo relative clause, which states two parsing principles in so many words, was opened for the first time. Written 2026-10-17, 4:46am run (PT).
Update 2026-10-22, 4:46am (T49).HYPOTHESIS The focus position is right above the MVS, and the MVS can be W alone. In NLA (9)–(11b) … bee dibé síníloh? the focus noun dibé follows an instrumental PP, which is read as merged on the clause side, between Top and Foc (IP-1, IP-2). In W&J (50b) the PP is part of what the verb describes, and both nouns precede it. See can a noun stand between a postposition and its verb?
Update 2026-10-22, 12:46pm (T49h).HYPOTHESIS The MVS-as-D1 wording on this page is superseded by CYC v2 (A/B as spell-out §3): D1 = W, sealed; the MVS is the set of PP sites inside QualP, ordered before W by C1′. The strings derived on this page do not change. A full rewrite of this page's text is task T50g (postpositions by verb).
Ground rules (Grant's). No analysis is assumed to be fully right; each may be partly right. Silence counts against nothing; only contradicting data count. Grant does not speak Navajo. Every Navajo form below is copied from a source opened this run, or from a site page that records it with page and example number. No Navajo form, judgment or index is invented; where a test needs one, it is listed as OPEN. Pre-VPISH and transitional trees (W&J's S/S′/IP trees, Platero's Aspects-style deep structures) are labelled, restated, and never counted as evidence about where nouns attach. Labels: SOURCEDHYPOTHESISGRANT'S CONJECTUREOPEN.
Rule-set letters and notation. The site convention is kept: A = word-internal (the left-headed prefixal verb word; Merge, Agree and Move inside it, and its spell-out, cycle D1), B = clause level (nouns, clause heads, linking, clause spell-out, cycle D2), Σ = role-schema step. Notation: the Syntax · Semantics · Interface bot still has no notation-conventions file. Its BRIEF.md (copied 2026-10-16, 7:22pm) has no notation section. The “Display convention” in its VERB-QUEUE.md covers sense subscripts on multi-sense verb labels, and no such label is used here. So the provisional Minimalist notation stays: copies in angle brackets, Spec,X, ⌢ for “precedes”. ρA is the slot with the most Proto-Agent entailments for its role in the action, and ρX is the other slot (patient or oblique object). [top] and [foc] are discourse features on slots.
Bottom line, in plain words
SOURCEDW&J 2000 name no topic or focus heads. On pp. 266–273 the only labels are S, S′, IP, V′, “V stem”, COMP, “FOCUS Adjunct” and “TOPIC Adjunct”, the tags T and F under the pronouns, the “verb-sentence”, and the “Maximal Verb Complex (Willie 1991)” (p. 270). There is no Top, Foc, C or T head and no vP. Topic and focus are positions, adjoined to S: “The position immediately preceding the verb is the focus position” (p. 268), and the “outer” nominal is the “topic-adjunct” (p. 268). The =go clause is S′ = [IP COMP], a daughter of S beside the main IP ((48), (49), pp. 269–270). So the search for head names that the audit hoped would point to Rizzi-style heads comes back empty. The spine below is ours.
HYPOTHESISHistorical placement hypothesis — not the current verdict. The verb-plus-postpositions unit (the MVS, W&J's “Maximal Verb Complex”) is the top of the word's own spine. That is Hale's QualP, and it is also the first spell-out cycle D1. Mode/Subject is the T head inside it (Hale 2003 p. 10: M = Infl/T). Above it, at clause level (D2), a silent Focus head hosts the noun right before the verb, a silent Topic head above that hosts the outer noun, and a higher focus position hosts a fronted question word. The =go clause is a CP (with =go as its C) adjoined above every matrix noun in the printed data. The CYC rule from A/B as spell-out then gives the observed orders (topic, focus, verb word, clause-final =go) from height alone, with no extra statement. The site's old rule SB3 (“topic noun first, if the topic is the higher adjunct, site assumption”) now follows from W&J p. 268. One new consequence: W&J's (50) puts an inverse bi- on a separate postposition word, and they call the verb-plus-postpositions unit the domain of the voice alternation. That older model placed the first cycle at the MVS. This is not established: B1/B2/B3 remain OPEN, and the later W correction below already retracts the sealed-MVS claim.
HYPOTHESIS“Parallel processing” is two things of grammar plus one thing of parsing. (PP1) Two all-third-person transitive clauses linked by =go must share their two referents. This is grammar, not a parsing habit. It holds in sentences with no nouns at all (Willie (7)–(9); W&J (49)). An overt new noun does not rescue a violation (*(12)), only morphology (a fourth person) or indefiniteness (léiʼ) does. And Platero's own parsing principles are stated over strings of nouns and verbs, so they cannot even apply to these noun-free sentences. (PP2) Which referent goes into which slot is also stated as grammar by W&J: “the focus pronoun of the first clause is coindexed with the focus pronoun of the second” (p. 270). (PP3) Which verb a noun belongs to in a string with two verbs is parsing. Platero 1973 says so himself: his Elementary Parsing Principle (179) is one of the “principles which speakers appear to use in interpreting sentences” (p. 83), part of “linguistic performance” (p. 94). He adds that he takes no position on its “exact status” in the description of Navajo (p. 93).
SOURCEDA new datum for an old open question (T12a). W&J 2000 (49) prints the same string as Willie 1991 (9a), yiztał=go nábidííłtsʼin, with the other set of indices: ‘When hei kicked himj, hei was hit by himj.’ That is reading R2 on the =go chains page: the same referent is the kicker in the first clause and the one hit in the second. What stays constant across the two clauses is the topic, not the role in the action. Willie's printed (9a) indices and her p. 149 wording (“coindexed according to theta role”) give R1. Her prose on p. 136 and in (15a) gives R2, and W&J's indices and prose give R2. The conflict is now between two printings of one sentence, so it still needs a speaker. But R2 is no longer only Willie's prose against her own example.
HYPOTHESISFor Grant's thesis. Every step that links the two clauses works on the verb's pronoun slots and their topic/focus status. In (49) there are no nouns at all. Nothing here contradicts the weak thesis (role-bearing pronoun slots are the verb's arguments; nouns are optional add-ons linked to them). Under STRATA (Grant's conjecture), topic/focus status has to sit on the slots before nouns are chosen (Layer 3). That is consistent with the conjecture, and it adds one thing it did not say: discourse status is a property of slots.
Willie, MaryAnn & Eloise Jelinek 2000. “Navajo as a discourse configurational language.” In Fernald & Platero (eds.), The Athabaskan Languages, pp. 252–287 (as in the Jelinek collection scan, box [local working copy, not in the repository]; the copy under metal-doping-theta-roles/lit/_pdfs/ is byte-identical). Re-read on page images printed pp. 266–273 at 130 dpi, with (48)–(50) and the rotated text block on p. 269 at 300 and 500 dpi. PDF 146 = p. 266, 147 = p. 267, 148 = p. 269, 149 = p. 270, 150 = p. 271, 151 = p. 272, 152 = p. 273. p. 268 is bound out of order at PDF 153. The scan has no text layer. The lower half of p. 269 is printed rotated, and the bottom of tree (48) with its Navajo line is cut off by the page edge, so (48)'s form is not reproduced.
Platero, Paul R. 1973.A Study of the Navajo Relative Clause. MIT M.S. thesis (advisor Kenneth Hale), the basis of Platero 1974, “The Navajo relative clause”, IJAL 40:202–246. Newly opened. MIT DSpace PDF (handle 1721.1/116490), 102 PDF pages; PDF = typed page + 5. The OCR layer was used only to find passages. Prose and the one Navajo line used here, (178), were read on page images (90 dpi pages; (178) at 400 dpi). Read: §1.9 -go complements (pp. 31–34, prose only), and the end of the deletion chapter, pp. 82–94 with footnotes pp. 95–96. The 1974 IJAL printing was not opened, so all page numbers are the thesis's.
Not opened: Platero 1976 (Willie's source for the term “parallel processing”, p. 135); Speas 1990; Hale, Jelinek & Willie 1991; Rizzi's left periphery (pointed to only by the W&J abstract found by search earlier). The term “parallel processing” does not occur in the 1973 thesis (searched).
2. What W&J 2000 pp. 266–273 actually name, and what that notation is
Label on the page
Where
What it labels
Restated as (ours)
S
(43), (44), (46), (47), (48), (49)
The clause; also the node that FOCUS and TOPIC adjuncts and the =go clause attach to
No single node: the clause-level D2 domain (TopP / FocP / CP)
IP, V′, “V stem”
all trees
3ACC as daughter of IP (sister of V′); 3NOM as sister of the V stem inside V′
Hale 2003: ρX in Comp,V, ρA in Spec,v, M/S (= Infl/T) above vP, Qual above M/S
FOCUS Adjunct
(43), (44), (46), (47)
The noun right before the verb; “a focus position, not an object position” (p. 268)
T, F under pronouns; “Focusi-Topicj-V (Direct)”, “Topicj-Focusi-V (Inverse)”
(43)–(49)
Which pronoun is topic and which is focus
Features [top], [foc] on the slots (VOICE-TF, §4)
S′ = [IP COMP]
(48), (49)
The =go clause; COMP is the gloss of =go in (49)
CP with C = =go
“verb-sentence”
pp. 266, 269
The inflected verb as a clause (“the focus pronoun of the verb-sentence”)
QualP = D1
“Maximal Verb Complex (Willie 1991)”
p. 270
“The verb and any preceding post-positional phrases form a noninterruptible constituent … that is the domain of the voice alternation. This complex determines the interpretation of the adjoined nominals”
Optional fronting of a question word with -sh “to the leftmost position”
Move to Spec,FochiP above TopP
HYPOTHESISClassification. These are transitional trees (GB-style S/IP with S-adjunction, 1990s), not a flat S. Two features are worth keeping apart. (a) The agent pronoun 3NOM sits inside V′, so the subject starts inside the verbal projection. That is VPISH-compatible as drawn, though there is no vP. (b) 3ACC sits above 3NOM. That geometry is the same as the first step of Hale 1997's analysis (object raised to Spec,MP in both voices, IA-2 on inverse as Agree §5.2). But W&J give no argument for it on these pages, and it simply mirrors the linear order 3ACC-3NOM-V. Following Grant's rule for older trees, (b) is not counted as evidence that the object prefix c-commands the subject. T29b stays open.
SOURCEDWhat the prose says, in the words that matter here. “Nominals in Navajo third-person transitive sentences are interpreted as follows. A nominal immediately preceding the inflected verb is in the focus position. Whether this nominal is coindexed with the agent or patient pronominal argument depends upon the yi-/bi--alternation.” (p. 266). “A nominal must be coindexed with a pronominal head. By default, an ‘outer’ nominal must be coindexed with the pronoun that is not coindexed with the inner nominal; it is a ‘topic-adjunct’, or reintroduced topic.” (p. 268). “The two pronominal arguments of the first clause (and their adjuncts, if any) must be coindexed with the two pronominal arguments of the main clause … The voice alternation obligatorily operates over the topic/focus structure of both clauses in the sequence.” (p. 269). “In sum: in each voice alternate, the focus pronoun of the first clause is coindexed with the focus pronoun of the second.” (p. 270). “there can be no more than one occurrence of inverse bi- in a clause … These constraints on multiple relatives preclude multiple topic switches.” (p. 270).
3. Data
SOURCEDNew to the site's data (the trees of (43), (44), (47) were looked at only for their notation in the framework audit; (49) and (50) were not used before. W&J's gloss lines are already in capitals and are copied as printed; “3-PP” is their gloss for the postposition, whose meaning ‘among’ comes from their translation).
W&J 2000:267, (43) The Direct (yi-) Construction
łį́į́ʼ
yiztał
horse
3ACC-3NOM-kicked (Direct)
FOCUS
Focus-Topic-V
‘He kicked it, [the horse]F.’
W&J 2000:267, (44) The Inverse (bi-) Construction
łį́į́ʼ
biztał
horse
3ACC-3NOM-kicked (Inverse)
FOCUS
Topic-Focus-V
‘He was kicked by it [the horse]F / [The horse]F kicked him.’
W&J 2000:269, (47) (inverse with two nominals)
ʼashkii
łį́į́ʼ
biztał
boy
horse
3ACC-3NOM-kicked (Inverse)
TOP-ADJ
FOCUS
Topic-Focus V
‘The boy, he was kicked by [the horse]F.’
W&J 2000:270, (49) (the =go sequence with an inverse second clause)
yiztał=go
nábidííłtsʼin
3ACC-3NOM-kicked=COMP,†
3ACC-3NOM-hit (Inverse)
SOURCED † Segmentation note (2026-11-11 audit pass 4): Willie & Jelinek 2000:270 (49) prints yiztał=go with the gloss “3ACC-3NOM-kicked=COMP,”: the gloss names more morphemes than the form shows as separate pieces (fused or unwritten morphemes), so the two lines do not match one-to-one in the source. Kept exactly as printed.
‘When hei kicked himj, hei was hit by himj.’
Tree (49): first IP: 3ACCj F, 3NOMi T, V stem “-kick”; second IP: 3ACCi T, 3NOMj F, “V stem (Inverse) -hit”. So in the first clause the agent i is topic and the patient j is focus. In the second, the patient i is topic and the agent j is focus. Same string as Willie 1991:135–136 (9a), which she prints with ‘As hei kicked himj, hej was hit by himi.’ The barred ł of the second verb is faint even at 500 dpi. It is read as ł, matching Willie's printing.
W&J 2000:270, (50a, b)
a.
dibé
łį́į́ʼ
yitah
yíghááh
sheep
horse
3-PP
3-join (Direct)
‘The sheep went among the horses.’
b.
dibé
łį́į́ʼ
bitah
yíghááh
sheep
horse
3-PP
3-join (Inverse)
‘The horse went among the sheep.’ (As for the sheep, the horse, among them it joined).
W&J on (50b): “The topical oblique object pronoun bi- in the postpositional phrase in (50b) marks the phrase as disjoint in reference with the preceding focus nominal. The leftmost nominal dibé does not form a syntactic constituent with the postpositional phrase, although it is coindexed with the bi-postpositional object.” (p. 270). The verb word is the same in (a) and (b). Only the postposition's object prefix alternates.
OPEN — transcription withheld. Platero 1973 p.87 (178), with interpretation at p.93 (185), was previously printed here with uncertain tone marks normalized to the site’s usual spelling. Its Leipzig line was editorial, not the printed source gloss. The exact form/gloss is withheld from current evidence until the page image is rechecked; L7 below is only a historical semantic test, not a verified current derivation.
SOURCEDAlready on the site, used here: W&J (46) ʼashkii łį́į́ʼ yiztał ‘The boy, he kicked [the horse]F’ (p. 268); W&J (57)–(60) (pp. 272–273; wh-questions); Willie (7) yiztał=go néidííłtsʼin, (8) biztał=go nábiidííłtsʼin, (9a, b), *(10) yiztał=go náhodííłtsʼin, *(12) yiztał=go ʼatʼééd néidííłtsʼin, (13) jizloh=go ʼáádóó yiztał, *(14) hazloh=go ʼáádóó yiztał, (16b) yiztał=go ʼáádóó ʼatʼééd léiʼ néidííłtsin, and the intransitive (4) yah ʼííyáa=go ʼatʼééd dah neezdá (Willie 1991:133–140; glosses and translations on =go chains §3).
4. The spine (formal statement)
CURRENT CORE — WORKING DEFAULT, not a source facsimile
[CP [C′ [TP [DP a_i] [T′ T⁰ [vP [DP a_i (lower copy)] [v′ v⁰ [VP [V′ [DP t] V⁰]]]]]] C⁰]]
Specifiers left; C⁰ and V⁰ final, T⁰ and v⁰ initial in THIS hypothesis.
Merge VP, v/subject, then T; V→v→T; subject Spec,vP→Spec,TP; add C above TP.
V→v→T and subject raising are the working default in tensed clauses (2026-11-13).
With T⁰ initial here, a verb-final clause needs the remnant VP above T (HYPOTHESIS); a final T⁰ is the OPEN alternative.
LF: ∃e[PRED(e) ∧ Agent(e,a) ∧ Theme(e,t)] (generic affirmative transitive).
Top/Foc sites, =go’s category, M/S≈T, Qual and boundaries B1/B2/B3 remain OPEN.
The earlier bracket placed C first but called it final. This corrected comparator chooses final C,
without claiming the order uniquely proves that choice. Nonpronounced heads are not evidence.
HISTORICAL MODEL — NOT THE CURRENT BOUNDARY VERDICT
MVS-D1 The first spell-out cycle D1 is QualP together with the VP-internal PPs, i.e. Willie's MVS
(W&J p. 270: "noninterruptible", "the domain of the voice alternation").
The prosodic word can be smaller than D1 (bitah | yíghááh). (Hale n. 6: "verbal phonological
phrase" may be the better term; n. 12: where the word stops is open.) HYPOTHESIS
VOICE-TF In a transitive clause with two 3rd-person slots (Σ step):
direct: ρ_A [top], ρ_X [foc]; inverse: ρ_X [top], ρ_A [foc].
(W&J's tags under (43), (44), (46), (47), (49).) A slot with π ≠ 3 bears neither feature
(W&J p. 267: no direct/inverse alternation with a non-third-person argument; 4-NO-INV).
SOURCED (tags) / HYPOTHESIS (features)
FOC-LINK A noun in Spec,FocP is linked to the [foc] slot; a noun in Spec,TopP to the [top] slot.
(W&J pp. 266, 268: "A nominal must be coindexed with a pronominal head"; focus position =
immediately before the verb; "outer" nominal = topic-adjunct.) SOURCED as order / HYPOTHESIS as heads
HEIGHT Top > Foc. (W&J p. 268: the outer nominal is the topic.) SOURCED as order
FRONT A question word with -sh may move from Spec,FocP to Spec,Foc_hiP, above TopP
(W&J p. 272 (58b) reading 2: "optional focus movement of the Wh-word + Q to the leftmost position").
HYPOTHESIS (landing)
GO-ATT The =go clause C1 is a CP adjoined in the matrix D2 above every matrix noun in the printed data
(Willie (4), *(12), (16b): matrix nouns follow C1). Whether it adjoins to FocP or to TopP: OPEN (T31a).
FOC-ID In [C1=go C2], both transitive and all slots π = 3: idx([foc] in C1) = idx([foc] in C2).
(W&J p. 270.) With two slots per clause and GO-NO-NEW, the topics then match too.
SOURCED as W&J's statement / HYPOTHESIS as rule
CYC As on A/B as spell-out §3: in each domain, non-heads before heads (C1), non-heads higher-first (C2),
heads higher-first (C3). D1 = MVS, D2 = clause; an already spelled domain = one non-head.
Update 2026-10-21, 4:46am (T43e).HYPOTHESIS MVS-D1 is to be read as the first ordering domain (QualP with its VP-internal PPs), not as one sealed block. The sealed units are the verb word W and each PP (PC-1 on attached or separate postpositions); the yes/no ending can sit between them (NLA (6) Neésh hólǫ́?). No derivation on this page changes its string. “Cannot be interrupted” (W&J p. 270) is derived from height: nouns and particles are merged above the MVS (PC-2).
HYPOTHESISTP and CP. In this spine “TP” is M/SP. Hale 2003 identifies M with Infl/T (p. 10), and it is the head that agrees with the subject slot (the portmanteau). It is inside the word, below Qual. “CP” is the clause-level layer: the spelled C heads (=go, ya’, the relativizer) and the silent Top/Foc heads below them. Only within this historical hypothesis is the A/B transition placed between QualP and the first clause head; the data do not uniquely determine it. Everything Hale puts in the word (Qual, M/S = T, v, V, the pronoun slots, the preverbs) is A. Everything W&J put in S-adjunctions (topic, focus, =go clause) is B.
ρX Comp,V; ρA Spec,v (Hale p. 28); under IA-1* the inverse ρX raises to Spec,M/SP
A
SOURCED (Hale) / restated
Focus noun (right before the verb)
FOCUS Adjunct to S
Spec,FocP
B
order SOURCED, head HYPOTHESIS
Topic noun (outer)
TOPIC Adjunct to S
Spec,TopP
B
order SOURCED, head HYPOTHESIS
Fronted question word + -sh
“focus (leftward) movement … to the leftmost position”
Spec,FochiP
B
HYPOTHESIS
=go
COMP (W&J (49)); “complementizer enclitic” (Willie p. 132); Platero's -go complements also carry COMP ((73), p. 33)
C head of C1
B
category SOURCED as their label
=go clause
S′ daughter of S (W&J (48), (49)); “adjoined” (Willie pp. 132–140)
CP adjoined in matrix D2 above the matrix nouns; FocP or TopP OPEN
B
HYPOTHESIS
Nouns linked to a PP object
“does not form a syntactic constituent with the postpositional phrase” (W&J p. 270)
Spec,TopP / Spec,FocP, linked to the PP's [top] or [foc] slot
B
HYPOTHESIS
5. “Parallel processing”: what is grammar and what is parsing
5.1 How to tell them apart (stated before looking at the data)
GRAMMAR a condition on structures or on the referents of slots: it can apply where the string has no
nouns; it produces stars that lexical content cannot cancel; it is sensitive to morphology.
PARSING a procedure over surface strings (Platero: "In a subsequence of the form (NP) NP V …", (179));
it needs nouns in the string; it can be overruled by the rest of the sentence; speakers vary.
(criteria ours, HYPOTHESIS; Platero's own split, pp. 82–94)
SOURCEDPlatero's split, in his words. His deletion rule “will then operate to produce both the derivations we have identified as well-formed and those we have identified as ill-formed. Under this proposal, the rules of the grammar will not themselves distinguish the two types of derivations” (pp. 82–83). The difference comes from “certain principles which speakers appear to use in interpreting sentences” (p. 83). These are the Grammatical Relation Principle (172), p. 83: “In a string of the form (NP) NP V containing only third person NPs, the NP immediately preceding the verb is interpreted as (a) the object if the verb-word contains the object prefix /yi-/, (b) the subject if the verb-word contains the object prefix /bi-/”. And the Elementary Parsing Principle (179), p. 88: “In a subsequence of the form (NP) NP V, (NP) NP is clausemate with V if the subsequence is a well-formed simple sentence of Navajo.” (179) “can be overruled” where “there are not enough NPs to go around” (p. 92), but not in (178)-type sentences (pp. 92–93). “We do not take a position on the issue of the exact status of the surface interpretation principles in the description of Navajo” (p. 93). The results “highlight the importance in linguistic performance of perceptual strategy of the type studied by Fodor, Bever, and others” (p. 94).
5.2 Three things under one name
Part
Statement
Evidence it is grammar or parsing
Verdict
PP1 cast identity
GO-PP and GO-NO-NEW (=go chains §4): two all-3rd transitive clauses share their two referents
Holds in noun-free strings ((7)–(9), W&J (49)), where Platero's string procedures have nothing to apply to. *(12): an overt new noun ʼatʼééd does not rescue it, which a parsing default would allow. A fourth person (morphology, (13)) or léiʼ (indefiniteness, (16b)) does license a new referent
Grammar (an interface condition on slots) HYPOTHESIS
PP2 mapping
Which referent goes into which slot in C2. R1: by role in the action (Willie (9a) indices, p. 149). R2 = FOC-ID: by focus (W&J p. 270, (49) indices; Willie p. 136, (15a))
W&J state it as obligatory (“must”, “obligatorily”, p. 269); noun-free
Grammar; R1 vs R2 OPEN (T12a = T31b)
PP3 noun-to-verb assignment
In a string with two verbs, which nouns go with which verb (Platero (179); H&P 1974 (44))
Defined over NP–V subsequences; overrulable (p. 92); speaker variation (H&P (40)–(43))
Parsing (SOURCED as Platero's classification)
(and) noun-to-slot inside one clause
Platero's GRP (172) = W&J's focus rule
Follows in this spine from FOC-LINK + VOICE-TF: the preverbal noun is in Spec,FocP, and the [foc] slot is ρX under yi-, ρA under bi- (L1, L2)
Derivable as grammar here. Platero called it an interpretive strategy and left its status open (p. 93), so this is not a contradiction of him
5.3 What FOC-ID derives, and what it does not
HYPOTHESIS FOC-ID plus VOICE-TF plus one default, “a lone 3rd-person slot next to a fourth person counts as [top] if it is the agent and [foc] if it is the patient”. The default follows W&J p. 266, “a thematic hierarchy whereby agents are topical in transitive sentences”, but it is our extension, not their statement (T31c).
(7) both direct: [foc] = patient j in both ✓. (8) both inverse: [foc] = agent j in both ✓ (Willie's indices agree).
W&J (49): [foc] = j in C1 (patient) and in C2 (agent) ✓, with the R2 indices as printed. Willie's (9a) indices (R1) ✗. Her (9b) “(back)” ✓, her printed (9b) indices ✗.
*(12) ✓: C2's two slots both carry [top]/[foc], both are fixed by C1, and nothing is left for a new girl.
*(10) yiztał=go náhodííłtsʼin ✓: C2's fourth-person slot has no [foc]/[top], so C1's focus has no partner in C2.
(13) jizloh=go ʼáádóó yiztał ✓ (with the default): C1's only marked slot is the patient j, [foc]; C2's [foc] is the direct patient, j. C2's [top] has no antecedent, so a new agent k is allowed.
*(14) hazloh=go ʼáádóó yiztał ✓ (with the default): C1's only marked slot is the agent, [top], so C1 has no [foc]. C2's [foc] (the patient) has no antecedent, so FOC-ID cannot be met.
(16b) … ʼatʼééd léiʼ néidííłtsin ✗ without an added clause: a new indefinite focus is allowed. W&J p. 273 say focus can be “new information”, so FOC-ID must be limited to definite foci. R1 needs the same clause.
HYPOTHESISConsequence for GO-PAT. The =go chains page left open whether GO-PAT (patients corefer) is stated over a role or over the object position (T12b). FOC-ID gives a third possibility: over focus. R1 derives (13)/*(14) with no default. FOC-ID derives them only with the agent-topic default. So the two remain tied on (13)/(14), and the indices of (9a) = (49) are the only place where they differ.
HYPOTHESISTwo ways to implement FOC-ID, both grammar. (i) Binding: one silent focus operator at the matrix edge binds the [foc] slot of both clauses. This is Speas's ATB constraint (p. 264; subject-to-subject, object-to-object parallelism, as recorded on Speas 1986) restated as binding, not movement, so the adjunct island does not arise. It needs C1 inside the operator's c-command domain and predicts that a matrix topic noun can precede C1 (T31a). (ii) Discourse anaphora on adjoined clauses: the focus of an adjoined C1 is resumed by its host. This needs no c-command, and it fits the construction split already on the site: GO-PP holds in =go clauses and relatives (Willie (27), p. 166) and fails under ‘say’ (reported speech (22)). Neither implementation is a parsing strategy.
6. Derivations L1–L8, step by step
Format as on the inverse as Agree. Each step is Merge, Agree, Move, Link or Spell-out, tagged Σ, A (D1 = MVS) or B (D2). The older IA-3 comparison is retained as a historical contrast, with IA-1* differences noted; it is not the current primary V→v→T derivation. Full L1–L8 reconstruction on the primary skeleton is OPEN. They never change the string. Role-in-the-action entailments for ‘kick’ are the D3 sets (dictionary); those for ‘hit’, ‘join’, ‘bite’, ‘rope’ are ours, from the translations (HYPOTHESIS). Verb words are not segmented beyond the source's gloss line.
Σ Σ(√KICK) = {ρA (volition, causation, movement), ρP (contact, affected)}, both π = 3. No rank difference is given, so the voice is free (RANK→EPP, third line); the speaker picks the direct. VOICE-TF: ρA [top], ρP [foc].
AMerge V with ρP (Comp,V). Merge v; Merge ρA in Spec,v. Merge M/S; Agree M/S ⟶ ρA [3]. Merge Qual (W&J give no segmentation; no Qual exponent is claimed).
ASpell-out D1 (= MVS). OBV-top: both slots 3, ρP not [top], so y-. CYC C1: the pronoun before the heads. → yiztał.
BMerge Foc (silent) with D1. Mergełį́į́ʼ in Spec,FocP. Link (FOC-LINK): łį́į́ʼ ↔ the [foc] slot = ρP. W&J: the patient pronoun “has focus … and serves as an anaphor to the nominal” (p. 266).
BSpell-out D2: N = {łį́į́ʼ, [D1]}, H = {Foc}. C2: łį́į́ʼ ⌢ [yiztał]. Result: łį́į́ʼ yiztał ✓, horse = the one kicked ✓.
Moves: 0 (IA-3 and IA-1*, since IA-1* raises nothing in the direct).
BSpell-out D2: łį́į́ʼ biztał ✓, horse = kicker ✓ (W&J p. 266: “it cannot be coindexed with the explicitly topicalized pronoun bi-”).
HYPOTHESIS L1 and L2 together are Platero's GRP (172) as a theorem: the preverbal noun is the object under yi- and the subject under bi-, with no parsing step.
BMerge Foc; łį́į́ʼ in Spec,FocP; Link to [foc] (46: ρP; 47: ρA).
BMerge Top (silent) with FocP; ʼashkii in Spec,TopP; Link to [top] (46: ρA ‘The boy, he kicked the horse’; 47: ρP ‘The boy, he was kicked by the horse’).
BSpell-out D2: N = {ʼashkii, łį́į́ʼ, [D1]}, H = {Top, Foc}. C2 (higher first): ʼashkii ⌢ łį́į́ʼ ⌢ [D1] ✓ for both.
HYPOTHESIS The old SB3 (“if the topic is the higher adjunct, site assumption”) is now HEIGHT + CYC C2, with the height taken from W&J p. 268.
BMerge C = =go (COMP) over C1's D1 (no nouns, so no Spec,TopP/FocP is filled). Spell-out D2(C1): N = {[yiztał]}, H = {=go}. CYC C1 (non-heads before heads): yiztał=go. C1 is now one non-head.
A C2: Merge, Agree as L2 (IA-1*: ρP raises). Spell-out D1: OBV-top gives b- → nábidííłtsʼin (not segmented further).
BMerge Foc, Top in C2 (silent, empty specifiers). Merge C1 by adjunction (GO-ATT; node OPEN).
B FOC-ID: idx([foc] C2) = idx([foc] C1) = j. GO-NO-NEW: C2's other slot = i. So C2's ρA = j and ρP = i: ‘hei was hit by himj’, as W&J print it. (Under R1, the role-matching alternative, C2's ρA = i and ρP = j, Willie's (9a) indices. Same string.)
BSpell-out D2: N = {[C1], [D1 of C2]}, H = {Foc, Top}. C2: C1 is higher → yiztał=go ⌢ nábidííłtsʼin ✓.
HYPOTHESIS No step mentions a noun. The cross-clause relation is computed entirely over slots and their discourse features.
L5. W&J (50b) dibé łį́į́ʼ bitah yíghááh (the inverse on a postposition)
Σ Σ(√JOIN, with P ‘among’) = {ρA (the one who goes: movement, volition), ρX = the PP object (location; no Proto-Agent entailment)}. Both π = 3. Inverse: ρX [top] (‘As for the sheep …’), ρA [foc].
AMerge P (-tah) with ρX. PP sub-domain: OBV-top on the P object: [top], so b-. CYC (pronoun ⌢ P) → bitah.
AMerge V with the PP inside VP. Merge v, ρA in Spec,v; Merge M/S; Agree M/S ⟶ ρA [3]; Qual.
ASpell-out D1 (= MVS): N = {[bitah]}, H = verbal heads. C1: bitah ⌢ yíghááh. The PP is phonologically a separate word inside D1 (MVS-D1).
BMerge Foc; łį́į́ʼ in Spec,FocP; Link to [foc] = ρA (the horse goes). Merge Top; dibé in Spec,TopP; Link to [top] = ρX (W&J: dibé is coindexed with the bi- object but not a constituent of the PP).
BSpell-out D2: dibé ⌢ łį́į́ʼ ⌢ [bitah yíghááh] ✓.
HYPOTHESISWhat (50b) does to the inverse options. Under IA-3 nothing moves and the [top] feature is on the P object from the start, so b- is spelled in place (step 2), with no extra assumption. Under IA-1* the topic slot should raise to Spec,M/SP. But bi- is pronounced on the postposition. So either the lower copy is pronounced, or the PP is not spelled until after v's EPP has applied, and CYC does not yet fix that timing. This is the same escape the da- trilemma needed in its low-Pl branch (inverse as Agree K5). IA-0 (inverse linking) first-merges ρX in Spec,v, so it cannot put bi- on P unless bi- is agreement (e3). It is a derived tension, not a falsification (T31e).
L6. W&J (58b) háí-sh Jáan yiztsʼǫs, reading 2 ‘Who was it, that John kissed?’ (fronting)
ΣA Direct: ρA [top], ρP [foc]. D1 = yiztsʼǫs.
BMerge Foc; háí-sh in Spec,FocP, linked to [foc] = ρP (the one kissed). Merge Top; Jáan in Spec,TopP, linked to [top] = ρA.
BMerge Fochi; Moveháí-sh → Spec,FochiP (FRONT; W&J: “optional focus movement of the Wh-word + Q to the leftmost position”, p. 272). The copy in Spec,FocP keeps the link.
BSpell-out D2: C2: háí-sh ⌢ Jáan ⌢ [yiztsʼǫs] ✓ (lower copy silent). Reading 1 (‘Who kissed John?’) is the unmoved parse: háí-sh in Spec,TopP (agent) and Jáan in Spec,FocP (patient). Same string.
OPEN In the inverse (60b) the fronted reading is out: “The bi-pronoun marks a topic/focus inversion that is incompatible with focus movement of the Wh-word” (p. 273). The spine says where the moved word lands, but nothing in it blocks the move in the inverse. Willie p. 208 also bans the parallel direct reading with two humans, which conflicts with (58b) (T19a). So this is left as T31h.
L7. Platero (178) Ashkii łééchąąʼí bishxash-ę́ę łį́į́ʼ yizloh: two legal structures, one parse
BStructure P (the parse Platero allows): the relative is a domain with head -ę́ę (Willie pp. 153, 158). Inside it, łééchąąʼí is in Spec,FocP, linked to [foc] = ρA (the dog bites), and ashkii is in Spec,TopP, linked to [top] = ρP. The relative is in the main Spec,TopP, linked to the main ρA, and łį́į́ʼ is in the main Spec,FocP, linked to ρP. Reading: ‘The boy that the dog bit roped the horse.’
BStructure Q (built by the same Merge steps, excluded by Platero):ashkii is in the main Spec,TopP. The relative contains only łééchąąʼí and has a different head (Platero's (180)).
BSpell-out D2 by CYC: P and Q both come out as Ashkii łééchąąʼí bishxash-ę́ę łį́į́ʼ yizloh.
Parsing (not a syntax step): Platero's (179) brackets the longest well-formed (NP) NP V subsequence, ashkii łééchąąʼí bishxash, as one clause. So only P is understood (p. 88; “cannot be overruled”, pp. 92–93).
HYPOTHESIS In this spine the syntax builds both structures and spells them identically, as Platero's grammar did (pp. 82–83). The choice is made by a string procedure that needs nouns. This is the parsing half of the split, and it is a different object from PP1/PP2 in L4.
L8. Schematic: where a matrix topic noun would go relative to the =go clause (no data)
B C1 adjoined to matrix FocP (below Spec,TopP): CYC gives Ntop ⌢ [C1=go] ⌢ Nfoc ⌢ [D1].
B C1 adjoined to matrix TopP (above Spec,TopP): CYC gives [C1=go] ⌢ Ntop ⌢ Nfoc ⌢ [D1].
OPEN Letters only. The opened pages print no sentence with a matrix topic noun and a =go clause together. Such a sentence would decide GO-ATT, and it would test binding implementation (i) of FOC-ID, which needs the first order (T31a).
7. Ties to CYC and to the inverse options
HYPOTHESISCYC (A/B as spell-out §3). Historical CYC comparison only, not the current boundary verdict. Two assumptions were added. (a) Where D1 stops. The spell-out-rules page left the word boundary open (Hale n. 12). W&J's “Maximal Verb Complex … domain of the voice alternation” and (50b) place it at the MVS: VP-internal PPs are inside D1 as already spelled non-heads. That revises SB1 there (“everything outside W: nouns, PPs …”): grammatical PPs are D1 non-heads, not D2. This is also the site's boundary candidate B3 (headedness §3), and the nominal adjunction page already found three properties converging on B3. (b) The clause heads. D2 has silent Top/Foc heads. They are heads, so by C1 they would be pronounced last, but they have no exponent, so the string is unaffected. Their specifiers are ordered by C2. CYC's prediction that da’ and doo are phrases (T29a) now has a natural home: Spec,FochiP or Spec,CP.
Inverse option
(49) FOC-ID / topic matching
(50b) bi- on the postposition
IA-0 inverse linking
Configurational: the topic is the highest slot in both clauses
✗ unless bi- is agreement (e3)
IA-1* raising, inverse only
Configurational: the topic is in Spec,M/SP (inverse) or Spec,v (direct), the highest slot either way. Speas's subject-to-subject parallelism (p. 264) and W&J's topic-to-topic then coincide
Tension: needs lower-copy pronunciation or late PP spell-out
IA-2 Hale 1997
Configurational: topic in Spec,QualP/E in both voices
Same tension, plus freezing (PDF p. 21) OPEN
IA-3 in situ (Hale 2003)
Featural: [top]/[foc] identity. Fine, since FOC-ID is stated over features
✓ with no extra assumption
HYPOTHESIS Neither datum eliminates an option. (49) is neutral, since it can be stated either way. (50b) is a small point for IA-3, which joins the da- trilemma (also solved by IA-3 with nothing extra) and leaves T30d (scope of (58)) as the real decider.
8. Results
Datum
R1 (match by role)
R2 = FOC-ID
Parsing default (“assume same cast”)
(7), (8)
✓
✓
✓
(9a) = (49): Willie's indices
✓
✗
says nothing about mapping
(49): W&J's indices
✗
✓
says nothing
*(10) 4th person in C2
✓ (patient π mismatch)
✓
✗ (π = 4 should override a default)
*(12) new noun in C2
✓ (cast)
✓
✗ (an explicit noun should override a default)
(13)
✓
✓ with default
✓
*(14)
✓
✓ with default
✗
(16b) léiʼ
needs clause
needs clause
✓
Willie's prose p. 136, (15a) “switch in thematic roles”
✗
✓
neutral
Willie p. 149 “coindexed according to theta role”
✓
✗
neutral
HYPOTHESIS The parsing default fails three starred sentences, so PP1 is not a parsing default. R1 and R2 each fit one printing of (9a) = (49) and one half of Willie's wording. Decider: a speaker's reading of yiztał=go nábidííłtsʼin.
9. What each source gets right, and what it misses (on this test only)
Source
Gets right
Misses or leaves open
Willie & Jelinek 2000
Topic and focus as two ordered positions, focus right before the verb (p. 268). Voice decides which slot is focus (p. 266). Cross-clause identity stated over topic/focus and over pronouns, not nouns (pp. 269–270), with explicit indices (49). The MVS as the voice domain, with the oblique inverse (50)
S-adjunction with no heads, so there is no account of why focus is preverbal. The ACC-over-NOM geometry is drawn without argument. (49) contradicts Willie's (9a) indices without comment. A new indefinite focus (Willie (16b)) is not discussed on these pages
Willie 1991
The construction facts and stars (7)–(16). Rescue by fourth person and by léiʼ. The construction split (=go and relatives vs ‘say’)
“Parallel processing” names two different things (PP1, PP2). Her indices (R1) and her prose (R2) disagree
Platero 1973
A clean grammar/performance split: the grammar over-generates, and string-based principles select (pp. 82–94). Explicit overruling conditions (p. 92). Honest about status (p. 93)
Aspects-style deep structures with NPs in S (pre-VPISH; not used as evidence about attachment). His GRP (172) can be derived as grammar (L1–L2), so calling it a strategy is not forced
Speas 1986
An ATB-type parallelism (p. 264) is the right shape for PP2 if “subject” means the highest slot (IA-1*, IA-2)
ATB as movement out of an adjunct; the subject-to-subject wording loses W&J's topic-to-topic where the two come apart (under IA-3)
Hale 2003
The D1 spine; M = Infl/T; nouns “adjoined to the full verb word (or verbal clause)” (p. 27), compatible with Spec,FocP/TopP right above QualP
Silent on topic/focus heads and on clause chaining. Counts against nothing
10. Grant's two phenomena, Dowty, and STRATA
Phenomenon 1 (right-headed clause, left-headed word).HYPOTHESIS The older model proposed a precise spine position, not an established changeover: between QualP (the top of D1 = MVS) and the first clause-level head (Foc). Below it, Hale's heads Qual > M/S(=T) > v > V are pronounced head-first (CYC C3 in D1). Above it, the only spelled heads are C-type particles, pronounced last (CYC C1 in D2), and the silent Top/Foc heads order their nouns by height. GRANT'S CONJECTURE the older model restated “the word is the changeover point” with “word” read as the MVS (verb plus grammatical postpositions). The prosodic word can be smaller (bitah | yíghááh).
Phenomenon 2 (roles, not words, as arguments).HYPOTHESIS Every grammatical step in L4 refers to slots and their [top]/[foc] features. The only step that refers to nouns as nouns (L7 step 5) is parsing. W&J's own wording is pronoun-based: “A nominal must be coindexed with a pronominal head” (p. 268), and “the two pronominal arguments of the first clause … must be coindexed with the two pronominal arguments of the main clause” (p. 269). Nothing here contradicts the weak thesis. The strong thesis already failed elsewhere and is not retested.
Dowty baseline. Grant's stance, followed here: Dowty's proto-roles are the mainstream baseline, and the Navajo topic/rank layer extends it. In (49) on W&J's reading, the same referent has Proto-Agent entailments for its role in the action in C1 and Proto-Patient ones in C2. What carries over is its topic status. That fits the site's standing view that proto-role entailments are fixed by each verb's action and that discourse only shifts which slot is topic (and so which one is spelled bi-). Event-structure roles (Rappaport Hovav and Levin style) play no part in this test. They are compatible with every row.
STRATA as the third option.GRANT'S CONJECTURE tested beside the two older views:
View
Restated in this spine
What this run's data say
Adjunct (Willie, W&J)
Nouns first-merged in Spec,TopP/FocP (D2), linked to slots
Fits (43)–(50) directly
Inside the clause (Speas)
Nouns first-merged in Spec,v/Comp,V inside D1, moved to Spec,FocP/TopP
Fits the orders with one move per noun. (50b) needs dibé extracted from the PP (W&J assert non-constituency, which is an analysis, not a datum). Nothing contradicts it outright
(49) and Willie (7)–(9) show discourse status working on slots with no nouns. That is consistent, and it adds “discourse status before Layer 3”. Layer 3 = D2 comes after the subject bundle's last Merge, so STRATA here inherits the Condition C ordering question (Condition C by site §6.4, T28a)
Follow-up (2026-10-19, 1:52pm). Which word-internal facts are Merge sites and which are Agree, what counts as a phase inside the word (D1 and the PP), and whether any template position is [+MFS]-like: Merge and Agree inside the word (T40a–j).
11. Tasks T31a–h
ID
Question
What would settle it
T31a
OPEN Is the =go clause adjoined to matrix FocP or TopP? Is there a sentence with a matrix topic noun and a =go clause (L8)?
OPEN = T12a: R1 or R2 for yiztał=go nábidííłtsʼin? Willie (9a) and W&J (49) print different indices
A speaker; Platero 1976; Hale, Jelinek & Willie 1991
T31c
OPEN In a clause with a fourth person, is the remaining 3rd-person slot topic or focus? (FOC-ID's account of (13)/*(14) depends on it)
Willie Ch. 4–5 on the fourth person and topic; speaker
T31d
OPEN Is the link from Spec,FocP to a [foc] slot deep inside D1 (direct ρP in Comp,V) syntactic Agree, which phase impenetrability would block, or an interpretive link? Under IA-1* only the inverse topic reaches the edge
Theory-internal; Bruening 2005 on binding into the inverse
T31e
OPEN (50b): under IA-1*, how does the P object's bi- get spelled on P (lower copy, or late PP spell-out)? Joins T30c
More oblique-inverse data (Willie Ch. 2–3 PPs); scope of bi--PP objects. Partly worked on postposition units PU4: the inverse b- is on P in four pairs (T32h)
T31f
OPEN Do lexical PPs (Willie p. 39, LPP) sit in D1 (MVS) or D2? Is a noun ever printed between a lexical PP and the verb?
Willie pp. 39–70 re-checked for LPP strings. Partly answered on postposition units: Willie p. 40 makes only grammatical postpositions members of the MVS (T32i). Update 12:46pm: Willie (92)/(93) put a noun between a grammatical PP (bił/yił) and the verb, against p. 40 + (44a) as stated; options M1–M4 on one-place verb + P §8 (T33d)
T31g
OPEN Open Platero 1976 (Willie's source for “parallel processing”) and the 1974 IJAL printing of the thesis; check whether PP1/PP2 are stated there as grammar or as processing
Library copy
T31h
OPEN Why does the inverse block wh fronting ((60b), p. 273)? The spine locates the landing site but does not derive the block; joins T19a
Speaker judgments on (58b)/(60b)-type pairs
Changes made elsewhere for this page. The framework audit T27l row is marked done, with notes on rows F7–F9. The =go chains T12a row and §5.2 carry the W&J (49) datum. A/B as spell-out §3 carries a note on MVS-D1 (D1 = the MVS; SB1 revised for grammatical PPs). Inverse as Agree carries a note on (50b). See also open questions (T31a–h) and sources.
Follow-up (2026-10-17, 7:42am). What a postposition is (P, functional, T or C?), the verb + postposition unit and the inverse bi- on P are worked on the new page postposition units (T32a–j).
Sources opened or re-checked for this page (2026-10-17, 4:46am PT). W&J 2000 printed pp. 266–273 on page images (PDF 146–153 of the Jelinek collection scan; p. 268 at PDF 153). Platero 1973 MIT thesis (DSpace 1721.1/116490), pp. 31–34, 82–96, newly opened. Willie 1991, Hale 2003, Speas 1986 and Hale & Platero 1974 as recorded on the site with page and example numbers.
Restated on the unified structure (2026-11-15). Nouns are restated as arguments merged in their theta positions (VP, Spec,vP) and, for the subject, raised to Spec,TP by the working default; a lone noun is linked by the H4″ default. Adjunction and pronominal-argument analyses below are labelled contrast; where nouns sit for Condition C is to be re-run on the unified spine. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). Flat-S and other pre-VPISH notation below is labelled contrast. Derivations below in which the verb or subject does not raise are labelled contrast; the working default raises V→v→T and the subject to Spec,TP. Status on the spine: see unified structure, §2 and §4, first rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.
2026-11-11 audit pass 2. Wording that used VPISH as the name of a spine, tree, skeleton or set of slots named a structure after the hypothesis. VPISH is a hypothesis name, never a node; the structure is a vP/TP spine with the subject first-merged in left Spec,vP. Only the wording changed.
2026-11-11 audit pass 3. Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference. See the pass 3 report.
The framework audit placed Willie's 1991 dissertation in class (b), “transitional”, from a few pages and a noisy OCR layer. Task T27o asked for a proper reading of her Chapters 1–2 on page images, to look for trees and for an explicit statement of her framework. This run read every page of the abstract, Chapter 1 and Chapter 2 (printed pp. 10–51) on page images, plus her own summary of the framework in Chapter 9 (pp. 220–221) and two Chapter 3 pages that Chapter 1 points to (pp. 92–93). It then restates her framework rule by rule in the site's spine, and tests the restated rules on the new Chapter 2 forms. Written 2026-10-18, 4:46am run (PT).
Ground rules (Grant's). Willie, Jelinek and Hale are each treated as possibly partly right. A framework statement is not data. Only contradicting data count against a hypothesis, and silence counts against nothing. Grant does not speak Navajo. Every Navajo form below is copied from a Willie 1991 page image read this run (300 dpi crops), with page and example number. No form, gloss or judgment is invented, and tests that need sentences nobody printed are given in letters only. Dowty's proto-roles are the baseline, and “role in the action” names a participant's Dowty entailments. Every rule name coined here is ours and a HYPOTHESIS.
Notation. This page uses the site's convention, since no conventions file has arrived from the Syntax · Semantics · Interface bot. Σ = role-schema step. A = word level (the verb word plus its grammatical postpositions, spelled out as cycle D1 = the MVS). B = clause level (nouns, clauses, enclitics; cycle D2 and up). The spine is the one on the clause spine §4: V < v < M/S (= T) < Qual (= D1 = MVS) < Foc < Top < C. ρ = a pronoun slot ⟨Eρ, posρ, φρ⟩ (proto-role schema); ρA is the slot first-merged in Spec,v and ρX the one in Comp,V. CYC = the spell-out rule of A/B as spell-out §3 (C1 non-heads before heads; C2 non-heads higher-first; C3 heads higher-first). Glosses. Willie's gloss line is copied exactly. Her order for a verb is object.subject.stem (“1s.3s.cut” = 1sg object, 3sg subject). Our Leipzig column only rewrites that order with OBJ/SBJ labels. Verbs are not segmented beyond what she prints.
Bottom line, in plain words
SOURCEDTrees. Willie draws no tree anywhere in Chapters 1–2 (every page read). An automatic scan of all 243 page images for the slanted lines of tree diagrams also found none elsewhere in the thesis, but that check is weak (§1). Her only structural devices are a prefix table ((1), p. 28), statements of linear order (“the subject is the innermost (rightmost) argument”, p. 27), and a list of properties of the Maximal Verb-Sentence ((44), p. 50).
SOURCEDHer framework, stated outright. She says it in three places. “In this thesis, I will adopt Jelinek's (1984, 1990) proposal on Pronominal Argument languages; in other words, my analysis will treat nominals in Navajo as adjuncts to the sentence, occupying A′ positions, while the pronominal prefixes on the verb occupy the A positions” (p. 22). “In terms of the Government and Binding approach (Chomsky, 1981) the MVS includes incorporated pronouns that satisfy all the argument positions that the verb is subcategorized for. These incorporated pronouns leave a trace in argument positions. The pronouns and trace form a chain that receives the theta role and case assigned by the verb” (p. 25). And in Chapter 9: “it is the resulting chain consisting of the pronoun and its trace that for convenience in exposition I am calling a pronoun” (p. 221). So the class (b) placement stands, and it can now be stated exactly. The machinery is Government and Binding (A and A′ positions, traces, chains, theta roles and Case assigned by the verb, government, LF). But the clause is just “the sentence”. There are no S, IP or VP labels, and the subject's position is never given structurally.
HYPOTHESISWhat the restatement shows. (1) Willie's “pronoun (chain)” is exactly this site's role slot. The trace marks where the slot is first merged, which fixes its role, and the pronoun marks where its features are pronounced inside the verb word. With Merge and no head movement (Hale 2003), the two are one object, so nothing needs to “incorporate”. (2) Her three consequences of the pronominal-argument view (p. 22–23: nouns are not case-marked; subordinate clauses are adjoined, not embedded; no movement to or from argument positions) all follow from one statement in our spine: nouns enter only after the verb word's cycle (D1) has been built and spelled out. By then the heads that assign Case have finished, and the argument positions are sealed inside a finished word. That is a direct result for the word as the changeover point between rule set A and rule set B. (3) Her order facts (the patient's prefix is always left of the agent's, p. 50; the agentive-passive ʼadi- is “to the right of (internal to) the object pronoun”, p. 36) follow from CYC only if every subject marker is pronounced on a head (M/S) and every object marker is a non-head (rule SUBJ-HEAD, §5). (4) Two forms do not fit as she glosses them. yi-sh-kʼá (Ch. 2 (4a)) and yiiłtsą́ (10) are glossed with a 3rd-person object and a 1st-person subject and begin with yi-. Yet her own p. 28 says the yi- object appears only in direct (all-3rd-person) clauses, and (8) shédléézh, also with a 3rd-person object and a 1st-person subject, has no such prefix (T35a). (5) The reflexive comes before plural da- (ʼá-da-diidleesh, (14), REF.PL), but the reciprocal comes after it (da-’ahiidleesh, (18), DIST.REC). So the two anaphors cannot sit in the same slot of the word (T35b). Update 2026-10-18, 12:46pm: resolved on reflexive vs reciprocal. Kibrik 2005 (from Young & Morgan 1987:70–71) segments the reflexive into two pieces, an outer ʼá- above the distributive and an inner di-, with the 4th-person subject between them in one form. The reciprocal ʼahi- is one piece in the object position. CYC then derives both orders (R1′).
HYPOTHESISFor Grant's thesis. Willie's own definition (p. 221: the chain “I am calling a pronoun” is what gets the theta role and Case) is the weak thesis in 1991 vocabulary. The verb's arguments are role-bearing slots, and nouns are optional add-ons linked to them. Her strong wording, “nominals are not arguments in Navajo” (p. 45), is contradicted in her own text: lexical postpositions “take lexical objects” (p. 50). That is the same split the site found (the strong version fails, the weak one stands). STRATA fits her architecture for its first and last layers (roles are fixed in the verb word; nouns are chosen later, “for reasons relating to the structure of the discourse”, pp. 51–52). Chapters 1–2 say nothing about a separate event-structure layer, and that silence counts neither for nor against it.
1. Sources opened this run, and the tree check (2026-10-18, 4:46am PT)
Willie, MaryAnn 1991.Navajo Pronouns and Obviation. PhD dissertation, University of Arizona (box copy, 243 PDF pages; hdl.handle.net/10150/185691). Newly read on page images (110 dpi, every Navajo line at 300 dpi): Table of contents p. 5; Abstract p. 10; Chapter 1, pp. 11–23; Chapter 2, pp. 24–34 (§§2.0–2.1.2.2). Pp. 35–36 were first read by the 2026-10-11 run; this run crop-checked (17)–(19) on p. 35, which that run had read at 150 dpi only. Re-read this run: pp. 45–51 (§2.5–2.6, already analysed on pleonastic & psych), p. 71, pp. 92–93 (Chapter 3 (76)–(78), which Ch. 1 p. 17 points to), pp. 199–200, pp. 220–221 (Chapter 9 §9.1). Page offset: PDF = printed + 5 for pp. 4–51 (printed 10 = PDF 15), and PDF = printed + 6 for pp. 71–221 (printed 221 = PDF 227). Pp. 37–44 were read on page images by the 2026-10-11 and 2026-10-17 runs and are used as recorded.
Hale 1997 workshop notes (box [local working copy, not in the repository], born-digital text layer), PDF p. 16 with n. 9, re-read for the 3rd-person object rule (OBV) already on the inverse as Agree and EPP.
Cited by Willie in these pages, not opened: Jelinek 1990; Hale, Jelinek & Willie 1991; Sandoval & Jelinek 1990 (the site has S&J 1989); Willie 1989, 1990; Frishberg 1974; Young 1977/1991 (the informal survey); Akmajian & Anderson 1970; Rice 1986; Cook 1984; Young & Morgan 1987; Hale 1977, 1987; Kari 1976; McDonough 1990; Belletti & Rizzi 1988; Chomsky 1981. Nothing below is attributed to them except as Willie reports it.
SOURCEDTree check, method. (a) Visual: every page from printed p. 10 to p. 51 was read on a page image. None has a tree, bracketing or diagram. (b) Automatic, whole thesis: every PDF page was rendered at 100 dpi, and a line detector counted slanted line segments (7°–83°). As a control, Willie & Jelinek 2000's tree pages scored 23 (p. 267, two small trees) and 194 (p. 268). Willie's text pages score 0–19, with no outliers except the library cover sheets (PDF 1–2). The highest text page, PDF 99 = p. 93, was opened and has no tree. (c) The OCR layer has no hits for INFL, Spec, c-command, D-structure, Projection Principle or Theta Criterion. It has hits for “Government and Binding” (p. 25 and the references), “A-bar” (pp. 199, 221), “trace”/“chain” (pp. 25, 221), “LF” (pp. 200, 218) and “S-structure” (p. 199). HYPOTHESIS A small sparse tree could score like the control's 23 and be missed, so (b) and (c) are weak. Only (a) is a full check, and only for Ch. 1–2 (T35h).
2. Her framework, in her words SOURCED
#
Page
What Willie says (quoted from the page image)
Framework vocabulary
W1
10 (Abstract)
“It is argued that Navajo is a pronominal argument language, where arguments are incorporated pronouns, while nominals are adjuncts.” “Identifying the pronominal inflection in the Navajo verb complex as having argumental status, and identifying the role of definiteness in these arguments, provides a unified account of Navajo syntax.”
pronominal argument; adjunct
W2
11–13
Earlier work: Hale 1973's “subject/object inversion”, under which the two arguments “switch grammatical relations”; “In more recent work, it has been proposed that the arguments switch theta roles” (Sandoval and Jelinek 1990; Willie 1990). “I will present a new and different analysis of this alternation in this thesis.” Earlier investigators “interpreted the yi-/bi- prefixes as agreement with noun arguments” and “presupposed empty categories if one or both of the nominals were "missing" (Hale et al, 1977; Platero, 1980; Woolford, 1986; Speas 1990).”
“There is disagreement in the current literature on the argumental status of nominals and pronouns in languages that have obligatory pronominal affixes … In this thesis, I will adopt Jelinek's (1984, 1990) proposal on Pronominal Argument languages; in other words, my analysis will treat nominals in Navajo as adjuncts to the sentence, occupying A′ positions, while the pronominal prefixes on the verb occupy the A positions. This work will show that the interpretation of a nominal depends upon its coindexing with a pronominal argument of the verbal sentence that follows it.”
A / A′ positions; coindexing; adjunction to “the sentence”
W4
22–23
The consequences: “a) the fact that nominals are not case-marked; b) that subordinate clauses (except for the complements of indirect speech verbs, which have their own syntax) are adjoined clauses, not embedded clauses; c) that there is no movement to and from A positions, and therefore no "raising" and Wh- movement.”
Case; adjunction; A-movement
W5
24
“the Navajo inflected verb is interpreted as a complete functional complex. That is, all the argument positions that the verb is subcategorized for are registered in a set of obligatory pronominal prefixes … The Navajo verb cannot appear uninflected for its arguments; there are no non-finite verb forms.” The “classifier” “is typically a valence marker that marks the verbs as transitive or intransitive.” Goal arguments are postpositional phrases, “some of which also appear phonologically attached to left margin of the inflected verb, while others immediately precede the inflected verb.”
subcategorization; “complete functional complex”
W6
25
The verb and its associated postpositional phrases “form a complex that behaves as a unified constituent under fundamental rules of the syntax. I will call this constituent the Maximal Verb-Sentence, the MVS.” Then the Government-and-Binding paragraph quoted in the bottom line, ending: “any nouns that may also appear in the sentence are not in argument positions, but are non-argumental adjuncts, that cannot receive case. These nouns must be licensed by a pronoun (chain) with which they are coindexed (see Hale, Jelinek, and Willie 1991).”
Morphology: “I will not repeat this work; I will assume the analysis given in Young and Morgan.” Focus: “the syntactic functions of these pronominal and anaphoric prefixes, and … the coindexing rules whereby co- and disjoint reference … are marked.”
(division of labour)
W8
27
“The subject is the innermost (rightmost) argument, closest to the verb stem, while the object is to the left of (external to) the subject. These pronominal prefixes are definite in reference.”
linear order only
W9
28
Table (1) (§4). “third person subject pronouns are always phonologically null, while objects vary between yi-/bi- according to the animacy hierarchy. I will take a somewhat different position here. The yi- object pronoun appears only in Direct constructions, while the bi- pronoun appears only in Inverse constructions … The Inverse bi- is clearly cognate with the third person bi- pronouns that appear as possessors and postpositional objects”. “If the number of the subject is three or more, the Distributive prefix da- is added to the dual number of the verb.”
Direct / Inverse
W10
32
“Da- is not an argument, but marks a semantic feature of the subject argument.” Anaphoric prefixes “appear to the left of the subject pronoun in the prefix array. These anaphors mark arguments that have patient theta roles, and cannot have independent reference. These anaphoric prefixes do not vary with person or number. When an anaphor is present, it always requires a -d or -l classifier, marking the verb intransitive. No pronoun or nominal that is coreferent with the reflexive may be adjoined to the sentence.”
anaphor; patient theta role
W11
34
“When the subject of a reflexive is plural, it is also possible to obtain a reciprocal reading.” “When a reciprocal anaphor is used, the subject is necessarily two or more in number, and only one reading is possible.”
—
W12
36
In the “agentive passive”, “the form ʼadi- … occurs to the right of (internal to) the object pronoun, proving that it marks the unspecified agent, and is not an object.”
linear order as evidence for role
W13
220–221
“Typologically, Navajo is a pronominal argument language (Jelinek 1984); that is, all the direct arguments are registered in the Navajo verb via incorporated pronouns. It is a morphological property of Navajo that all closed class, functional elements are incorporated into the lexical items with which they are associated by government. The argumental pronouns that are incorporated into the Navajo verb leave traces in argument positions; it is the resulting chain consisting of the pronoun and its trace that for convenience in exposition I am calling a pronoun. Because it is these incorporated pronouns and their traces that are assigned theta roles and case, nominals in Navajo must occupy A-bar positions. Nominals are adjuncts related to their pronominal heads via a relation of predication, and are comparable to "discontinuous" relatives.”
incorporation under government; trace; chain; A-bar; predication
W14
71, 199–200
The Inverse “is comparable to a passive in that the subject has a Patient theta role, but it differs from the passive in that it is a transitive sentence that has both agent and patient arguments (Chomsky 1981)” (p. 71). Wh-words “always appear as adjuncts in A-bar positions, and thus cannot undergo Wh- movement … I assume that Wh- movement takes place at LF in Navajo” (pp. 199–200). (Already on the site; re-checked.)
LF; S-structure; A-bar
HYPOTHESIS “Complete functional complex” (W5) is also the name of a binding domain in later Government-and-Binding work. Willie gives no citation at p. 24, and that work was not opened, so whether she means the technical sense or just “complete by itself” is OPEN (T35i). The site's nominal-adjunction page quotes her p. 51 use, “by itself is a grammatical sentence”, which suggests the second reading.
3. Classification, sharpened
Question
Answer from pp. 10–51, 71, 92–93, 199–200, 220–221
Status
Flat S tree (pre-VPISH notation)?
No tree of any kind. The clause is called “the sentence” or “the verbal sentence”. Nouns are “adjuncts to the sentence” (p. 22) or “adjoined to the sentence” (p. 32). Per Grant's rule, there is no flat-S tree here to label. The claims are about A vs A′ positions and coindexing, so they can be restated directly.
SOURCED
Government and Binding?
Yes, by name (p. 25: “In terms of the Government and Binding approach (Chomsky, 1981)”). Markers: A/A′ positions, trace, chain, theta role and Case “assigned by the verb”, government, subcategorization, LF, S-structure.
SOURCED
VP-internal subject (VPISH)?
Not stated. Subject vs object is given only by prefix order (“innermost (rightmost)”, p. 27) and by theta-role labels. Nothing in these pages says where the subject's trace is.
SOURCED (silence)
Where are the nouns attached?
“A′ positions”, adjoined to “the sentence”. In Chapter 3 (p. 52, on nominal adjunction) they are “at the left of the MVS”. No height is given, so Willie's text alone cannot decide the attachment-site questions of Condition C by site.
SOURCED / HYPOTHESIS (consequence)
Movement?
Pronouns: “incorporated”, leaving traces (pp. 25, 221). Nouns: “no movement to and from A positions” (p. 23). Wh: none overt, but movement “at LF” (p. 200).
SOURCED; the tension is §6
Class on the audit
(b) transitional, confirmed. More exactly: treeless Government and Binding inside a Jelinek-style pronominal-argument typology, with no statement about where the subject is first merged.
HYPOTHESIS
4. Data: the forms of Chapters 1–2 (and Ch. 3 (77)–(78)) SOURCED
Every form was transcribed from a 300 dpi crop. Willie's ' is written ’. Her hand-added nasal hooks are kept where they are visible. The “Ours (Leipzig)” column only relabels her gloss. A blank in her gloss column means she printed no gloss line.
4.1 Chapter 1: the animacy hierarchy, reported (pp. 14–17)
#
Form (as printed)
Willie's gloss
Ours (Leipzig)
Translation (as printed)
Ch1 (1), p. 14
ʼashkii łį́į́ʼ bizta
boy horse 3-3-kicked
boy horse 3.OBJ(b).3.SBJ.kicked
“the boy had himself kicked by means of a horse” (Witherspoon's translation, as Willie reports it)
Ch1 (2), p. 15
ʼiiʼni łį́į́ʼ yiyiisxí
lightning horse 3.3.killed
lightning horse 3.OBJ(y).3.SBJ.killed
The lightning killed the horse.
Ch1 (3), p. 16
łį́į́ʼ ʼiiʼniʼ biisxį́
horse lightning 3.3.killed
horse lightning 3.OBJ(b).3.SBJ.killed
The horse was killed by lightning.
Ch1 (4a), p. 17
shiłį́į́ʼ ʼiiʼniʼ biisxį́
1SG.POSS-horse lightning 3.OBJ(b).3.SBJ.killed
The lightning killed my horse.
Ch1 (4b), p. 17
ʼiiʼniʼ shiłį́į́ʼ biisxį́
lightning 1SG.POSS-horse 3.OBJ(b).3.SBJ.killed
(same translation as (4a))
Willie's comments. On (2)–(3) (p. 16): “Both these entences [sic] have the same meaning, despite the difference in word order. There is variation across speakers with respect to the acceptability of sentences like these. Young (1977;1991) found this to be true in an informal survey … where he found only about half of the speakers obeying the animacy hierarchy.” On (4) (p. 17): “when a possessive pronoun is prefixed to a nominal, such as łį́į́ʼ "horse" below, the order of the nominals in the sentence is irrelevant … This freedom in word order seems to be possible only for a limited number of sentences, suggesting that the distance between the referents on the animacy scale still seems to be a factor. I explore this matter further in Chapter 3 below.” Ch. 1 (5) on p. 19 is a Sarcee form from Cook 1984 and is not used.
4.2 Chapter 3, pp. 92–93: the follow-up Ch. 1 points to
#
Form
Willie's gloss
Ours (Leipzig)
Translation
(77a), p. 92
shilééchaʼí hastiin yiyiisxį́
1.dog man 3.3.killed
1SG.POSS-dog man 3.OBJ(y).3.SBJ.killed
My dog killed the man.
(77b), p. 93
hastiin shilééchaʼí yiyiisxį́
man 1SG.POSS-dog 3.OBJ(y).3.SBJ.killed
The man killed my dog.
(78a), p. 93
shilééchaʼí hastiin biisxį́
1.dog man 3.3.killed
1SG.POSS-dog man 3.OBJ(b).3.SBJ.killed
My dog was killed by the man.
(78b), p. 93
hastiin shilééchaʼí biisxį́
man 1SG.POSS-dog 3.OBJ(b).3.SBJ.killed
The man was killed by my dog.
P. 92: Frishberg's point “is that if a possessive pronoun is prefixed to a noun referring to a lower being, it is not necessary for this noun to be second in the sentence”; after (77a): “If the order of the nouns is reversed, the meaning changes.” P. 93: “When the possessive pronoun is present, the nominal is now maximally definite, and the two arguments are of equal rank.” (Spelling note: shilééchaʼí is printed without nasal hooks here.)
4.3 Chapter 2: the prefix table and the definite direct arguments (pp. 28–35)
Subject sg.
Subject duopl.
Object sg.
Object duopl.
1.
-sh-
-ii(d)
shi-
nihi
2.
-ni-
-oh-
ni-
nihi
3.
-0-
-0-
bi/yi
bi/yi
4.
ji-
ji-
ho/ha
ho/ha
Table (1), p. 28 (“The underlying forms of the referential pronominal prefixes as given by Young and Morgan”, p. 27), copied cell by cell, hyphens as printed.
#
Form (as printed)
Willie's gloss
Ours (Leipzig)
Translation (as printed)
(2), p. 29
déyá
1s.will go
1SG.SBJ.will.go
I will go
(3), p. 29
naashnish
1s.working
1SG.SBJ.working
I am working. (Hale 1977, per Willie: can also mean ‘he worked’, with -Ø- subject and -sh- marking the perfective)
(4a), p. 30
yi-sh-kʼá
3s.1s.grind:IMPF
3SG.OBJ.1SG.SBJ.grind:IPFV
I am grinding it.
(4b)
ná-sh-kʼááh
iter.3s.1s.grind.IMPF.
ITER.3SG.OBJ.1SG.SBJ.grind.IPFV
I usually grind it.
(4c)
yíkʼą́
3s.1s.grind.PERF.
3SG.OBJ.1SG.SBJ.grind.PFV
I ground it
(4d)
deeshkʼááł
3s.1s.grind.FUT.
3SG.OBJ.1SG.SBJ.grind.FUT
I will grind it.
(4e)
wóshkʼááł
3s.1s.grind.OPT.
3SG.OBJ.1SG.SBJ.grind.OPT
I wish to grind it.
(5), p. 31
ashkii, shishgish
[none under the noun] 1s.3s.cut
boy 1SG.OBJ.3SG.SBJ.cut
The boy cut me.
(6)
ashkii, nishgish
boy 2s.3s.cut
boy 2SG.OBJ.3SG.SBJ.cut
The boy cut you.
(7)
nishédléézh
2s.1s.painted
2SG.OBJ.1SG.SBJ.painted
I painted you.
(8)
shédléézh
3s.1s.painted
3SG.OBJ.1SG.SBJ.painted
I painted him.
(9), p. 32
dayiiłtsą́
PL.3.3.saw
DISTR.3.OBJ.3.SBJ.saw
They saw him/them. *He saw them.
(10), p. 33
yiiłtsą́
3.1s.saw
3.OBJ.1SG.SBJ.saw
I saw it.
(11)
ʼádiistsą́
REF.1s.saw
REFL.1SG.SBJ.saw
I saw myself.
(12)
yiichʼid
1-dual-scratching
1DU.SBJ.scratching
We are scratching it.
(13)
ʼádiichʼid
REF-1-scratching
REFL-1.SBJ-scratching
We are scratching ourselves.
(14), p. 34
ʼádadiidleesh
REF.PL.1dl.painting
REFL.DISTR.1DU.SBJ.painting
We are painting ourselves. (14′) We are painting each other.
(15)
ʼádadiilzhííh
REF.PL.1dl.blackening
REFL.DISTR.1DU.SBJ.blackening
We are blackening ourselves. (15′) We are blackening each other.
(16)
ʼahiidleesh
REC.1dl.painting
RECP.1DU.SBJ.painting
We are painting each other.
(17), p. 35
ʼahiichʼid
REC.1dl.scratching
RECP.1DU.SBJ.scratching
We are scratching each other.
(18)
daʼahiidleesh
DIST.REC.1dl.painting
DISTR.RECP.1DU.SBJ.painting
We are painting each other.
(19)
daʼahiichʼid
DIST.REC.1dl.scratching
DISTR.RECP.1DU.SBJ.scratching
We are scratching each other.
Willie's comments: (5)–(8) show “a third person subject with first person object”, “a second person object”, then a first-person subject with second- and third-person objects (p. 31). (9): “In direct constructions, when da- appears, the subject must be plural (three or more), and the number of the object is irrelevant” (p. 32). (18)–(19): “subjects that are three or more in number, since the Distributive prefix da- is present” (p. 35). In (14)–(15) the gloss line puts “PL” right after “REF”, and the same morpheme (da-) is glossed “PL” in (9), (14), (15) and “DIST” in (18), (19). (12) yiichʼid has no object in its gloss, though the translation has ‘it’. (5) and (6) have a comma after ashkii, and unlike (1) they have no initial ’.
WF-1 SLOT = CHAIN Willie's "pronoun (chain)" <pro_i , t_i> (pp. 25, 221) is one slot
ρ = <E_ρ, pos_ρ, φ_ρ>:
t_i's position = pos_ρ (first Merge: fixes the role in the action)
pro_i's position = where φ_ρ is pronounced inside D1
With Merge and no head movement (Hale 2003 ms. p. 30, "There is no
head-movement in the traditional sense"), the two are ONE occurrence.
"Incorporation" = spell-out of φ_ρ in D1 by CYC, not movement.
WF-2 θ AND CASE θ: read off pos_ρ (Spec,v = ρ_A; Comp,V = ρ_X; Obl(P)).
"Case assigned by the verb" (p. 25) => Agree inside D1:
M/S [uφ] ⟶ ρ_A (subject φ on M/S: the portmanteau, Hale n. 11);
ρ_X: either no probe (Hale 2003 as on /inverse-agree/: Obj = the pronoun itself)
or Qual/E (Hale 1997 pp. 9–10, 16) (choice OPEN, as on /inverse-agree/)
WF-3 NO NOUN AT pos_ρ No noun is first-merged at any pos_ρ. (= the schema's (G);
Willie p. 22: "the pronominal prefixes on the verb occupy the A positions")
WF-4 LATE NOUNS A noun N is merged only in D2 (rule set B): in Spec,FocP, Spec,TopP, or
adjoined, and it is LINKED to exactly one slot ρ ("coindexing", p. 22;
"predication", p. 221). LINK needs a definite, referential slot whose φ
matches N's (p. 18: 4th person, "like first and second, ordinarily excludes any
adjunct that does not match it in "phi" features"; p. 32 anaphors; p. 35 indefinites;
p. 48 "a bare nominal cannot appear in a sentence without a personal
pronoun to which it is adjoined"). (= SLOT-TYPE-LINK of /pleonastic-psych/)
WF-5 B IS INERT Because of WF-4, when N is merged, D1 has been spelled out: its probes
(derives W4) (M/S, v, Qual) are valued and its pos_ρ are inside a finished domain. Hence
(a) no noun is ever an Agree goal of a Case/φ probe → "nominals are not
case-marked" (p. 22 a);
(c) no noun moves to or from a pos_ρ → no A-movement or "raising" of
nouns (p. 23 c). Movement in D2 is A′ only: FRONT (/clause-spine/) for
an overt question word, covert scope "at LF" (p. 200);
(b) a clause enters a sentence the way a noun does: adjoined, or as a
D-closed CP linked to a slot (D-LINK, /relative-cp/), never first-merged
at pos_ρ (p. 22 b). Willie's exception, "the complements of indirect
speech verbs, which have their own syntax", is OPEN (T35f).
WF-6 SUBJ-HEAD Every SUBJECT-slot exponent is pronounced on a HEAD of D1 (M/S): -sh-, -ni-,
-ii(d)-, -oh-, -0-, ji- (table (1)) and the agentive-passive 'adi- (p. 36).
Every OBJECT-slot exponent is a NON-HEAD of D1: shi-, ni-, nihi-, yi-/bi-/Ø,
ho-/ha-, and the anaphors 'ádi-, 'ahi- (p. 32).
With CYC C1 (non-heads before heads), this yields Willie's
"the patient always to the left of the agent" (p. 50; pp. 27, 36–38).
Without WF-6, C2 would put a non-head ρ_A (Spec,v, which c-commands
Comp,V) BEFORE ρ_X: the wrong order for 'adi- (p. 36).
WF-7 3-OBJ 3.OBJ (ρ_X with π = 3), pronounced in D1:
→ y- if c-commanded by a 3rd-person ρ in the same word (OBV; Hale 1997 p. 16)
→ Ø if c-commanded by a 1st/2nd/4th-person ρ and pronounced in E
→ b- otherwise (SOURCED: Hale 1997 PDF p. 16 and n. 9)
Willie's p. 28 ("yi- … only in Direct constructions") and Abstract
(the alternation is "confined to sentences with all third person
arguments") agree. Her (8) shédléézh fits the Ø case. Her (4a) yi-sh-k'á
and (10) yiiłtsą́, glossed 3 + 1s, begin with yi- (T35a).
WF-8 ANA An anaphor value on ρ_X ('ádi- reflexive, 'ahi- reciprocal) has no person or
number (p. 32), is bound by ρ_A inside D1, requires v = d/l ("marking the
verb intransitive", p. 32), and cannot be LINKED (p. 32). It occurs only on a
slot with "patient theta roles" (p. 32) → restated: only on ρ_X, the slot
with the most Proto-Patient entailments (an extension of Dowty 1991:572,
not a departure). 'ahi- needs #(ρ_A) ≥ 2 (p. 34).
WF-9 PL da- is not a slot (p. 32). It is an adjunct that reads the number of ρ_A
((9): "*He saw them"). Its position: adjoined to QualP (Hale 2003 ms.
pp. 17, 39; P-probe on /inverse-agree/) or lower (P-low).
Rule
What it restates
Willie data it covers
Status
WF-1, WF-2
pp. 25, 221 (the chain)
all of (4)–(19)
HYPOTHESIS (restatement)
WF-3, WF-4
pp. 22, 25, 48
(5), (6); Ch. 1 (1)–(4); (77)–(78)
HYPOTHESIS (WF-4's φ-match clause is SOURCED, p. 18)
WF-5
pp. 22–23 (a)–(c)
a derivation of her three consequences from one timing fact
HYPOTHESIS; new
WF-6
pp. 27, 36, 50
(5)–(8), (11), (13); ʼadi- (p. 36)
HYPOTHESIS; new; fails on nothing printed in Ch. 2 except the da- order (WD6). Replaced 2026-10-18, 8:46pm by SH-3 + SH-4 on subject on the head: subject φ is on M/S because of Hale 2003 n. 11's portmanteau condition plus VPISH, and ji-, ’-, areal ho- are not M/S exponents (they precede the qualifier: Hale 2001 (22b, c))
WF-7
Hale 1997 n. 9; Willie p. 28
(7), (8) fit; (4a), (10) do not, as glossed
SOURCED rule / OPEN mismatch
WF-8
pp. 32, 34
(11), (13)–(19)
HYPOTHESIS. Revised 12:46pm: right for the reciprocal; the reflexive is two pieces, not one exponent of ρX (RR-1)
p. 45 is true of the verb's arguments only. A lexical P's object is a noun argument. This is the strong vs weak thesis split, inside Willie's own text.
SOURCED clash / HYPOTHESIS reading
X4
yi- object “only in Direct constructions” (p. 28); the alternation is “confined to sentences with all third person arguments” (p. 10)
(4a) yi-sh-kʼá “3s.1s.grind:IMPF”; (10) yiiłtsą́ “3.1s.saw”; but (8) shédléézh “3s.1s.painted” has no yi-
Either the yi- of (4a)/(10) is not the object prefix (Willie defers to Young & Morgan, not opened, p. 26), or WF-7 needs an exception for these verbs. Willie segments (4a) as yi-sh-kʼá and glosses yi as “3s”, so the clash is on her page.
Headings and contents, p. 5: 2.3 indefinite, 2.4 postpositional, 2.5 pleonastic, 2.6 summary (there is no 2.2)
A numbering slip. The site cites the headings (2.3, 2.4, 2.5).
SOURCED (fidelity only)
HYPOTHESISX1/X2 and the inverse. If Willie's “no movement to and from A positions” is read as covering pronoun slots too, it favours IA-3 (in situ; Hale 2003) over IA-1* (the patient slot raises in the inverse) on the inverse as Agree and EPP. It is a framework statement, not data, so it does not count against IA-1*. Her “comparable to a passive in that the subject has a Patient theta role” (p. 71) is equally compatible with both (T35e).
7. VPISH-aware derivations WD1–WD8
Format as on the clause spine. Statements about where things merge, move or are pronounced are ours (HYPOTHESIS). The strings, glosses and judgments are Willie's (SOURCED). Role labels follow her translations, and their entailments are read Dowty-style (role in the action). V-P = raising of ρA from Spec,v to Spec,M/S inside the word (the site's working choice). Under V-H (in situ) nothing below changes in the strings.
WD1. (5) ashkii, shishgish ‘The boy cut me’: a 3rd-person noun links past a 1st-person slot
AMerge V (unsegmented stem) with ρX (Comp,V). Merge v; ρA in Spec,v (VPISH). Merge M/S; Agree M/S ⟶ ρA: [3] → -0- (table (1), line 3). Move ρA → Spec,M/S (V-P; optional for the string).
ASpell-out D1 by CYC: ρX [1sg] is a non-head → shi-, first (C1). The heads follow: M/S (silent 3), v, V. WF-6: the subject exponent is on M/S, so it follows the object. String shishgish (not segmented further; Willie's gloss is 1s.3s).
BMergeashkii in D2. Link: WF-4 φ-match. ρX is 1st person and “excludes any adjunct that does not match it” (p. 18), so the only target is ρA [3]. There is no direct/inverse choice, since not both slots are 3rd person (VOICE-TF on the clause spine applies only to 3-on-3).
BSpell-out D2: ashkii ⌢ [D1]. Willie's comma after the noun is copied. Whether it marks a pause (a topic) or is just punctuation is not said (T35g).
WD2. (7) nishédléézh vs (8) shédléézh: a 3rd-person object under a 1st-person subject is silent
Σ Σ(√PAINT) = {ρA [1sg], ρX [2sg] in (7), [3sg] in (8)}.
A as WD1 step 2: Agree M/S ⟶ ρA [1sg] → -sh- (table (1)).
ASpell-out D1. (7): ρX [2] → ni-, then the heads: ni⌢shédléézh. (8): ρX [3] is c-commanded by a 1st-person ρA. WF-7 gives Ø if the object is pronounced in E (Hale 1997 n. 9), so the string is shédléézh ✓. With a b- output instead, WF-7 would predict *b…-shédléézh, which is not attested in these pages.
B no nouns.
OPEN Whether (8) supports Hale's “in E” clause in particular, or just “Ø with a non-3rd subject”, needs the verb segmented (Young & Morgan, T35a).
WD3. (4a) yi-sh-kʼá and (10) yiiłtsą́: where the restatement fails as glossed
Σ Σ(√GRIND) = {ρA [1sg], ρX [3sg]}: the same φ as (8).
A as WD2. WF-7 predicts that ρX [3] under a 1st-person ρA → Ø (or b-), never y-.
A Willie prints yi-sh-kʼá with yi glossed “3s”. If that yi is the object, WF-7, her own p. 28, and the Abstract all fail on (4a). If it is not the object (some other prefix that Willie's gloss merges with the object), all three survive, and (4a) is like (8). Our strings decide nothing between these two options. The same holds for (10) yiiłtsą́ “3.1s.saw” beside (9) dayiiłtsą́ “PL.3.3.saw”, and for (12) yiichʼid, glossed with no object at all.
OPEN T35a: segment (4a), (10) and (12) from Young & Morgan 1987 (Willie's stated source for morphology, p. 26). Only a contradicting segmentation would count against WF-7, and none is printed here.
WD4. (10) yiiłtsą́ ‘I saw it’ vs (11) ʼádiistsą́ ‘I saw myself’: an anaphor value closes the slot to nouns
A (11): ρX is valued [ana] (WF-8): no person, no number. Merge v = d-type classifier (p. 32: an anaphor “always requires a -d or -l classifier”). Bind ρX to ρA inside D1 (local). Agree M/S ⟶ ρA [1sg].
ASpell-out D1: ρX[ana] is a non-head → ʼádi- first (C1), “to the left of the subject pronoun” (p. 32) ✓. String ʼádiistsą́.
B No noun or pronoun coreferent with the reflexive may be merged (p. 32). WF-4: LINK needs a slot with independent reference, and ρX[ana] has none. Compare (10): ρX [3] is linkable (a noun could be added; none is printed for (10)).
HYPOTHESIS The anaphor's restriction to “patient theta roles” (p. 32) is a Σ-level condition: the value [ana] is available only to the slot that the proto-role ranking makes the object. That is Dowty's selection principle with one Navajo addition, an extension and not a departure.
WD5. (16) ʼahiidleesh and (14) ʼádadiidleesh ‘we are painting each other / ourselves’: one reading vs two
Σ Σ(√PAINT) = {ρA [1, du], ρX}.
A (16): ρX = [ana, recip]. It needs #(ρA) ≥ 2 (p. 34) and gets one reading. (14): ρX = [ana], plus Pl da- reading ρA's number (WF-9). With a plural binder, the plain anaphor allows a reflexive or a reciprocal reading ((14), (14′); p. 34).
HYPOTHESIS The reading facts need no noun at all. Coreference between the two participants is settled inside the word, as on Speas 1986 ((4.9)/(4.10)).
WD6. (14) ʼá-da-diidleesh vs (18) da-’ahiidleesh: the two anaphors sit on opposite sides of da-
A Input, SOURCED: in (14) and (15) the gloss order is REF.PL; in (18) and (19) it is DIST.REC. The strings agree: ʼá…da… vs da’a…
A CYC on the site's spine. Pl da- is a non-head adjoined to QualP (WF-9, P-probe). The anaphor on ρX (Comp,V) is a non-head lower down. C2 (higher non-head first) → Pl ⌢ ρX for both anaphors. (18) da-ʼahi- ✓. (14) predicts *da-ʼádi-…, against the attested ʼá-da-di-… ✗.
A If Pl is adjoined lower (P-low, to vP), it still c-commands ρX. Same prediction, same failure.
HYPOTHESISThree ways out, none tested: (R1) the reflexive's ʼá- part is merged above Pl (an outer element that binds ρX), and only its di- part sits low, so the reflexive is two pieces. (R2) The reflexive is one piece in ρX, and da- is placed by a morphological rule after syntax (it would then be outside CYC). (R3) The reciprocal ʼahi-, unlike the reflexive, is not in ρX but in a position below Pl. Willie does not segment (14). Which pieces ʼádadiidleesh has is a question for Young & Morgan's prefix template (T35b). The clash is real on the printed glosses, and it is the first Ch. 2 datum that CYC with a single anaphor position cannot derive.
HYPOTHESISUpdate 2026-10-18, 12:46pm (T35b done). R1 holds, in a sharper form. The reflexive is two pieces: ʼá- (Kibrik 2005 position 15, Young & Morgan Ic, disjunct, above the distributive) and di- (Kibrik's “transitivity decrease”, position 8, a head). Kibrik's (8a) ʼá-zh-dí-l-zhééh has the subject prefix between them. The reciprocal is one piece in the ordinary object position. R2 is unnecessary and R3 beside the point. Derivations RD1–RD8 are on reflexive vs reciprocal §7.
WD7. (9) dayiiłtsą́ ‘They saw him/them’, *‘He saw them’: da- reads the subject only
Σ Σ(√SEE) = {ρA [3, pl], ρX [3, #?]}.
A M/S Agrees with ρA. Pl da- Agrees with ρA's number (P-probe, or P-role: a probe relativized to the agent slot, GRANT'S CONJECTURE → HYPOTHESIS on /inverse-agree/). ρX [3] under a 3rd-person ρA → y- (WF-7, OBV).
A Spell-out: Pl ⌢ y- ⌢ heads → dayiiłtsą́ ✓. *‘He saw them’ ✓: da- cannot read ρX (p. 32: “the number of the object is irrelevant”).
WD8. Ch. 1 (4b) ʼiiʼniʼ shiłį́į́ʼ biisxį́ vs (78b) hastiin shilééchaʼí biisxį́: the same order, opposite linking
Σ Σ(√KILL) = {ρA (causation), ρX (change of state)}, both [3].
A Inverse: b- on ρX (WF-7 “otherwise” case; IA-1* or IA-3). VOICE-TF: ρX [top], ρA [foc]. Spell-out D1 = biisxį́.
B FOC-LINK (the clause spine) links the first noun (topic) to ρX and the noun before the verb (focus) to ρA. (78b): hastiin → ρX, shilééchaʼí → ρA: ‘The man was killed by my dog’ ✓. (4b): ʼiiʼniʼ → ρX, shiłį́į́ʼ → ρA: ‘my horse killed the lightning’ ✗. The printed meaning is ‘The lightning killed my horse’, the reverse linking.
B So in (4b) the noun-to-slot assignment follows what is plausible (Willie p. 16: “horses do not kill lightning”), not position. In (77a)/(77b), where both assignments are plausible, “If the order of the nouns is reversed, the meaning changes” (p. 92).
HYPOTHESIS This fits the clause spine's split of “parallel processing” into grammar and parsing. Noun-to-slot assignment behaves like parsing: it can be overruled by the rest of the sentence, and speakers vary (Young's survey, p. 16). FOC-LINK is then a parsing default, not a grammatical condition, at least when one assignment is implausible. Willie herself says the freedom holds “only for a limited number of sentences” (p. 17), so the default holds elsewhere (T35g).
8. What each analysis gets right, and what it misses
Analysis
Gets right (on the data above)
Misses or leaves open
Willie 1991 (GB, pronominal argument)
The verb word alone is a complete clause, with all the verb's participants as obligatory prefixes (p. 24; every form in §4.3). Nouns are optional and need a matching definite slot (p. 18, p. 32, p. 48). Anaphors and indefinites close a slot to nouns. Subject and object order is stated precisely (p. 27, p. 50). da- is not an argument (p. 32, (9)). The MVS as a unit that includes the grammatical postpositions.
No structural subject position (no VPISH statement either way). Traces and incorporation are machinery that Merge does not need (WF-1). “No movement” is ambiguous (X1, X2). The strong wording (p. 45) overreaches her own lexical postpositions (X3). The yi- glosses of (4a)/(10) clash with p. 28 (X4). No attachment height for nouns, so her text cannot decide Condition C.
Jelinek's pronominal-argument proposal, as Willie adopts it (p. 22)
A positions = the pronouns; nouns in A′ positions; no Case on nouns. All three hold on every form in Ch. 1–2.
Says nothing about the A′ height or the spine. Its consequence (b), that clauses are adjoined, has an exception Willie only names (indirect-speech complements, T35f). Jelinek 1990 not opened (T35d).
This site's restatement (WF-1 to WF-9 with CYC)
Derives W4 (a)–(c) from one timing fact (WF-5). Derives patient-before-agent order, including ʼadi- (p. 36), from WF-6 + CYC C1. (8) fits WF-7. (9), (11), (13), (16)–(19) fit.
Fails on (14)/(15) vs (18)/(19) (WD6) without R1–R3 (fixed 12:46pm by R1′, reflexive vs reciprocal). Cannot decide (4a)/(10) without a segmentation (WD3). WF-6 is a stipulation: why subject φ is on a head and object φ is not is not derived (T35c).
Hale 1973 / Creamer / Frishberg (as Willie reports them, pp. 11–17)
An animacy/control ranking constrains yi-/bi-. Frishberg: the possessed-noun effect ((4), (77)–(78)).
Treated yi-/bi- as agreement with noun arguments and needed empty categories for missing nouns (p. 13). These are pre-VPISH, transformational sources (framework audit (a)); not tested here beyond Willie's report.
9. Grant's thesis, Dowty, STRATA, and the changeover
GRANT'S CONJECTURE → HYPOTHESISWeak thesis. Willie's definition of what she calls a pronoun (p. 221: the chain that is “assigned theta roles and case”) is a role-bearing position plus a feature bundle. That is the site's slot ρ, and it is what the weak thesis says verbs take. Nothing in Ch. 1–2 contradicts it. Every form has its participants as prefixes, and every noun ((5), (6), Ch. 1 (1)–(4), (77)–(78)) is optional and linked to one of them. Strong thesis. Willie states it (p. 45, “nominals are not arguments in Navajo”). Her own p. 50 (lexical postpositions take lexical objects) contradicts it, as the site found independently. It fails for P and holds for V.
HYPOTHESISDowty. Willie's role labels (Agent, Patient, Experiencer, p. 47) are used as a mainstream baseline. Two Navajo additions extend Dowty and do not depart from him: the anaphor value is restricted to the most patient-like slot (WF-8), and noun linking is sensitive to the slot's kind (personal vs areal vs indefinite, WF-4). Who “can act on whom” (Witherspoon, p. 14) is a ranking over the participants' roles in the action, as on who acts on whom.
HYPOTHESISSTRATA (theta roles first, then event-structure roles, then noun selection). Willie assigns theta role and Case in one step to the chain (p. 25), but her architecture already separates the layers that STRATA puts first and last. Roles are fixed in the verb word, and nouns are added later, licensed by coindexing, “for reasons relating to the structure of the discourse” (pp. 51–52). WF-5 makes that order derivational: the noun layer cannot reach back into D1. The only event-structure hook in Ch. 1–2 is the classifier as “valence marker” (p. 24), together with the anaphor's need for d/l (p. 32). Neither tests a separate middle layer. WD8 adds one point for STRATA's late noun layer: which noun goes with which slot can be decided by plausibility after the slots and voice are fixed. The three views (adjunct / inside-clause / STRATA) still give the same judgments on these data.
HYPOTHESISThe changeover point (research focus 1). Willie's “complete functional complex” (p. 24) and MVS (p. 25) are the domain of rule set A. Her three consequences (p. 22–23) are a list of what rule set B cannot do: give Case, embed into an argument position, or move into or out of one. On this site all three follow from the word cycle D1 being finished before B starts (WF-5). That is the strongest single argument so far that the changeover is a cycle boundary and not just a change of head direction. WF-6 refines rule set A: inside D1, the order is decided by head vs non-head (subject agreement on a head; objects, anaphors, Pl as non-heads) before c-command. WD6 is the first place where that is not enough.
HYPOTHESISPhrase-level packaging. The MVS includes the verb and its grammatical postpositions (p. 25; (44), p. 50), so the unit that rule set A builds is larger than the word. That is where a verb + postposition pair can be stored as one item, as on postposition units.
10. Tasks T35a–i
#
Task
Needs
T35a
OPEN Is the yi- of (4a) yi-sh-kʼá, (10) yiiłtsą́ and (12) yiichʼid the 3rd-person object prefix? Settles X4 and whether WF-7 (Hale 1997 n. 9) holds for these verbs
Young & Morgan 1987 (template, these paradigms); Kari 1976
T35b
DONE 2026-10-18 (12:46pm) (reflexive vs reciprocal): R1′, a two-piece reflexive (ʼá- high, di- a head) and a one-piece reciprocal in the object position. Follow-ups T36a–h. Original question: why does the reflexive precede da- ((14), (15)) but the reciprocal follow it ((18), (19))? Decide R1–R3 of WD6
Young & Morgan's prefix positions for ʼádi-, ʼahi-, da-
T35c
DONE 2026-10-18, 8:46pm (subject on the head): WF-6 derived from PORT (Hale 2003 n. 11) under VPISH; the deictic subjects are the left-over deictic part of the subject slot, adjoined to the highest head; ho-/ha- objects keep their whole bundle (SH-5). HYPOTHESIS. Original task: Derive WF-6 (subject exponents on a head, object exponents as non-heads) instead of stipulating it; check fourth-person ji- (subject) and ho-/ha- (object) against it
Hale 2003 n. 11; the fourth-person pages' forms
T35d
OPEN Open Jelinek 1990 and Hale, Jelinek & Willie 1991 (cited pp. 22, 25): does either draw a tree or give the A′ height? (joins T27n)
those papers
T35e
OPEN Does Willie's “no movement to and from A positions” (p. 23) cover pronoun slots? If so, she sides with IA-3 over IA-1* (framework, not data). Read Ch. 7 pp. 180–185 on “raising”
Willie Ch. 7 (on the box)
T35f
OPEN Where does Willie treat “the complements of indirect speech verbs, which have their own syntax” (p. 22)? Compare with Platero's open direct-discourse CPs (OPEN-C)
OPEN FOC-LINK as a parsing default: collect every printed 3-on-3 sentence where the order and the meaning disagree (Ch. 1 (4b); others?). Does the comma in (5)/(6) mark a topic?
Willie Ch. 3 pp. 85–95; Creamer; Witherspoon (on the site)
T35h
OPEN Visual tree check for Ch. 3–9 (pp. 52–219). The automatic scan found nothing, but it is weak
Willie page images (on the box)
T35i
OPEN Does “complete functional complex” (p. 24) carry the technical binding-domain sense?
the GB source of the term (not opened)
Sources: Willie 1991 (Arizona PhD dissertation; box copy, page images and 300 dpi crops under [local working copy, not in the repository], not committed): pp. 5, 10–36, 45–51, 71, 92–93, 199–200, 220–221. Hale 1997 (box copy [local working copy, not in the repository]), PDF p. 16 and n. 9. Hale 2003, Platero 1973, Willie & Jelinek 2000 and Dowty 1991 as recorded on the site. Cross-links: framework audit (T27o), open questions (T35), sources.
Restated on the unified structure (2026-11-15). The framework is restated in §1: VPISH (arguments merged in the verb’s scope, subject in Spec,vP) with theta roles, and V→v→T with the subject raised to Spec,TP as the working default in tensed clauses; each pattern is what else moves, with an LF whose scope matches c-command. Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). Flat-S and other pre-VPISH notation below is labelled contrast. Derivations below in which the verb or subject does not raise are labelled contrast; the working default raises V→v→T and the subject to Spec,TP. Status on the spine: see unified structure, §1. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.
2026-11-11 audit pass 2. Wording that used VPISH as the name of a spine, tree, skeleton or set of slots named a structure after the hypothesis. VPISH is a hypothesis name, never a node; the structure is a vP/TP spine with the subject first-merged in left Spec,vP. Only the wording changed.
2026-11-11 audit pass 3. Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference. See the pass 3 report.
Grant asked which sources, analyses, rules, trees and derivations on this site belong to older frameworks (flat S, deep/surface structure, transformations, X-bar without a VP-internal subject) and which are VPISH/TP/CP or Merge-based, and what restating each would involve. This page classifies every source we opened, from the pages we opened, and flags our own rules that borrow older assumptions. Written 2026-10-15, 9:50pm run (PT), after the 9:20pm revision of Jelinek & Baker attach.
Ground rules (Grant's). Each source is treated as possibly partly right. An older notation is not a refutation: only data that contradict a hypothesis count against it. Judgments, examples and glosses stay valid whatever framework reported them; analyses are what need restating. A classification is made only from pages actually opened. Where a source is known only through another work, it is marked unverified and not placed.
Bottom line, in plain words
HYPOTHESISSorted by what was opened: 4 sources are (a) pre-VPISH (Jelinek 1984, Sandoval & Jelinek 1989, Hale & Platero 1974, Creamer 1974); 6 are (b) transitional (Speas 1986, Willie 1991, Willie & Jelinek 2000, Diesing & Jelinek 1995, Hale 1997 workshop notes, Hale 1999), plus Speas 1990 on prior-run notes only; and the minimalist or VPISH-aware ones are Hale 2003 (with a GB residue), Hale 2001, Baker's agreement/dislocation paper, Goad & Travis 2021, and the Bruening, Abe, Marantz, Embick & Noyer and Déchaine & Wiltschko papers. Ten groups of named sources were not opened (Hale 1973, Kari, Young & Morgan, Sapir & Hoijer, Rice 2000, Barss et al., Hale & Keyser, Platero and Perkins, Speas 1990 Ch. 4, Baker 1996, §3) and are not classified.
HYPOTHESISDate does not decide the class. Hale 1999 draws S → Pro VP trees with d-structure (pp. 171–172); Hale's own 1997 workshop notes call d-structure and s-structure “ultimately fictions” and the Merge-and-Move theory “conceptually superior” but keep using them (PDF pp. 25–26); Hale 2003 uses Merge (p. 3) yet still mentions the Case Filter as a motive for movement (p. 12). The audit therefore reads markers on the page, not years.
HYPOTHESISThe five biggest items on our own site that need restating (details in §5):
The Condition C derivations and A3 (F1). The “nouns outside the subject's reach” view is Hale's tree (9), where the nouns are adjoined to S above a subject that is a daughter of S (pp. 171–172). That is S-adjunction in a pre-VPISH tree. In a vP/TP/CP spine the question is which projection the nouns adjoin to: CP, TP/M/SP, vP, PP, or a left-peripheral specifier. Baker's Kinande trees (16) and (19) show exactly that choice drawn (pp. 14–17), for Kinande, not Navajo.
Rule sets A and B (F3). They are written as linear-order rules for heads (“H⌢C”). That is a head-direction parameter in the X-bar style. In Hale 2003's terms the hierarchy comes from Merge and the order from spell-out, so the rules should be restated as statements about spell-out, and tested against an antisymmetric alternative (not opened). Done 2026-10-16, 12:46pm (T27j): A/B as spell-out; Kayne 1994 now opened.
The inverse analysis (F5, F6). The site's “passive-like” wording and the case story (ergative agent, absolutive patient) come from Hale 1973 as reported (inversion “between deep structure and surface structure”, Creamer p. 29) and from Willie 1991 (Case assigned to pronouns before incorporation, pp. 72, 100). In VPISH terms it is a choice of which slot M/S probes and raises, with a minimality problem when the patient crosses the agent.
Nominal adjunction, wh-questions and clause chains (F7–F9). They rest on Willie 1991's GB machinery: the MVS as a sentence-like unit nouns are “adjoined to” (p. 51), wh-words in A-bar positions at S-structure with scope at LF (pp. 199–200), and Platero's “parallel processing” (p. 135). Each needs a position in the spine: where the MVS sits, which left-peripheral head hosts topic and focus (W&J's trees (43)–(48), pp. 267–269, are S-adjunction over IP), and whether parallelism is syntax or parsing.
Relative clauses (F10). Hale & Platero 1974's deletion rule (9) and raising rule (12) are transformations over S and NP. Willie 1991 treats the enclitic as the head (pp. 153, 158). A VPISH/TP/CP statement would make the relative a clause under a D-like head and ask whether the noun raises into its specifier.
HYPOTHESISWhat does not need restating. The Navajo data (every form on the site, with the judgments in it) and the descriptive generalizations that report them: Creamer's rankings, Witherspoon's ranking sentences, the yi-/bi- interpretation principles as facts, Willie's coindexing facts, Hale 1999's (10)–(12), and the verb-template order. The site's role-slot (Σ) schema and V-H/V-P derivations are already VPISH/Merge-based; they are flagged only where they inherit an older motive (Case-driven raising, S-level adjuncts).
HYPOTHESISCorrections made while auditing. (1) The 9:20pm page said Willie & Jelinek 2000 was “abstract only”. That was wrong: pages 264–278 were opened as page images in earlier runs, and trees (43)–(48) on pp. 267–269 were re-read now. Corrected on Jelinek & Baker §12, open questions and sources. (2) A leftover “Supports A3 for this Warlpiri analysis” in the sources table (Jelinek 1984 row) is replaced by “no verdict on A3”. (3) Two files on the box labelled Barss et al. 1989 and Perkins 1978 were wrong files (a Wittgenstein excerpt, an Eddington & Lachler stem paper, a Utah honors thesis) and were not used.
Clause built by subdividing S; or rules that move or delete NPs between deep and surface structure; subject not VP-internal
Flat S with V, AUX and noun daughters; “deep structure”, “surface structure”, “transformation”, deletion/raising rules over S and NP phrase markers; INFL “governs” the subject with no VP for it
(b) transitional
GB-era, X-bar, early VPISH: heads and specifiers, but S-structure/LF, Case Filter, government or barriers; or functional projections without v/Merge
Merge-built tree, external argument in Spec,v/vP, TP/CP above, no D-/S-structure
Merge, v/vP with a Spec,vP subject, TP/CP trees, Agree or Spec–Head by feature, no mention of d-structure as a level
(n) not a syntactic-theory source
Ethnographic, semantic, typological or descriptive; no trees or rules over phrase markers
Absence of the markers above after reading the cited pages
(u) unverified
Known only through another work, or listed without opening
Not placed. A classification “by date” is not made
How the markers were found. Text-layer keyword counts (INFL, VP-internal, Spec, X′, Merge, vP, TP/CP/IP, c-command, deep/surface structure) on the box copies, then page images of the pages where a hit or a tree occurs. Where a scan has no text layer (Hale 1999, W&J 2000, S&J 1989, Witherspoon), I looked at the page images only. A zero count on an OCR layer is weak evidence (Willie 1991's layer is noisy), and is stated as such. Page numbers are the printed ones unless the row says “PDF p.”.
2. Sources opened, with class
Source
Class
Evidence on the pages opened
Data that stays valid
Used on
Jelinek 1984, “Empty categories, case, and configurationality” (Warlpiri)
(a)
SOURCED Flat S: CPPs “sisters to the verb under S” (tree (28), pp. 50–51; p. 73). “INFL governs the subject”; objects “appear in a VP constituent and subjects do not” (p. 44). GB vocabulary on a flat S.
Warlpiri examples; the typological claim (73b) as a claim
Sandoval & Jelinek 1989, “The bi-construction and pronominal arguments in Apachean” (Jicarilla, Navajo)
(a)
SOURCED Flat S trees (42a′), (42b′): S with NP, PP and V daughters, p. 352. “The nominals are adjuncts to the sentence, not under a PP node” (p. 351). Verbs classified by how many adjunct slots they allow, V, V′, V″, “V′ = V + Postposition” (pp. 346–347). Empty-category talk (pro, PRO) at p. 345.
Examples (1)–(9), (30)–(35); coindexing facts; goals (a)–(d) on p. 339
SOURCED Deletion rule (9) and raising rule (12); “passive-like subject-object inversion transformation (cf. Hale 1973)”, p. 15; “The problem … resides in the surface structure, not in the transformational rules”, p. 21; tree diagrams with S and NP nodes and NPs in deep-structure positions (e.g. (19), p. 15). (Printed pages mapped from the text extract: extract page + 5.)
Relative forms (1)/(4); island facts (p. 13); the interpretation facts behind (28) and (44)
SOURCED “The Navajo rule of subject-object inversion … occurs between the deep structure and surface structure” (p. 29); “the first noun phrase in the deep structure is the subject” (p. 31).
The 28 sentences (c1)–(c28) and the eight status groups
Speas 1986, Adjunctions and Projections in Syntax (MIT diss.)
(b)
SOURCED IP, CP and DP after Fukui & Speas; the maximal projection of a lexical category is V′, not VP; “specifiers of Functional heads are … always moved from within their complement” (p. 89, page image). Projection Principle, CHAIN with a non-θ tail (pp. 201–202), ATB (p. 264). The PF-movement option for prefixes is raised and declined (p. 201).
Navajo examples (2.95), (2.96), (4.6)–(4.10), (4.119)–(4.121) and her judgments (from Platero)
Opened in an earlier run (front matter, Ch. 1–3, Warlpiri pp. 159–172). The box copy is no longer present, so not re-checked now; Ch. 4 (Navajo) was never available. Treat the class as unverified for the VPISH question.
Willie 1991, Navajo Pronouns and Obviation (Arizona diss.)
(b)
SOURCED “In terms of the Government and Binding approach (Chomsky 1981) the MVS includes incorporated pronouns … These incorporated pronouns leave a trace in argument positions” (p. 25); Case on pronouns before incorporation (p. 72); external vs internal argument (p. 100); wh-movement “at LF”, S-structure adjunct positions (pp. 199–200). Text-layer search found no INFL, VP-internal subject, Spec or X′ (noisy OCR; weak evidence). Checked on page images 2026-10-18 (T27o): no trees in Ch. 1–2; the framework is named outright (Jelinek's pronominal-argument proposal, A vs A′ positions, p. 22; the chain “I am calling a pronoun”, p. 221); the clause is only “the sentence”, with no statement of where the subject is first merged (Willie's framework).
Every Navajo example (the dictionary forms from Ch. 2–8) and the coindexing facts
Willie & Jelinek 2000, “Navajo as a discourse configurational language”
(b)
SOURCED Trees (43)–(44), p. 267, and (47)–(48), p. 269: S with FOCUS / TOPIC Adjuncts over IP; IP over 3ACC and V′; V′ over 3NOM and V stem; S′ with COMP in (48). Labels S, S′, IP, V′; no TP, CP or vP.
Examples (57)–(62), (70)–(77); the animacy and wh facts
Diesing & Jelinek 1995, “Distributing Arguments” (NLS 3:123–176; Egyptian Arabic and others, not Navajo)
(b), VPISH-aware
SOURCED “the availability of a VP-internal subject position as well as the VP-external position is assumed” (p. 125, citing Kuroda 1988, Pollock 1989); two subject positions (p. 134); TP over AspP over VP, subjects through Spec,AspP and Spec,TP (tree (74), p. 169). Conditions stated on S-structure and LF (pp. 123, 134). Read by text-layer search plus the cited pages only.
—
not yet used on the site; relevant to T26g
Hale 1997, “Remarks on the syntax of the Navajo verb”, Parts I–III (NLA workshop, Tsaile, July 1997; MIT papers hale023.pdf; 61 pp.)
(b), knowingly
SOURCED Trees with EP, MP, VP (inverse (15), PDF p. 25). “The syntactic levels of d-structure … and s-structure … are ultimately fictions, artifacts of an earlier tradition”; the Merge-and-Move theory is “at least conceptually superior”, but the traditional representations are kept (PDF pp. 25–26). The subject is a “distinguished adjunct” of VP, “in a minimal sense ‘internal’ to VP, as in the VP-internal Subject Hypothesis” (n. 20, PDF p. 36). Case and the EPP as motives for raising to Spec of Infl/Tense (PDF p. 46). C-government and barriers (PDF p. 7). Printed folios not checked.
Navajo verb words and their segmentation
new on the site (this page)
Hale 1999, “Conflicting truths”
(b)
SOURCED Trees (8), (9), pp. 171–172 (page images): S over Proi and VP; the relative NREL adjoined to S in (9); “d-structure argument positions” (p. 171). The subject pro is a daughter of S, not in Spec,vP.
(1)–(3), (7), (10)–(12) and the judgments reported
Hale 2001, “Navajo verb stem position …” (LI 32:678–693)
(c)-leaning
SOURCED “Merge relation” for Voice and Mode/Subj (p. 680); “two raising operations” derive the surface order (p. 680); adjuncts “asymmetrically c-command the verb” (p. 689); head-initial extended projection (p. 689). No d-structure, no Case Filter in the pages read.
Template order data; the Iterative/Plural order facts
Hale 2003 ms., “On the significance of Eloise Jelinek's Pronominal Argument Hypothesis”
(c), with a GB residue
SOURCED “the single operation now called Merge within the Minimalist Tradition” (pp. 3–4); spine Qual > M/S > v > V (pp. 31–32); pronoun in Spec,v or Spec,V (pp. 27–29). Hale–Keyser argument structure assumed (pp. 44, 65). But raising the object to Spec,M is considered as forced by the Case Filter (the object “governed by C”) and rejected (pp. 11–12; the Case passage is on p. 12).
SOURCED TP/vP trees with proi and Agri+T, subject in Spec,vP (16a–b), pp. 14–15; dislocated subjects adjoin to TP below Focus, dislocated objects to CP (19), pp. 16–17. On Navajo only n. 16 (PDF p. 47).
(c) (not Navajo; recorded from the runs that opened them)
Bruening: Voice [EPP], movement (30), p. 21. D&W: pro-DP/φP/NP. Abe: derivational c-command (3). Marantz, E&N: Late Insertion. Not re-read this run; marked “prior-run notes”.
Witherspoon 1977 ch. 3; Dowty 1991; Smith, Perkins & Fernald 2007; Kibrik 2005; the NLA grammars (Fernald et al.); Gessner 2003; Goertz et al. 2006
(n)
Ethnographic semantics (Witherspoon: reports “a rule of subject-object inversion” from Hale 1973 at p. 63, adds none of his own); lexical semantics; aspect semantics; typological template; descriptive grammar. No phrase-structure trees or rules read. (Kibrik: one stray TP/CP/IP token, no tree.)
All their data
many
Other sources listed on /sources/: Ramchand slides and Bruening 2010 (review of Ramchand 2008); Potts (review of Borer)
(c) (non-Navajo; opened in full in the 1:07pm run, except the books themselves)
init/proc/res and “arguments introduced by functional heads”: minimalist-era argument-structure work. Not re-read now.
Typological universals, typological database, a Greenlandic grammar, child-language corpus study, a how-to note, a book review (secondhand for Rice 2000), a survey chapter (prior run), a phonology handout whose only syntax is (16)–(17). No trees read.
Haugen & Siddiqi; Stepanov; Stockwell et al.; Bruening & Al Khalaf; Hale & Platero NPI ms. (earlier labelled Hale & Perkins); the binding-theoretic YI- thesis
(u)
Abstract or search snippet only (see sources). Not placed.
Grant's list named several of these. None was opened, so none is classified. What is known secondhand is given so the follow-up is cheap.
Source
What is known, and from where
Why it is not a class
Hale 1973, “A note on subject-object inversion in Navajo”
Described as “a rule of subject-object inversion” read “from a surface level point of view” (Witherspoon p. 63); an “inversion transformation” (Hale & Platero p. 15); a rule between deep and surface structure (Creamer p. 29). Vocabulary of the transformational period.
Three secondhand reports of the vocabulary; the original not read. (a)? unverified.
Kari 1976, 1979, 1989
Template and zone descriptions (Goad & Travis 2021, p. 3; Smith et al. 2007; Kibrik 2005)
Not opened. The works are descriptive; whether they use trees is unknown.
Young & Morgan 1943, 1987
Reference works; Hale 1997 (PDF p. 1) says that through “Sapir, Hoijer, and … Young and Morgan” it is well known that the verb word is a stem preceded by prefix categories in a fixed order
Not opened. This sources only the verb-centred template description, via Hale 1997. It does not source Grant's remark about WWII-era verb-centred accounts, and says nothing on their use of trees.
Sapir & Hoijer 1942; Hoijer 1945
Bibliographic only (ERIC ED102820 and ED101549 reference lists; Hale 1997 PDF p. 1 names Sapir and Hoijer)
Not opened
Rice 2000, Morpheme Order and Semantic Scope
Head-final verb, stem raising and right-adjunction to Voice, then Mode/Subj, as rendered by Hale 2001 pp. 679–680, trees (4)–(5); the Scope Hypothesis (Bird 2001; Goad & Travis 2021 p. 2)
Not opened. Hale's rendering is an adjunction tree, so Grant's guess of (c) is not supported by what was opened. Unverified.
Barss, Hale, Perkins & Speas 1989 (Aspects of Logical Form in Navajo) and 1991/92 (Logical Form and Barriers in Navajo)
Titles and publisher pages only (10.1515/9783110852394-009; 10.1007/978-94-011-3472-9_2). Barriers vocabulary in the 1992 title
Not opened; no open PDF found. The two box files labelled for it were wrong. Unverified.
Hale & Keyser 1993, 1998; Hale & Platero 1996
Hale 2003 assumes the Hale–Keyser theory of argument structure (pp. 44, 65). Hale 1997 builds lexical projections with specifiers and complements (PDF p. 36)
Not opened; the earlier extract of H&P 1996 is no longer on the box. Grant's guess of (b) is unverified.
“Inversion rule between deep and surface structure” (Hale 1973 via Creamer p. 29); Case on the inverse patient (Willie pp. 72, 100); EXT/INT arguments
Noun attachment
Willie's (1)–(36) coindexing judgments; W&J word order
“Adjoined to S” (Jelinek, S&J, W&J, Hale 1999); “adjoined to the MVS” (Willie p. 51); Baker's CP/TP adjunction for Kinande
Condition C
Hale (10)–(12); Speas (4.15), (4.119)–(4.121) judgments
The tree in which the subject c-commands the noun: S → Pro VP (Hale (8), (9)); A1–A10
Questions
Willie (2)–(32); W&J (57)–(60)
A-bar adjunct positions at S-structure; wh-movement at LF (Willie pp. 199–200)
Relatives
H&P (1)/(4), island facts; Willie Ch. 6 (1)–(26)
Deletion (9) vs raising (12) as rules; Platero's parallel parse as a rule
Word order
WALS 26A, 83A, 85A; NLA particle positions
The head-direction parameter behind rule sets A and B
5. Our own rules and derivations that borrow older assumptions
HYPOTHESIS Each row says what the page borrows, from which opened source, and what restating it in VPISH/TP/CP terms would involve. The restatements are ours, labelled HYPOTHESIS; none is a claim about a source. The pages are already Merge/vP-based in their bones (the Σ schema, Spec,v and Comp,V first-merge sites, V-H and V-P); the flagged items are the inherited bits.
“Nouns adjoined to S above the subject”, with the subject pro as a daughter of S sister to VP (Hale 1999 trees (8), (9), pp. 171–172); the tree definition of c-command (the page notes the derivational definition gives the same answers)
Replace “S” by named spine nodes. Run the Condition C derivations with the noun adjoined to each of: CP; TP/M/SP; vP; PP; a left-peripheral topic/focus specifier. Subject merged in Spec,vP and optionally raised to Spec,M/SP (V-P), so its c-command domain is vP at merge and M/SP after raising. A3 then holds for CP/TP/specifier attachment and fails for vP/PP. Compare Baker's TP/CP choices for Kinande (pp. 14–17). Done (T27i, 2026-10-16): the expected split “A3 holds for CP/TP/specifier, fails for vP/PP” is right except for vP, which is outside the reach of a subject that stays in Spec,v and inside only if the subject raises (table).
Already labelled pre-VPISH with no verdict (9:20pm). Remaining work is T26g and T26b: W&J's (43)–(48) (S-adjuncts over IP, pp. 267–269) are the nearest to a translatable tree.
Head-direction rules, LIN(H, C) = H⌢C or C⌢H, i.e. a head parameter of the X-bar style; Greenberg-style OV/postposition facts read as head–complement facts; B3 = Willie's MVS (Willie p. 25)
Fix hierarchy by Merge (Qual > M/S > v > V, Hale 2003 pp. 31–32) and treat order as spell-out: rule set A becomes a statement about how a word-sized spell-out domain linearizes its heads, rule set B about how clause-level material is linearized. Re-ask whether the “word” is a spell-out domain (a phase-like unit) instead of a rule boundary. Test against an antisymmetric account (Goad & Travis p. 6; Kayne not opened). Done (T27j): restated as SA1–SA5 / SB1–SB5 and the single rule CYC; antisymmetric versions AS-1, AS-2 tested (A/B as spell-out).
Raising of the object to an outer Spec,M/SP “for Case” was Hale's considered-and-rejected option (Hale 2003 pp. 11–12; Case Filter, p. 12; Hale 1997 PDF p. 46 “Case and the EPP”). Clause-level DPs adjoined at the “left edge” (B) echo W&J's S-adjuncts
Motivate V-P movement by an EPP/φ-probe on M/S rather than by Case; keep V-H as the Hale default. Put B-layer topic/focus DPs under named left-peripheral heads or as TP/CP adjuncts (see F1, F8).
“Passive-like” subject–object inversion between deep and surface structure (Hale 1973 as reported, Creamer p. 29, Hale & Platero p. 15); EXT/INT argument and Case on the pronouns (Willie pp. 72, 100)
Voice as a choice of which slot M/S probes: direct = ρA in Spec,v; inverse = the higher-ranked or definite slot, raised to Spec,M/SP over ρA (V-P). State the minimality issue (patient crosses agent). One published model of an inverse as a Voice head with an EPP is Bruening 2005 (p. 21, Algonquian; opened). Test with da- (number & scope). Done (T27k, 2026-10-16): the inverse as Agree and EPP states IA-0 to IA-3; the minimality problem is avoided by an outer Spec,v under raising, and it reappears in the direct voice under Hale 1997.
Nouns “adjoined to the MVS” (Willie p. 51); MVS as a sentence-like unit (p. 25); coindexing principles (17), (20)
State where the MVS sits in the spine (M/SP? the whole extended projection?) and which node the noun adjoins to. Replace coindexing by linking to a φ-bundle (Σ). Same question as T26g. 2026-10-17: done on clause spine §4.1: MVS = QualP = D1; nouns in Spec,FocP/Spec,TopP, linked to [foc]/[top] slots.
Wh-words as A-bar adjuncts at S-structure, scope at LF, no wh-movement (Willie pp. 199–200); focus movement (W&J pp. 272–273)
Q and Focus heads in the left periphery (the W&J abstract, found by search, points to Rizzi); wh-in-situ with Q-Agree instead of LF movement; “no long-distance wh” as locality of the Σ-slot (the site's WH-SLOT) or a phase edge. 2026-10-17: W&J name no Q/Focus heads (pp. 266–273); placement on clause spine §4 (Foc above QualP, Fochi above TopP for fronting); the inverse block on fronting is T31h.
“Parallel processing” as an interpretive rule (Willie p. 135, crediting Platero 1976 and Speas 1990, unopened); =go clauses “adjoined” (Willie pp. 132–140); “inner speech” clauses “loosely adjoined” (p. 213)
Say which node the =go clause adjoins to (TP or CP) and whether parallelism is syntax (ATB-like, Speas p. 264) or a parsing preference. Unverified until Platero 1976 is read. 2026-10-17: split on clause spine §5: cast identity and mapping are grammar, noun-to-verb assignment is parsing (Platero 1973, opened); Platero 1976 still unopened (T31g).
Deletion rule (9) and raising rule (12) as transformations (Hale & Platero pp. 11–24); relativizer as a “determiner” (p. 27); Willie's enclitic head (pp. 153, 158)
Relative as a clause (CP) selected by a D-like enclitic; ask whether the internal noun raises to Spec,CP (raising) or is matched; keep islands as data. No Navajo-specific modern treatment opened. 2026-10-17: done on relative as CP: Opρ (a slot copy) raises to Spec,CP under D -ę́ę/-ígíí; islands from Platero 1973 Ch. II; matching kept as the runner-up (T34a).
Speas's CHAIN and ATB (pp. 201–202, 264); weak crossover and Condition C as A/A′ diagnostics (Ch. 2, Ch. 4); V′ without VP (p. 89)
Re-run WCO and Condition C with a Spec,vP subject and derivational c-command. FB's “bundle plus linked noun” already restates her CHAIN as Agree between a bundle and a noun; make that explicit.
“Raising” modelled on Dative Movement (Willie p. 182)
PR-SLOT (b), a two-slot schema with no movement, versus (a), movement of the possessor from Spec of the object to a higher specifier; (a) would be an A-movement test for VPISH (T16a).
First-merge sites Spec,V vs Spec,v (Hale 2003 pp. 27–29), which rely on the Hale–Keyser theory (pp. 44, 65; not opened)
Not pre-VPISH. Flag only that the “position follows from the entailments” idea rests on an unopened theory (T27h).
H&P 1996 lists via Gessner and Goertz
HYPOTHESISAlready VPISH/Merge-based, no restatement flagged: the Σ schema and the V-H/V-P pair; number & scope; the fourth-person pages; feature bundles (FB-1 to FB-4, FB-6); the verb dictionary (its first-merge sites are HYPOTHESIS, its forms are sourced). FB-5 is the exception: its exemption of linked nouns depends on A3 (F1).
6. New OPEN tasks
#
Task
Source to open (not opened yet)
T27a
OPEN Open Hale 1973 itself and classify it; if it is transformational, say so from the page
Hale 1973 (Kachru et al. eds., pp. 300–309); the NLA Hale archive (fernald.domains.swarthmore.edu/nla/halearch/) lists course notes (1977) and a relative-clause file
T27b
OPEN Open Barss, Hale, Perkins & Speas 1989 and 1991/92 and classify (c-government and barriers expected from the title; not verified)
OPEN Open Young & Morgan 1987 and 1943 for trees or rules; and Sapir & Hoijer 1942, to source (or not) Grant's remark on older verb-centred descriptions
Young & Morgan; Sapir & Hoijer 1942; Hoijer 1945
T27d
OPEN Open Rice 2000 and classify it (Grant guessed minimalist; Hale's rendering is an adjunction tree)
Rice 2000 (CUP)
T27e
OPEN Open Kari 1976/1979/1989 and classify
Kari
T27f
OPEN Get Speas 1990 Ch. 4 (Navajo, pp. 202–281) and check whether its subject is VP-internal; compare with Speas 1986 p. 89
Speas 1990
T27g
OPEN Open Platero 1974, 1978, 1982 and Perkins 1978 and classify the generalizations behind LINK-NEAR and parallel parse
Platero; Perkins
T27h
OPEN Open Hale & Keyser 1993/1998 and Hale & Platero 1996 and classify; check what the causative-split page depends on (F13)
Hale & Keyser; Hale & Platero 1996
T27i
DONE 2026-10-16 (4:46am) F1: re-run D6 and the Condition C derivations with adjunction sites CP, TP/M/SP, vP, PP, left-peripheral specifier, using derivational c-command only; record which sites make Hale (10)–(11) predicted and which Speas (4.119). Result (Condition C by site): what matters is whether the noun is in the tree before the subject's last Merge, not whether it is an adjunct. Rows ARG, PP, VP, vP with subject raising, M/SP below the raised subject, and a moved left-edge noun predict Hale; vP with a subject that stays in Spec,v (Hale 2003), M/SP above the subject, QualP (“the full verb word”), CP and a first-merged left-edge noun predict Speas; none predicts both; (11b) needs A10 at every site. A3 holds in 5 of 10 non-argument configurations; D6 stays compatible, undecided. New tasks T28a–f
none new (T26g, T24)
T27j
DONE 2026-10-16 (12:46pm) F3: restate rule sets A/B as spell-out statements over Hale 2003's hierarchy; test an antisymmetric alternative. Result (A/B as spell-out): A = Hale's skeleton filling (SA1–SA5), B = SB1–SB5; both reduce to one rule applied in two cycles (CYC: non-heads first, then heads, each highest-first; the finished word is one non-head in the clause). CYC derives Hale (1), (13), (20b), (31a), (34) and the NLA/W&J clauses with no movement, and derives Hale's lexical-head-final split. The word is a spell-out cycle (stipulated); exceptions are Hale's own (n. 13 obliques, 2001 deictic infixation); da’ must be a phrase. Antisymmetry (Kayne, opened) gets the conjunct-head order free but needs 6–8 raisings and 14 silent heads for the five words, and does not make the word a boundary; IRP (G&T) does, by a word-only rule. New tasks T29a–j
DONE 2026-10-16 (8:46pm) F5/F6: restate the inverse as a probe/EPP choice on M/S; compare V-P and V-H on da- and on the minimality problem. Result (the inverse as Agree and EPP): four options tested. IA-1* (Bruening-style optional [EPP] on v, raising in the inverse only, M/S agrees with the agent after the patient has moved) and IA-3 (Hale 2003, no movement, a topic feature decides y-/b-) fit every form on the site. IA-2 (Hale 1997, newly used: object to Spec,MP in both voices, then EPP to Spec,EP) has a crossing in the direct and a freezing clause that blocks da-. IA-0 (inverse linking) is disfavoured by da-. Raising plus CYC plus dabiiłtsą́ needs da- to be role-targeted, bi- to be an agreement head, or da- to copy M/S's number. INV-5 follows from closeness under raising if the PP is above the theme. Ranking = a comparison at Σ that chooses the feature. New tasks T30a–h.
Hale 1997 PDF pp. 9–25 (newly used); Bruening 2005 pp. 1–6, 13–24 (re-opened); Hale 2003 pp. 10–17, 27
T27l
DONE 2026-10-17 (4:46am) (clause spine). F7/F8/F9: place the MVS, the nouns, the =go clause and topic/focus in the spine; re-read W&J pp. 267–273 for any head names; separate syntax from parsing in “parallel processing”. Result: W&J pp. 266–273 name no Top/Foc/C/T heads (only S, S′, IP, V′, COMP, FOCUS/TOPIC Adjunct, “Maximal Verb Complex”). Restated: MVS = QualP = spell-out cycle D1 (W&J p. 270, (50b)); silent Foc and Top above it host the preverbal and outer nouns; a fronted question word goes higher; the =go clause is a CP adjoined above the matrix nouns (FocP vs TopP open). “Parallel processing” = cast identity (grammar: holds with no nouns, *(12)) + mapping (grammar; W&J (49) prints R2 indices for Willie's (9a) string) + noun-to-verb assignment (parsing: Platero 1973 (179), pp. 83–94, newly opened). Tasks T31a–h.
DONE 2026-10-17 (8:46pm) (relative as CP). F10: restate the relative as CP under a D-like enclitic; decide raising vs matching on the island and “missing noun” data. Result: Platero 1973 Ch. II (pp. 14–55, page images) supplies the islands. The relative tie reaches any depth through direct-discourse complements and a -go complement ((55), (78)), but not into one conjunct (*(80), *(83)) or into a relative inside a relative (*(90), *(91) vs (88)); -ígíí nominalized clauses are doubtful, with speakers split (?(96), ?(100)). The pattern is the same whether the noun is inside or after the relative (p. 41), and coreference ignores it ((115), (117)). Restated: [DP [CP Opρ [K … ] Crel] -ę́ę/-ígíí]. A silent copy of a pronoun slot moves to Spec,CP (OP-ρ); the noun never moves ((12): it stays mid-clause; often there is none). Island = a CP under a determiner (ISL-D; Platero (104) as drawn in (105)) or a conjunct. Only slots are accessible (ACC): subject, object, the postposition object, the possessor; never the bare noun of an enclitic phrase (*(110) vs (111)). Raising vs matching: the islands show movement rather than coreference, so plain binding fails, but they do not separate raising from matching. The “missing noun” data rule out overt noun raising. Matching with the enclitic as a determiner mis-orders a noun after the relative (§6), so matching survives only with REL as C. Adopted: raising of the slot (H&P's “only the determiner raises”); matching is the runner-up. The decisive reconstruction data are unprinted (T34a). New tasks T34a–i.
OPEN Read Diesing & Jelinek 1995 and any later Jelinek paper for a PAH restatement in a TP structure (the “minimalist reanalysis of Jelinek” question in T26g)
Diesing & Jelinek 1995 (partly read); later Jelinek papers
T27o
DONE 2026-10-18 (4:46am) (Willie's framework). Read Willie 1991 Ch. 1–2 on page images for any trees or an explicit statement of her framework (the OCR layer is noisy). Result: pp. 10–51 read on page images (plus pp. 71, 92–93, 199–200, 220–221). No tree in Ch. 1–2. An automatic slanted-line scan of all 243 pages, calibrated on W&J pp. 267–268, found none elsewhere (weak; T35h). The framework is stated outright: Jelinek's pronominal-argument proposal (“nominals … as adjuncts to the sentence, occupying A′ positions, while the pronominal prefixes on the verb occupy the A positions”, p. 22), in Government-and-Binding terms (“These incorporated pronouns leave a trace in argument positions. The pronouns and trace form a chain that receives the theta role and case assigned by the verb”, p. 25; “the resulting chain … I am calling a pronoun”, p. 221). Its consequences: nouns are not case-marked; subordinate clauses are adjoined, except indirect-speech complements; “no movement to and from A positions” (pp. 22–23). No S/IP/VP labels, and no statement of where the subject is first merged. Class (b) confirmed: treeless GB inside a pronominal-argument typology. Restated (WF-1 to WF-9): the chain = the site's slot ρ. The three consequences follow from nouns entering only after D1 is spelled out (WF-5). Patient-before-agent order, including the agentive-passive ʼadi- (p. 36), follows from CYC if subject exponents are on a head (WF-6). Open clashes: yi-sh-kʼá (4a), yiiłtsą́ (10) vs p. 28 and Hale 1997 n. 9 (T35a); reflexive before da- (14) but reciprocal after it (18) (T35b). New tasks T35a–i. Follow-up 2026-10-18 (12:46pm):T35b done on reflexive vs reciprocal. The reflexive is two pieces (Kibrik 2005 < Young & Morgan 1987:70–71), so CYC derives both orders. The passive's ʼadi- is the deictic indefinite subject plus the shared di-, which puts WF-6 back on the table (T36g).
Willie 1991 pp. 10–51, 71, 92–93, 199–200, 220–221 (opened); Hale 1997 n. 9
T27p
OPEN Add Baker's trees (16), (19) to the T26g comparison and ask whether Navajo nouns could pattern with Kinande dislocated subjects (TP) or objects (CP); needs Navajo data, not a verdict from Kinande
Baker ms. pp. 14–17 (opened); Navajo judgments
7. Next steps
Do T27i first Done 2026-10-16, 4:46am: Condition C by site. A3 survives only as a timing condition (A3′); D6 stays compatible, undecided, and now turns on attachment height and on whether the subject pronoun moves (T28a, T28b).
Then T27j and T27k (rule sets A/B and the inverse), which are rewrites of our own rules. T27j done 2026-10-16, 12:46pm (A/B as spell-out); T27k done 2026-10-16, 8:46pm (the inverse as Agree and EPP). All F-row restatement tasks from T27i–k are now done; T27l–p remain.T27l done 2026-10-17, 4:46am (clause spine); T27m–p remain.T27m done 2026-10-17, 8:46pm (relative as CP); T27n–p remain.T27o done 2026-10-18, 4:46am (Willie's framework); T27n and T27p remain.
Open the unopened originals in the order Hale 1973, Barss et al., Young & Morgan, Rice 2000, Platero, Speas 1990 Ch. 4, Hale & Keyser; reclassify the (u) rows as each is read.
Do not claim any WWII-era source for the verb-centred remark until Young & Morgan 1943 or Sapir & Hoijer 1942 is opened.
Opened or re-checked for this page (2026-10-15, 9:34 to about 10pm PT): Jelinek 1984 and the Jelinek collection scan (box [local working copy, not in the repository]; W&J 2000 pp. 267, 269 = PDF 147–148; S&J 1989 pp. 345–352 = PDF 57–64; Diesing & Jelinek 1995 pp. 125, 134, 169 = PDF 280, 289, 324); Speas 1986 p. 89 (page image); Willie 1991 text layer and pp. 25, 71, 200; Hale & Platero 1974 and Creamer 1974 (ERIC extract); Witherspoon p. 63; Hale 1999 pp. 171–172 (page images); Hale 2001 and Hale 2003 (text extracts, box [local working copy, not in the repository]); Baker ms. pp. 14–17; Hale 1997 (ling-phil.mit.edu/papers/hale/papers/hale023.pdf, downloaded for this audit). Cross-links: open questions, sources, Jelinek & Baker attach, Hale derivation.
Which noun is which? Nominal adjunction: nouns attach to role slots, and the verb can overrule word order
2026-11-11 audit pass 2. Apostrophe fixes only. Navajo glottal stops typed as ' or ’ (for example in prefix names like ʼa- and ʼeʼe-, or words like Awééʼ) now print as ʼ (U+02BC), and a build step no longer turns the closing quote of an English translation inside a Navajo row into a glottal stop (‘…the girls.’, not ‘…the girls.ʼ). No form, gloss or translation wording changed.
2026-11-11 audit pass 4. Interlinear rows whose form and gloss do not line up one-to-one were re-checked against the source page. Each is marked † with a note under its table: SOURCED where the source prints it that way, OPEN where the source cannot confirm it. See the pass 4 report.
Restated on the unified structure (2026-11-15). Nouns are restated as arguments merged in their theta positions (VP, Spec,vP) and, for the subject, raised to Spec,TP by the working default; a lone noun is linked by the H4″ default. Adjunction and pronominal-argument analyses below are labelled contrast; where nouns sit for Condition C is to be re-run on the unified spine. Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §2 and §4, first rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.
This page reads the first two sections of Willie's (1991) Chapter 3, “The Third Person and Nominal Adjunction”: §3.0–3.1 (pp. 51–63) and §3.2, “Adjoining nominals to oblique arguments” (pp. 63–70). The site had used only the §3.3 inverse part of this chapter (pp. 71–100), and Willie's linking rule had come to the site mainly through Willie & Jelinek 2000:268. The test is a basic one for Grant's thesis. If the verb's arguments are role slots and not words, what decides which noun goes with which slot? Willie's answer, on pages opened this run, has three parts. (1) A noun “by itself cannot be an argument -- it must be licensed by a pronoun” (p. 52). The sentence is complete without it, and it is added “for reasons relating to the structure of the discourse” (p. 51). (2) By default, position decides. A lone noun goes with the object (17). With two nouns the inner one goes with the internal argument and the outer one with the external argument (20). With a postposition, the innermost noun goes with the postposition's object (31). (3) The verb's meaning can overrule position. ‘boy’ + ‘picked (round objects)’ means ‘the boy picked them’, not ‘he picked the boy’ (22). The run gives four results. (a) How many slots the verb has decides how a string of nouns is parsed. ‘[the old man's car] is ruined’ is one noun phrase. ‘[the old man] [his car] ruined’ is two, and the transitive verb also allows the one-phrase reading (§6, SCHEMA-PARSE). This supports Grant's thesis. (b) The default links are nested around the left edge of the verb-plus-postpositions unit (the MVS). The noun nearest the MVS goes with the argument whose prefix is leftmost in it (§8, MIRROR-LINK). Willie's own (36) breaks this nesting when meaning demands it. (c) Some overrides come from the verb's lexical meaning, and some come from knowledge about the world (“horses typically are led”). Dowty (1991:572, n. 16) separates these two, and only the first belongs to a role schema (§9). (d) Nouns attach to the MVS as a whole, never inside it, which bears on Willie's promise that the MVS “behaves as a unit under certain syntactic processes” (§10, MVS-UNIT).
This site writes a verb's arguments as a set of role slots, Σ(√) = {ρA, ρP, …}, each with Dowty entailments (proto-role schema). Prefixes value the slots inside the word (rule set A). Nouns are linked to slots at clause level (rule set B; schema C3, “each overt DP links to exactly one ρ, and its linear position is fixed by rule set B”). C3 was a HYPOTHESIS resting on Willie & Jelinek 2000 and §3.3. This page checks it against Willie's full account of how nouns are linked in the direct voice. It asks three things:
Phenomenon 2 (roles, not words). Are nouns really outside the argument structure? And when a noun's link is not fixed by order, is it fixed by the slot's entailments?
Phenomenon 1 (rule set A vs B). Clause-level order is head-final (nouns to the left), and the prefix order inside the word is fixed. How do the two orders line up, and around which boundary?
VPISH. Does the order of the nouns show anything about where the slots first Merge?
2. Sources opened this run (2026-10-11, ~8:55pm PT)
Willie, MaryAnn 1991.Navajo Pronouns and Obviation. PhD dissertation, University of Arizona (the scan on the box). Fresh 150 dpi page images of printed pp. 51–71 (PDF 57–77), i.e. all of §3.0–3.2 and the first page of §3.3. Page offset: PDF = printed + 6 for Chapter 3 (printed 52 = PDF 58, from the page headers). The printed number “51” occurs twice. PDF 56 is the last page of Chapter 2 and PDF 57 the first page of Chapter 3. That is why Chapter 2 runs at + 5 (as found on the pleonastic & psych page) and Chapters 3–4 at + 6. Every Navajo line of (1)–(36) was checked on 300 dpi crops, with extra zoom on yiyííbį́į́ʼ (22)–(23), the l/ł of the ‘ruin’ forms, shiiniltsą́ (9)–(10), and łííʼ (32), (36). No OCR was used for any form. The OCR layer was used only to find the chapter and section starts.
Dowty, David 1991. “Thematic proto-roles and argument selection.” Language 67(3):547–619. Page image of printed p. 572 (the stored 150 dpi image dpg-27.png): (27)–(28) and n. 16, which separates entailments of the predicate from inferences drawn from the whole sentence.
Transcription and glosses. Navajo as printed, with the glottal stop as ’ and the nasal hook as ą/į/ǫ. Willie's gloss lines are copied exactly, misprints included and marked [sic]. Her gloss order for verbs is object.subject (“3s.1s.ruined” = it, by me) and for postpositions object.P (“3.for”). Hyphens inside nouns (bi-chidí) are hers. Verbs are not segmented beyond her glosses.
Cited by Willie in these pages, not opened: Young & Morgan 1987:120 (“simple passive”, for (12)); Willie 1989 (Direct vs Inverse); Chomsky 1981.
3. Nouns are adjuncts, licensed by a pronoun (§3.0–3.1) SOURCED
P. 51: “In this chapter, I will describe the process whereby nominals are adjoined to the MVS, and how they are properly coindexed with the pronominal arguments.” Pp. 51–52: “The MVS is a complete functional complex, and by itself is a grammatical sentence. Nominal expressions are always added to the MVS for reasons relating to the structure of the discourse, not for reasons of grammaticality. When for discourse reasons the speaker desires to remove any potential ambiguity from the reference of a pronominal argument, a coreferential noun may be included in the utterance, and these nominals are in adjunct positions.” P. 52: “The noun by itself cannot be an argument -- it must be licensed by a pronoun; but it forms a complex expression with the pronoun, since it modifies the pronoun.” P. 52 (§3.1.1): “Navajo is a verb final language, and adjoined nominals appear at the left of the MVS. Except for a few terms that refer to humans, nouns are not marked for number, and nouns and free pronouns are never marked for case, since these features are marked in the incorporated pronouns.” Pp. 69–70: “nouns in Navajo are not in argument positions, where they are directly governed and assigned case by the verb or postposition, but are linked via predication to the incorporated pronouns, like "discontinuous" relatives. While a noun in and of itself cannot fill an argument position, it forms a complex constituent with its pronominal head at the level of the interpretation of the sentence.” P. 70: “a nominal may be adjoined for each of the pronouns in the MVS.”
GRANT'S CONJECTURE matched by SOURCED. Grant's thesis is that Navajo verbs take theta roles, not lexical items, as arguments. These statements are the most direct support for it that the site has page-checked. The arguments are the incorporated pronouns, one per slot. A noun is an optional modifier of one of them and is never the argument itself. The site's wording is that the slots are role slots with Dowty entailments, while Willie's is that they are “pronominal arguments”. Whether these differ is T7 (open questions).
4. Free and emphatic pronouns (§3.1.1–3.1.2) SOURCED
Willie first takes the free pronouns. They “have a particular function, that of marking referential contrast or emphasis, and therefore do not appear in "unmarked" constructions” (p. 53). With a 1st- or 2nd-person argument, “only a free pronoun with the same "phi"-features (person and number) may be adjoined” (p. 53). That gives her principle (2): “A adjoined [sic] nominal must agree in semantic ("phi") features with the pronoun it is coindexed with” (p. 53). It is “fundamental to the interpretation to all adjoined nominals in Navajo” (p. 54).
Willie 1991:53–56, (1), (3)–(5), (7)–(10)
(1)
nihí
yiicha
we
2d.crying [sic?]
‘We (are the ones who) are crying.’
(3)
Tʼáábí
Mary
yiztał
3.EMPH
Mary
3.3.kicked
‘He himself kicked Mary.’
(4)
Tʼáábí
ʼádílcha
3 EMPH
self-3-cry
‘He made himself cry.’
(5)
Tʼáábí
yicha
3.EMPH
3.cries
‘He himself is crying. (on his own)’
(7)
tʼááshí · tʼáání · tʼáábí
tʼáánihí · tʼáánihí · tʼáábí
(two columns, no gloss line; “produced either by adding a particle tʼáá, meaning "just" or by a shortened form -a”, p. 55)
(8)
shí
bi-chidí
yílchxǫʼ
I
3.car†
3s.1s.ruined
‘I (am the one who) ruined his car.’
(9)
shí
shiiniltsą́
I
1s.2s.saw
‘I (am the one) you saw.’
(10)
*ní
shí
shiiniltsą́
SOURCED † Segmentation note (2026-11-11 audit pass 4): Willie 1991 (8) prints bi-chidí with the gloss “3.car”: the gloss joins with periods or colons pieces the form separates with hyphens, so the two lines do not match one-to-one in the source. Kept exactly as printed.
P. 56: “With transitives, it is possible to add a free pronoun that agrees in person and number with only one of the arguments.” “Constructions with two free pronouns are judged unacceptable, since the function of the adjoined free pronoun is contrastive reference.” P. 54: “It is the reflexive pronoun ʼádi on the verb that makes (4) a reflexive sentence, not the emphatic pronoun”.
HYPOTHESIS (ours) Two points for the site. First, a free pronoun is linked by its person value alone. In (8) shí ‘I’ can only go with the 1st-person slot, which is the subject, and in (9) only with the 1st-person slot, which is the object. That is PI-MATCH / SLOT-TYPE-LINK (pleonastic & psych §10), with Willie's (2) as its source. Order plays no part. Second, *(10) bans two free pronouns even though each would match a different slot. So the ban is not about slots. It is about one contrast per clause (T18h).
5. One noun, two nouns (§3.1.3–3.1.4) SOURCED
Willie 1991:57–62, (11)–(25)
(11)
(ʼashkii)
yicha
boy
3s.crying
‘(The boy) he is crying.’
(12)
yíchxǫʼ
3s.ruined
‘It is ruined.’
(13)
bi-chidí
yíchxǫʼ
his car
3s.ruined
‘His car, it is ruined.’
(14)
[hastiin sání bi-chidí]
yíchxǫʼ
man old 3POSS.car
3s.ruin PERF (p. 58); “3-3-ruined” when repeated on p. 60 [sic]
‘The old man's car is ruined.’
(15)
(naadą́ą́)
yíkʼą́
corn
3.1.ground
‘(The corn) I ground it.’
(16)
(bi-chidí)
yílchxǫʼ
(3.car)†
3s.1s.ruined
‘(His car) I ruined it.’
(18)
chidí
yiyíílchxǫʼ
car
3s.3s.ruined
‘He/she ruined the car.’
(19)
[hastiin sání]
[bi-chidí]
yiyííłchxǫʼ
man old
3POSS.car†
3.3.ruined
‘The old man ruined his car.’
(21)
[hastiin sání bi-chidí]
yiyííłchxǫʼ
man old 3POSS.car
3-3-ruined
‘He/she ruined the old man's car.’
(22)
ʼashkii
yiyííbį́į́ʼ
boy
3.3.picked round objects
‘The boy picked them. (c.f. berries)’
(23)
[ʼashkii bimá]
yiyííbį́į́ʼ
boy 3s.mother
3.3.picked round objects
‘The boy's mother picked them.’
(24)
ʼashkii
yiyííłtaʼ
boy
3.3.count plural objects
‘The boy counted them.’
(25)
ʼashkii
yóldon
boy
3.3.shooting
‘The boy is shooting at it.’
SOURCED † Segmentation note (2026-11-11 audit pass 4): Willie 1991 (16) prints (bi-chidí) with the gloss “(3.car)”: the gloss joins with periods or colons pieces the form separates with hyphens, so the two lines do not match one-to-one in the source. Kept exactly as printed.
SOURCED † Segmentation note (2026-11-11 audit pass 4): Willie 1991 (19) prints [bi-chidí] with the gloss “3POSS.car”: the gloss joins with periods or colons pieces the form separates with hyphens, so the two lines do not match one-to-one in the source. Kept exactly as printed.
P. 57: “The single argument of an intransitive may have an agent, patient or theme theta role.” Of (12): “Young and Morgan (1987:120) call this the simple passive. Only one nominal, coindexed with the patient subject, is allowed in these constructions.” P. 59: “The nominal in both these constructions must be coindexed with the object, since the subject is first person.” P. 59, (17): “When a transitive sentence has only one adjunct, that adjunct is interpreted as coindexed with the object (patient) argument.” P. 60: “When a transitive verb has two third person arguments, it is possible to add two nominal adjuncts. Such sentences are not common in ordinary discourse.” P. 60, (20): “When a transitive sentence includes two nominal adjuncts, the inner nominal is coindexed with the internal argument, and the outer nominal (by default) is coindexed with the external argument.” P. 61: “If the transitive verb describes an action which can only be performed by an animate subject, these semantic features of the verb can "override" the expected interpretation principle stated in (20).” “Since berries do not pick boys, this sentence poses no problems of interpretation, and the expected word order can be suspended. This applies to long complex nominals as well.” P. 62: “In (23), the initial string cannot be interpreted as two constituents, even though the verb is transitive, because of the semantic features of the verb.” “(22) through (25) have the interpretation of the single nominal being coindexed with the subject. This coreferentiality between the nominal and the pronominal follows from what we know about what humans can do, such as picking, counting, and the like.”
Two small points as printed. Willie's text says the override suspends principle (20), the two-noun rule, but (22) and (24)–(25) have one noun, so the rule actually overridden there is (17). And p. 62 has “the expected work order” [sic].
6. The schema parses the string HYPOTHESIS (on SOURCED data)
Willie sets (14) beside (19) and (21) on purpose (pp. 60–61). The nominal string hastiin sání bi-chidí ‘man old his-car’ is the same in all three:
Verb
Slots in Σ
Parse of the string
Reading
(14) yíchxǫʼ (intransitive)
1: ρTh
one constituent, “obligatory” (p. 60)
‘The old man's car is ruined.’
(19) yiyííłchxǫʼ (transitive)
2: ρA, ρTh
two constituents
‘The old man ruined his car.’
(21) yiyííłchxǫʼ (transitive)
2
one constituent, linked to the patient (p. 61)
‘He/she ruined the old man's car.’
(23) yiyííbį́į́ʼ (transitive)
2
one constituent only, because of “the semantic features of the verb” (p. 62)
‘The boy's mother picked them.’
SCHEMA-PARSE (ours). The number of slots in Σ, and what each slot can hold, sets an upper bound on how many constituents the clause-level string can be split into. With one slot only one parse is possible. With two slots both parses are possible, unless the slots' content excludes one, as in (23). For Grant's thesis this is the clearest case on the site so far. The nouns do not decide the structure. The role schema decides how the nouns are grouped. It follows from Willie's p. 70 statement (“a nominal may be adjoined for each of the pronouns in the MVS”) plus p. 60. We did not find it stated as a rule in her text.
7. Postpositions (§3.2) SOURCED
Willie 1991:63–70, (26)–(36)
(26)
ʼatʼééd
yichʼįʼ
yáłtiʼ
girl
3.to
3.speaking
‘He is speaking to the girl.’
(27)
tsé
yikááʼ
neezdá
rock
3.on
3.sat
‘He sat down on the rock.’
(28)
ʼashkii
yikááʼ
neezdá
boy
3.on
3.sat
‘The boy sat down on it./He sat on the boy.’
(29)
yá
yoolóós
3.for
3.3.lead [sing. +animate]
‘She is leading it for him.’
(30)
ʼashkii
ʼatʼééd
yá
néidííłtsʼin
boy
girl
3s.for
3.3.fist-hit
‘He hit the boy for the girl.’
(32)
łííʼ
yá
yoolóós
horse
3.for
3.3.lead [sing. +animate]
‘She is leading the horse for him.’
(33)
ʼashkii
yá
yoolóós
boy
3.for
3.3.lead [sing. +animate]
a. ‘He is leading the boy for her.’ b. ‘She is leading it for the boy.’ c. *‘The boy is leading it for her.’
(34)
ʼazdzą́ą́
ʼatʼééd
yá
yoolóós
woman
girl
3.for
3.3.lead [sing. +animate]
a. ‘He is leading the woman for the girl.’ b. ‘The woman is leading it for the girl.’ c. *‘The woman is leading the girl for him.’
(35)
ʼasdzáá
yiiʼ
yiyííłbéézh
woman
3.in
3.3.cooked
‘The woman cooked it in it.’
(36)
ʼashkii
ʼatʼééd
łííʼ
yá yoolóós
boy
girl
horse
3.for 3.3.lead [sing. +animate]
‘The boy is leading the horse for the girl.’
P. 63: “The interpretive rule given in (17) applies also to sentences where the MVS is composed of an intransitive verb and a postpositional phrase.” “the single nominal is coindexed with the object of the verb if there is one, or with the postpositional object if the verb is intransitive.” “If the semantic features of the noun are incompatible with its being an adjunct to the object of the postposition, by default it may be interpreted as an adjunct to the subject.” P. 64: “Grammatical postpositional phrases may be phonologically integrated into the MVS, or they may form the first word in the MVS”. “There are three argument positions in this sentence; we can add a nominal to any of them.” P. 65: “If the semantics of the verb permit a human goal, then the nominal that immediately precedes the MVS will be interpreted as an adjunct to the postpositional object”. (31): “In an MVS with one Postpositional Phrase, the innermost nominal is coindexed with the postpositional object in the Direct Construction.” P. 66: “If the innermost adjoined noun has semantic features that make for its being interpreted as an adjunct to the object of the verb, then the generalization given in (31) will be overridden.” Of (32): “This interpretation follows from the fact that horses typically are led, and do not lead people. Note that on this interpretation, the noun preceding the inflected postposition (yá) is not coindexed with it, and cannot be governed by it.” P. 67, of (33): “On (33a), the nominal ’ashkii is interpreted as coindexed with the object of the verb, but it is not in a position to be governed by the verb. Note also that this single adjoined nominal cannot be interpreted as a subject adjunct; I will return to this point in Sec. 3.2. below.” P. 68: “Principle (31) above accounts for the exclusion of the reading shown in (34c).” P. 69: “it is entirely a matter of the semantic features of the verb and the adjoined nominal whether the single argment [sic] is to be coindexed with the verbal or postpositional object, or neither.” “When the single nominal cannot be coindexed with either object that it precedes, because of the semantic features of the MVS, it must be coindexed with the subject.”
Points as printed. P. 67 promises a return to the subject point “in Sec. 3.2. below”, but p. 67 is already in §3.2, and we found no later return on pp. 67–70. (36) is the same sentence as Willie's (85) on p. 95 (inverse §2.5), but (36) spells ‘horse’ łííʼ with no nasal hooks, as (32) does, while (85) and (37) print łį́į́ʼ. ‘woman’ is ʼazdzą́ą́ in (34) and ʼasdzáá, with no hooks, in (35). The ‘ruin’ forms with a 1st-person or 3/3 subject have plain l in (8), (16) and (18) (yílchxǫʼ, yiyíílchxǫʼ) but ł in (19) and (21) (yiyííłchxǫʼ). Willie p. 39 says the transitive is marked by the classifier, and (26) there has ʼííłchxǫʼ. So plain l may be a misprint, but we leave it as printed (T18f).
8. Phenomenon 1: the default links are nested around the MVS edge HYPOTHESIS
SOURCED inputs. (i) Inside the MVS the exponents of the arguments are in a fixed order: “the patient always to the left of the agent, and the postpositional phrases leftwards of the direct arguments” (p. 50, P-LEFT on the pleonastic & psych page). (ii) Outside it, the default links are (17), (20) and (31).
Put the two side by side. Number the exponents from the left edge of the MVS inwards: e1 = the postposition's object prefix (if any), then the verb's object prefix, then the subject. Number the nouns from the MVS outwards: N1 is the noun right next to it. The default links are then:
MIRROR-LINK (ours). By default, the k-th noun counting outwards from the MVS links to the k-th exponent counting inwards from the MVS's left edge. The links are nested around that edge, so the clause-level order (rule set B) is the mirror image of the prefix order (rule set A plus the postpositions). Three things follow.
The pivot is the MVS edge (B3), not the verb word (B2). In (30) the postposition's object is e1, outside the verb word, and the nesting counts it. A nesting centered on B2 would link N1 to the verb's object, which is reading *(34c)-type. That is a second datum for T2, in the same direction as WORD-EDGE′: the slots are counted up to B3.
It is a default, and Willie's own data break it. In (36) the nouns are ‘boy girl horse’ and the links are boy ↔ subject (e3), girl ↔ PP object (e1), horse ↔ verb object (e2). So N1 ↔ e2 and N2 ↔ e1, and the links cross. Willie licenses this through (32): a horse is led and is not the one led for. (34b) is a milder case: N2 skips the unexpressed verb object and goes to the subject. So MIRROR-LINK describes the order nobody has to explain. It does not say what is possible.
What it does not show. The nested order is also what one expects if nouns are ordered by how high their slots are (subject highest, outermost). The two descriptions make the same predictions on these data. The (36) crossing is a counterexample for both. It shows that, for nouns, order is overridden by meaning, not by structure.
The segmentation in the diagram is schematic. Willie does not segment néidííłtsʼin here, and the position of e2 and e3 inside it is read off P-LEFT, not off this form. That yá is y- + -á follows her gloss “3.for”. The inverse is left aside here (T18d).
9. Phenomenon 2: when meaning overrules order, and which meaning
9.1 Dowty's distinction SOURCED
Dowty 1991:572, n. 16: “It is important here to distinguish entailments of the PREDICATE from what follows from any one sentence as a whole (e.g. entailments that may arise in part from NP meanings, etc.). For example, if Mary slapped John is true, and John is a normal human, then, slapping being the kind of action it is, we would conclude that John necessarily perceives something … But it does not follow that the direct object of slap is entailed to have the P-Agent property of sentience, since we can also felicitously say Mary slapped the table or Mary slapped the corpse. However, the object of awaken does have the P-Agent entailment of sentience, as is revealed by the anomaly of #Mary awakened the table/the corpse.”
9.2 Willie's overrides, sorted by Dowty's test HYPOTHESIS
Willie gives one cover term, “the semantic features of the verb” (pp. 61–62, 66, 69). By Dowty's n. 16 her overrides are of two kinds. (The sorting is ours. Willie does not cite Dowty.)
Example
What excludes the default link
Kind (Dowty n. 16)
Part of Σ?
(22), (23) yiyííbį́į́ʼ
The object must be “round objects” (gloss), and a boy is not one. Willie's reason is “berries do not pick boys” (p. 61).
Predicate property of ρTh (the classificatory stem). Willie's reason also uses the agent's volition (P-A (27a)).
Yes: a property of the slot
(24) yiyííłtaʼ
The object must be “plural objects” (gloss), and ʼashkii is one boy. Most nouns are unmarked for number, but “a few terms that refer to humans” are marked (p. 52), and ‘boys’ is ʼashiiké (Willie (63b), number & scope §2.4).
Predicate property of ρTh
Yes
(25) yóldon
Nothing in the gloss excludes ‘he is shooting at the boy’. Willie: “what we know about what humans can do” (p. 62).
Unclear: no predicate property is visible
OPEN (T18a)
(32) łííʼ yá yoolóós
“horses typically are led, and do not lead people” (p. 66). Both ‘it is led’ and ‘led for’ allow an animal, since the gloss is [sing. +animate] and grammatical postpositions allow animates (p. 65).
Sentence-level, world knowledge (“typically”)
No
(35) ʼasdzáá yiiʼ yiyííłbéézh
A woman is neither the thing cooked nor the container ‘in it’, so the noun goes to the subject (p. 69).
Mixed: ‘in’ needs a container (P property), and ‘cooked’ needs food (world knowledge, not entailed)
Partly
HYPOTHESISENT-LINK (ours). A noun can be linked to slot ρ only if it is compatible with the predicate properties of ρ, meaning its Dowty entailments plus the stem's sortal class (round, plural, [sing. +animate]). Among the compatible slots, the defaults (17)/(20)/(31) choose, and world knowledge can choose against them. Only the first step belongs to the role schema. Grant's thesis predicts the first step. It does not predict the second, and it does not need to, because linking nouns is not argument structure on Willie's account (p. 70: the link holds “at the level of the interpretation of the sentence”).
9.3 What cuts against Grant's thesis here
OPEN (25) ʼashkii yóldon ‘The boy is shooting at it’. If no slot property excludes ‘he is shooting at the boy’, then a linking choice that Willie states without qualification is not driven by the role schema. Either the schema has properties we cannot see in the gloss, or the linking is pragmatic (T18a).
OPEN *(33c) ‘The boy is leading it for her’. A boy can lead, and ρA of ‘lead’ (vol, sent) fits a boy perfectly, yet the reading is out. So compatibility with a slot's entailments is not sufficient for a link. Position also restricts. A lone noun before a PP + transitive verb may go with either object but not with the subject, unless both objects are excluded ((35)). Yet in (28), with an intransitive verb + PP, the lone noun ʼashkii can be the subject (‘The boy sat down on it’) even though nothing excludes the PP reading, and the text on p. 63 says the subject reading needs incompatibility (T18b). We do not resolve this.
10. The MVS as a unit (44d) HYPOTHESIS
SOURCED Willie's (44) (p. 50) lists four properties of the MVS. (d) is that “it behaves as a unit under certain syntactic processes (to be discussed in Chap. 3)”. The process Chapter 3 opens with is nominal adjunction “to the MVS” (p. 51), with nouns “at the left of the MVS” (p. 52). Two of its statements locate the nouns outside both the verb and the postposition. The noun before yá in (32) “is not coindexed with it, and cannot be governed by it” (p. 66). In (33a) ʼashkii “is not in a position to be governed by the verb” (p. 67). And grammatical postpositions never take a noun as object (p. 39, GPP/LPP).
MVS-UNIT (ours). Rule set B adjoins nouns to the MVS as a whole. It never adjoins them to the verb word or to a grammatical PP inside it, so no noun is ever between a grammatical PP and its verb. This is our reading of (44d). Willie does not say on these pages that adjunction is the process (44d) meant. No starred example with a noun inside the MVS is printed on pp. 51–70 (T18g). If MVS-UNIT holds, rule set B “sees” the MVS as one head: B3 is where clause syntax starts. Together with WORD-EDGE′ (Σ closes at B3) and MIRROR-LINK (the nesting pivots at B3), that makes three properties that converge on B3, against one (prefix linearization) at B2.
11. Rules
Notation as on the pleonastic & psych page. π(ρ) is a slot's value. Nk is the k-th adjoined noun counting leftwards from the MVS. ek is the k-th argument exponent counting rightwards from the MVS's left edge (P-LEFT).
ADJ-NOTARG A noun is never an argument: it "must be licensed by a pronoun",
is "linked via predication" to it, and is added for discourse,
not grammaticality. (pp. 51–52, 69–70)
ADJ-LEFT Adjoined nominals appear at the left of the MVS; not governed or
case-marked by V or P. (pp. 52, 66, 67, 69)
ADJ-ONE At most one nominal per pronoun in the MVS. (p. 70; (36))
ADJ-PHI An adjoined nominal agrees in "phi" features with its pronoun.
(p. 53, (2); (8), (9), (15), (16))
ADJ-FREE Free pronouns mark contrast/emphasis; at most one per clause.
(pp. 53, 56; (1), (8)–(10))
ADJ-OBJ One nominal, transitive verb: it links to the object (patient).
One nominal, intransitive verb + PP: it links to the PP object.
(p. 59, (17); p. 63, (26))
ADJ-TWO Two nominals: inner ↔ internal argument; outer ↔ external argument
"by default". (p. 60, (20))
ADJ-PP MVS with one PP (direct voice): the innermost nominal ↔ the PP
object. (p. 65, (31); *(34c))
ADJ-SEM "Semantic features" of verb, P and noun can override ADJ-OBJ,
ADJ-TWO and ADJ-PP. (pp. 61–62, 66, 69; (22)–(25), (32))
ADJ-SUBJ A lone nominal that fits neither object links to the subject;
with a PP + transitive verb it cannot otherwise be the subject.
(pp. 67, 69; *(33c), (35))
ADJ-AMB A lone human nominal before PP + transitive verb is ambiguous
between the two objects. (p. 67; (33))
Willie's statements and examples as cited. The rule names are ours. ADJ-OBJ and ADJ-TWO match INV-6 on the inverse page (there from pp. 72–73 and W&J 2000:268).
SCHEMA-PARSE. |Σ| and the slots' content bound the number of constituents in the adjoined string.
HYPOTHESIS
(14) vs (19)/(21); (23). §6.
MIRROR-LINK. Default: Nk ↔ ek (nested around B3).
HYPOTHESIS, default only
ADJ-TWO + ADJ-PP + P-LEFT. Broken by (36) and (34b). §8.
ENT-LINK. A link to ρ needs compatibility with ρ's predicate properties (Dowty entailments + sortal class). The defaults and world knowledge choose among compatible slots.
HYPOTHESIS
§9.2; Dowty 1991:572 n. 16. Not sufficient: *(33c). Unexplained: (25).
MVS-UNIT. Rule set B adjoins nouns to the MVS as a whole, never inside it.
HYPOTHESIS
(44a, d); pp. 51–52, 66–67; GPP/LPP. §10.
Schema C3 (schema): “linear position is fixed by rule set B”
HYPOTHESIS, now refined
Position fixes a default link only. ADJ-SEM and ENT-LINK can choose another slot.
12. Derivations L1–L8, step by step
Conventions as on the derivations page: Σ schema step, A word-internal linearization (prefixes left of the stem), B clause level (complements and adjuncts left of their head). V-P is the VPISH variant, in which the subject slot raises to Spec,M/SP inside the word. V-H keeps it in situ (Hale). Entailment labels are ours, from Willie's translations (Dowty 1991:572 (27)–(28)). Verbs are not segmented beyond Willie's glosses.
L1. (11) (’ashkii) yicha ‘(The boy) he is crying’ and (1) nihí yiicha
Σ Σ(√CRY) = {ρA}, one slot (vol?, sent; ours).
Merge ρA in Spec,v. Agree: π = 3 (gloss “3s.crying”). (V-P) Move ρA to Spec,M/SP. (V-H) no move. Same string.
A Spell-out yicha. The word closes (B2). No PP, so the MVS closes too (B3), and Σ is closed (WORD-EDGE′).
B Optional: adjoin ʼashkii at the left of the MVS (ADJ-LEFT). There is one slot, so the only link is ρA. ADJ-PHI: a noun counts as 3, and π = 3, so the link is allowed.
(1): π(ρA) is non-singular (gloss “2d” [sic?], translation ‘We’). B adjoin the free pronoun nihí. ADJ-PHI: its person and number must match the slot. ADJ-FREE: the reading is contrastive, ‘we are the ones who …’.
L2. (8) shí bi-chidí yílchxǫʼ, (9) shí shiiniltsą́, *(10) ní shí shiiniltsą́
Σ Σ(√RUINtr) = {ρA (vol, caus), ρTh (chg)}.
Merge ρTh in Comp,V and ρA in Spec,v. Agree: π(ρTh) = 3, π(ρA) = 1 (gloss “3s.1s”). (V-P) Move ρA to Spec,M/SP.
A Spell-out yílchxǫʼ (as printed; T18f). B2 = B3, so Σ closes.
B Adjoin bi-chidí, then shí, both on the left. Linking is by π, not by order. shí is 1, so it goes to ρA. bi-chidí is 3, so it goes to ρTh (ADJ-PHI; Willie p. 59 for (16)). Here the order (N2 = subject, N1 = object) happens to match MIRROR-LINK, but nothing depends on it.
(9): π(ρTh) = 1, π(ρA) = 2 (“1s.2s.saw”). shí links to the 1st-person slot, which here is the object: ‘I (am the one) you saw.’ So the same free pronoun goes with a different role in (8) and (9), and the role is decided by the value of the slot.
*(10): ní would link to ρA (2) and shí to ρTh (1). Both links are allowed by ADJ-PHI and ADJ-ONE. The clause is out by ADJ-FREE: there can be only one contrast.
L3. One string, two parses: (14) yíchxǫʼ vs (19)/(21) yiyííłchxǫʼ
Merge / Agree: (14) π(ρTh) = 3. (19)/(21) π = 3, 3 (“3.3”). (V-P) Move the subject slot (ρTh in (14), which is a patient subject, Willie p. 57; ρA in (19)) to Spec,M/SP. Prefix order is unchanged.
A Spell-out yíchxǫʼ / yiyííłchxǫʼ. Σ closes.
B (14): there is one slot, so ADJ-ONE allows one adjunct. The string must be parsed as one DP, [hastiin sání bi-chidí], and it links to ρTh. The bi- inside the DP refers to the old man, DP-internally.
B (19): there are two slots. Parse as two DPs. ADJ-TWO: inner [bi-chidí] ↔ ρTh, outer [hastiin sání] ↔ ρA: ‘The old man ruined his car.’ (21): parse as one DP, and ADJ-OBJ links it to ρTh: ‘He/she ruined the old man's car.’ SCHEMA-PARSE: the verb's |Σ| sets which parses exist.
Σ (18): {ρA, ρTh (chg)}. (22): {ρA (vol), ρTh: [round objects]}. (24): {ρA (vol, sent), ρTh: [plural objects]}. The sortal classes are from Willie's glosses.
Merge / Agree π = 3, 3 in all three. (V-P) Move ρA. A Spell-out. Σ closes.
B Adjoin one noun. ADJ-OBJ proposes ρTh. (18): chidí ‘car’ is compatible, so the link holds: ‘He/she ruined the car.’
B (22): ʼashkii is not a round object, so ENT-LINK blocks ρTh. The only other slot is ρA, and a boy is compatible with it, so ADJ-SEM gives ‘The boy picked them.’ (24): one boy is not “plural objects”, so the same reasoning gives ‘The boy counted them.’ (23): the same block applies to the two-DP parse, since [ʼashkii] [bimá] would put ‘mother’ in ρTh. So only the one-DP parse is left (SCHEMA-PARSE + ENT-LINK).
L5. (30) ʼashkii ʼatʼééd yá néidííłtsʼin ‘He hit the boy for the girl’
Σ Σ(√FIST-HIT) = {ρA (vol, caus), ρP (aff)} + Obl-á(ρBen) (sent?; ours). Under WORD-EDGE′, ρBen is part of Σ through the grammatical PP (T17d).
Merge P = -á ‘for’ with ρBen as its object. Agree: π(ρBen) = 3, spelled y- (Willie: “3s.for”). Ayá.
B-ordered, inside Σ: [yá] to the left of the verb word (P-LEFT). B3 closes, and Σ closes.
B Adjoin ʼatʼééd, then ʼashkii, at the left of the MVS. ADJ-PP: N1ʼatʼééd ↔ ρBen. The verb “permit[s] a human goal” (p. 65), so ENT-LINK is satisfied. N2ʼashkii ↔ ρP, as Willie's translation shows. MIRROR-LINK: N1 ↔ e1, N2 ↔ e2, nested around B3. Note that ADJ-TWO's “outer ↔ external” is not what happens here: the outer noun is the hit boy, not the hitter.
L6. (33) ʼashkii yá yoolóós: two readings, and not a third
B Adjoin one noun, ʼashkii. ENT-LINK: a boy is compatible with all three slots.
ADJ-AMB: it may link to ρP, giving (33a) ‘He is leading the boy for her’, or to ρBen, giving (33b) ‘She is leading it for the boy.’ So (31)'s “innermost ↔ PP object” is not forced for a lone noun.
ADJ-SUBJ: ρA is not available while an object is, so (33c) *‘The boy is leading it for her’ is out. Compatibility with the slot was not enough (§9.3).
L7. (34) ʼazdzą́ą́ ʼatʼééd yá yoolóós: a, b, *c
Σ, Merge, Agree, Move, A as in L6.
B Adjoin ʼatʼééd (N1), then ʼazdzą́ą́ (N2). ADJ-PP: N1 ↔ ρBen in both licit readings.
N2 may link to ρP, giving (34a) ‘He is leading the woman for the girl’ (nested: N2 ↔ e2). Or it may link to ρA, giving (34b) ‘The woman is leading it for the girl’ (N2 ↔ e3, skipping the unexpressed e2; still non-crossing).
*(34c) ‘The woman is leading the girl for him’ needs N1 ↔ ρP. That violates ADJ-PP, and Willie says so (p. 68). Contrast (32)/(36), where N1 ↔ ρPis allowed because the noun is a horse. So ADJ-PP yields to ADJ-SEM only when the noun does not suit the PP object. A girl suits both.
L8. (35) ʼasdzáá yiiʼ yiyííłbéézh and (36) ʼashkii ʼatʼééd łííʼ yá yoolóós
Σ (35): Σ(√COOK) = {ρA (vol, caus), ρTh (chg)} + Obl-ii’ ‘in’ (container). Willie's gloss “3.in” has a pronominal object, so we treat it as grammatical. The p. 39 lists were not re-checked for -ii’. Merge / Agree 3, 3, 3. (V-P) Move ρA. Ayii’ + yiyííłbéézh.
B (35): one noun, ʼasdzáá ‘woman’. ADJ-PP proposes the PP object, but a woman is not a container, so the link is blocked. ADJ-OBJ proposes ρTh, and a woman is not what is cooked, so that is blocked too (world knowledge, §9.2). ADJ-SUBJ: link to ρA. ‘The woman cooked it in it.’
B (36): three nouns, one per slot (ADJ-ONE). N1łííʼ: ADJ-PP proposes ρBen, and ADJ-SEM (the (32) reason) overrides that, so it links to ρP. N2ʼatʼééd ↔ ρBen. N3ʼashkii ↔ ρA. The links N1 ↔ e2 and N2 ↔ e1cross, so MIRROR-LINK is overridden (§8). Same string and reading as (85), p. 95.
13. VPISH
HYPOTHESIS Three points.
Nouns never sit where VPISH puts subjects. On Willie's account (ADJ-NOTARG, ADJ-LEFT) no noun is ever in Spec,vP or in a raised subject position. Nouns are outside the MVS and “not in a position to be governed by the verb” (p. 67). So VPISH, if it applies, applies to the slots and their prefixes inside the word (V-P), never to nouns. That is the PAH reading the site has used since the derivations page. Chapter 3 states it outright.
Noun order cannot tell V-P from V-H. MIRROR-LINK puts the subject's noun outermost. That is what one expects if the subject slot is the highest, and it is the highest in both V-P (after raising) and V-H (Spec,v above Comp,V). And (36) shows that noun order can be overridden by meaning, so it is not a reliable readout of slot height anyway.
The patient-subject intransitive. (12) yíchxǫʼ ‘It is ruined’ has a patient subject (p. 57), which in VPISH terms is an internal argument promoted to subject. One noun links to it ((13), (14)). Nothing on these pages shows whether the slot moves. The prefix gives only a 3, and noun order has one noun. This is the unaccusative question again (plain ł-causer), with no new test.
14. Open questions
ID
Question
What would settle it
T18a
(25) ʼashkii yóldon is read ‘The boy is shooting at it’. What excludes ‘He is shooting at the boy’? If nothing in the verb's slots does, the link is pragmatic, and one of Willie's override cases is not schema-driven.
A speaker's judgment on the object reading; other sources for yóldon (Y&M 1987). OPEN
T18b
*(33c): a lone noun cannot be the subject of PP + transitive verb even when it suits ρA. In (28), with PP + intransitive verb, it can (‘The boy sat down on it’), although p. 63 says the subject reading needs incompatibility with the PP object. Is the difference transitivity? And why does p. 63 illustrate “incompatible” with (27), where the noun is the PP object?
More lone-noun sentences with PP + intransitive verbs; Willie's promised return “in Sec. 3.2” (not found on pp. 67–70). OPEN
T18c
Is an override from world knowledge ((32) “horses typically are led”) grammar or processing? Dowty's n. 16 test: does the default reading come back in a context where a horse could be the beneficiary?
Speaker judgments in context; texts. OPEN
T18d
Does MIRROR-LINK hold in the inverse? On the inverse page, INV-6 links the noun next to the verb to the internal slot, which under bi- is ρA. In (86) ʼashkii ʼatʼééd łį́į́ʼ bá yoolóós ‘The girl is leading the horse for the boy’, the order is PP object, subject, object (inverse §2.5). Is that nesting with the internal/external labels swapped, or a different pattern?
Where bi- sits in the template; Willie §3.3 re-read with this question (pp. 94–96). OPENPartly answered 2026-10-13: no. In Willie's Ch. 8 (13b) John háísh bichʼįʼ yáłtiʼ the noun next to the inverse postposition is the speaker, and the outer noun goes with the postposition's object (wh-questions §7, MIRROR-LINK′).
T18e
(1) nihí yiicha is glossed “we 2d.crying” and translated ‘We …’. Ch. 7 (2d)–(2e) give nihi- as both “1-dual” and “2-dual”. Is “2d” a misprint for 1d, or a reading ‘you two’ that the translation does not show?
Y&M paradigms for ‘cry’ (dual); NLA Verb Modes (dictionary: cry). OPEN
T18f
‘ruin’ with an agent is printed with plain l in (8), (16), (18) and with ł in (19), (21) and p. 39 (26). Misprint, or a real difference between the 1st-person and 3/3 forms? Also, (14) is glossed “3s.ruin PERF” on p. 58 but “3-3-ruined” on p. 60, a transitive gloss for an intransitive verb.
Y&M 1987 for -chxǫʼ. OPEN
T18g
MVS-UNIT predicts that no noun can stand between a grammatical PP and its verb. Is *N [P-obj-P] N [V] (a noun inside the MVS) actually starred? (44a) says the MVS “cannot be interrupted”, but no starred example is printed on pp. 51–70.
A starred example (Willie elsewhere; W&J 2000); a speaker. OPEN
T18h
*(10) bans two free pronouns even though each matches a slot. Is the limit one contrastive focus per clause, tied to W&J's focus position “immediately preceding the verb” (2000:268)? In (8) the free pronoun shí is not immediately before the verb.
Free pronouns in other positions; W&J 2000 §11 re-read. OPEN
Fidelity notes. Forms are as printed. (36) łííʼ vs (85) łį́į́ʼ; (34) ʼazdzą́ą́ vs (35) ʼasdzáá; plain l vs ł in the ‘ruin’ forms; (14)'s two glosses. (4) has “3 EMPH” without a period, while (3) and (5) have “3.EMPH”. Willie's “A adjoined nominal” (2), “interpretation to all adjoined nominals” (p. 54), “expected work order” (p. 62), “single argment” (p. 69), “pronouns, like” spelled “proncuns” (p. 70, a scan artifact or misprint), “Sec. 3.2. below” (p. 67, written inside §3.2) and the unclosed parenthesis after “horse"” (p. 69) are all as printed. (7) has no gloss line, and its two columns are not labelled. We read them as singular | non-singular but do not print that as Willie's. Printed page 51 occurs twice in the scan (§2).
Not quite a word: is the Navajo verb's argument a feature bundle?
2026-11-11 audit pass 3. Status badges now use only the label set SOURCED, HYPOTHESIS, SPECULATIVE, OPEN, NONE and GRANT'S CONJECTURE; verdict or task words that had been styled as badges (for example Opened, PRE-VPISH, DONE, SUPPORTS, PARTLY ANSWERED) are now plain text or a label with a qualifier. See the pass 3 report.
Restated on the unified structure (2026-11-15). The framework is restated in §1: VPISH (arguments merged in the verb’s scope, subject in Spec,vP) with theta roles, and V→v→T with the subject raised to Spec,TP as the working default in tensed clauses; each pattern is what else moves, with an LF whose scope matches c-command. Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). Flat-S and other pre-VPISH notation below is labelled contrast. Status on the spine: see unified structure, §1. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.
Grant's intuition is that the Navajo verb doesn't take an ordinary noun as its argument. It takes a theta role, a feature bundle, or a “proto-word”, something that is not quite a word. This page states that as a precise hypothesis, compares it with the literature on pronouns as feature bundles, and tests it against the configurational objections. Written 2026-10-14 (PT).
Ground rules for this page (Grant's). Nobody has fully worked out Navajo arguments. Each existing analysis (Jelinek, Willie, Hale, Speas, Baker) may be only partly right, and none of them is treated here as the judge. For each account we show what it gets right and what it misses, and the hypothesis keeps the parts that work. A source that does not state the hypothesis is not evidence against it. Only data that contradict it count against it.
Bottom line.HYPOTHESIS FB (below) survives every Navajo datum on the site. It also absorbs the strongest configurational point: Speas's own analysis of a pronominal-argument language (Warlpiri) already says that only pronominal clitics discharge the verb's argument positions and that nominals are modifiers (Speas 1990:167–171). The live dispute is narrower than “roles vs words”. It is whether a Navajo noun links to its slot inside the verb's projection (Speas) or outside it (Jelinek, Willie). FB is compatible with both on most data. Condition C, extraction and quantifier scope would decide, and the Navajo chapter of Speas 1990 (pp. 214–281), with its relative-clause argument (pp. 221–237), is still unread. So the hypothesis is not refuted, but it has not yet been tested against her actual Navajo data.
Update 2026-10-14 (8:46pm): Speas's Navajo data now checked.HYPOTHESIS Her 1986 dissertation (Ch. 2 §2.2.9, Ch. 3, Ch. 4; page images) is read datum by datum on Speas 1986 on Navajo. Result: no datum contradicts FB (0 CUTS AGAINST, 3 SUPPORTS, 15 COMPATIBLE). Her two “nouns are arguments” tests, no weak crossover (2.95) and multiple questions (2.96c), target an A-bar-adjunct version of the PAH and do not bear on FB. Her Condition C data, (4.15a) and (4.119), lean toward FB, and she concedes that Navajo nominals “behave like adjuncts insofar as they show ATB effects” (p. 323). FB has to borrow two positional rules from Platero and Speas: FB-3a LINK-NEAR and FB-6 PAR (§6 there). The 1990 revision of her chapter is still unread, so FB-a is partly done.
Update 2026-10-15 (4:46am): Hale & Platero 1974 checked.HYPOTHESIS Their relative-clause and anaphora paper is read on page images on Hale & Platero 1974. No datum contradicts FB's core (0 CUTS AGAINST, 1 SUPPORTS, 16 COMPATIBLE). One add-on rule does change: the positional linking rule FB-3a(a) (LINK-NEAR) is a default, not a hard rule. Their parsing principle (44) is applied “with varying degrees of strictness”, and some speakers let a noun skip the nearest verb in context ((41)–(42)). SOURCED Their (28) already states FB-3a(b), that the noun next to the verb is the object under yi- and the subject under bi- (p. 17). Their determiner analysis of the relativizer (p. 27) fits FB: what a relative must contain is a pronoun-like element, and its noun is optional (§5). New on that page: FB-7 SURF-LINK (linking checked on the surface string, as in their architecture (50)).
Update 2026-10-15 (12:46pm, corrected 1:07pm): Hale 1999 “Conflicting Truths” checked (page; step-by-step derivation and assumption audit). HYPOTHESISA conditional tension, not a refutation. Hale's (10) and (11a) report that a null subject cannot corefer with a noun inside its theme phrase or inside a clause under ‘about it’, with two people and so no rank confound (pp. 172–173). FB-5 said a bundle never triggers Condition C on a linked noun. That is in tension with (10) and (11a) if ten assumptions hold (A1–A10: Condition C is what excludes the reading, the judgments are stable, an adjunct is out of the subject's reach, and so on). None is independently established for Navajo, and Speas's (4.119) has the opposite judgment on a sentence of the same shape. The 12:46pm wording (“that is wrong”, “the answer is now inside”) was too strong. FB-1 to FB-4 are untouched. FB-3c LINK-IN, FB-5′ and FB-6′ are conditional additions, needed only if the assumptions hold. D6 below is compatible, undecided.
FB-1 An argument of V (or of a grammatical P) is a bundle
β = [π person, # number, κ capacity/animacy rank, ρ role]
realized as a pronominal prefix (or ∅ for 3rd person).
FB-2 ρ is fixed by Σ(√) (the verb's proto-role schema); π, #, κ are
the referent's features; β is complete without any noun.
FB-3 A noun N carries features [π=3, #?, κ(N)] and no ρ. N links to
exactly one β whose features it matches (π, #, κ, and the stem's
sortal class on ρ); position and topicality choose among matches.
FB-4 The voice (yi-/bi-) compares κ of the two bundles: the higher-κ
bundle is external; bi- marks external ≠ ρA.
FB-5 β is a φ-level object (variable-like), not a D-level one (referential
R-expression); N may be D-level.
Clause
Status
Basis
FB-1: the argument is a prefix carrying person and number
SOURCED
Hale's Pronominal Argument Theory: “Person-Number morphology in the Navajo verb word has the status of full argument” (handout (16), p. 7); Willie 1991:51–52
FB-1: the bundle also carries κ and ρ
GRANT'S CONJECTURE formalized as HYPOTHESIS
Grant's “schema of theta roles … prior to realization as a lexical item” (project chat, message 1)
FB-3: nouns link by matching features
SOURCED in part
Willie's ADJ-PHI: “A adjoined [sic] nominal must agree in ‘phi’ features with its pronoun” (p. 53). Sortal match: (22) “round objects”, (24) “plural objects” (pp. 61–62). Position defaults: (17), (20) (pp. 59–60)
FB-4: voice compares rank
SOURCED as a description; κ as a feature of the bundle is HYPOTHESIS
HYPOTHESISWhat “not quite a word” means here. In D&W's terms (§2), a φP is “neither D syntax nor N syntax”. It “simply spell[s] out φ-features”, and it is a variable rather than a referring expression (2002:410–411). That gives Grant's “proto-word” a precise sense. The verb's argument is a φ-level bundle, smaller than a noun phrase (DP/NP), which the verb word can take in fully built. The noun is a full phrase that has to be linked to the bundle from outside.
2. Feature bundles in the literature
Source
What it says
Access
Use for FB
Jelinek 1984 (PAH)
Pronominal clitics/affixes are the arguments; nominals are adjuncts (Warlpiri tree (28), 1984 pre-VPISH notation: CPP sisters of Verb–AUX under S)
Opened 2026-10-15 (8:46pm) for attachment (page); earlier via Hale, Speas 1990:166
FB-1 core; no verdict on A3 (notation cannot be translated directly; T26g)
Hale (handout; 2003 ms.)
Person-number morphology is the full argument, “not ‘agreement’”. Category of true arguments: “D, an X-zero and X-max type” (handout (16)). An independent pronoun is a “contrastive adjunct” (ms.:5–6)
Opened
FB-1 sourced. Note Hale says D, where FB-5 says φ (§3)
Déchaine & Wiltschko 2002, “Decomposing pronouns,” LI 33:409–442
Three proform types. Pro-DP: argument only, definite, Condition C (R-expression). Pro-φP: argument or predicate, a variable (Condition B), “simply spell[s] out φ-features”. Pro-NP: predicate, constant (pp. 409–411, table (2)). The D/φ/N split “cuts across” phrasal pronouns, clitics and agreement (n. 1; §6). Obviation and different-subject marking are D-agreement, which obeys Condition C (§5)
Opened (Georgetown copy)
Gives FB-5 a typology and a binding prediction (§6, D6)
Baker 1996
Morphological Visibility Condition: every argument must be visible on the head by agreement or incorporation
Secondhand
FB-1 is a version of it. But Navajo lacks verbal noun incorporation (Hale ms.:13)
Preminger (e.g. 2014, Agreement and Its Failures)
φ-agreement as a probe that can fail without crashing; tests for distinguishing clitics from agreement
OPEN(cited, not opened)
His clitic-vs-agreement tests could tell whether Navajo prefixes are bundles in argument position (clitics) or agreement (Speas). Not yet applied
Classificatory stems
Stem choice encodes the Theme's shape, rigidity, animacy and number
Grant's Notes §3; Willie's glosses (22), (24)
The patient bundle's sortal features are built into the stem: FB-3's sortal match
3. Each account: what it gets right, what it misses
Account
Gets right (FB keeps)
Misses (on data, or not addressed)
Jelinek 1984 (PAH)
Prefixes are the arguments. Nouns are optional and added for discourse (D1–D2 on the thesis page). 8:46pm: Warlpiri attachment is S-level (T25c page)
Predicts free order of adjuncts. Navajo order is fixed by default (Willie (20), *(33c)). Rescued only by “discourse configurational” (W&J 2000). Paper is Warlpiri-centred; no Navajo Condition C
Willie 1991
Nouns are licensed by a pronoun and linked by φ-match and position (pp. 51–70). The inverse is chosen inside the word with no nouns present (pp. 86–88). da- follows the agent (pp. 82–84)
Some links rest on world knowledge (“horses typically are led”, p. 66) or on nothing visible ((25), p. 62). The inverse agent is stated as ergative, where W&J 2000 keep a constant NOM gloss (T7)
Hale (1973 via Willie; 2003)
The PA Theory statement. Bare nominal arguments of postpositions are in base position (ms.:51). The Condition C problem (Perkins) as a test
Calls the true arguments category D. That would make them R-expressions in D&W's sense, which sits badly with variable-like uses (fourth person, wh-binding: Willie p. 201). Leaves Condition C open himself
Speas 1986/1990
Navajo has ordinary hierarchical structure and subject/object asymmetries. “There is no one parameter responsible for nonconfigurationality” (1990:195). The prefixes must be infixes, accessible to syntax (1990: Ch. 4 §4.5.2, p. 247; 1986: Ch. 3, pp. 200–202, opened). Her own Warlpiri analysis says “Verbal grid positions can only be discharged by pronominal clitics” and nominals (KPs) are modifiers linked to a grid position by theta-identification (pp. 167–170, (73))
Checked 2026-10-14 (8:46pm) against the 1986 Navajo chapters (page): she declines the option that the prefixes are arguments that move into the verb at PF (1986:201) on two theory-internal grounds plus two Ch. 2 tests that do not bear on FB. Her (4.15d) Condition C case is confounded by rank. The abstract says overt nominals “behave like arguments” by standard tests (1986). How that squares with Willie's “a noun by itself cannot be an argument” (p. 52) is the open question (§5)
Baker 1996
Arguments must be visible on the head. The CED and quantifier predictions make the adjunct analysis testable
Built on Mohawk. Navajo lacks verbal noun incorporation, so Baker does not classify it as polysynthetic
Creamer 1974
The ranking is a noun's “capacity to act upon” others (p. 30), and “the nature of the verb plays a secondary role” (p. 29)
Verb-sensitive exceptions (lightning with ‘strike’, p. 33; the group-7 movement rule, p. 36) show that the capacity is read in the event, not just off the noun
HYPOTHESISWhat FB keeps from each. From Jelinek, Willie and Hale: the argument is the prefix, and the noun links to it. From Speas: the bundle sits in a real argument position in a hierarchical structure (Spec,v / Comp,V, as in this site's derivations), and noun linking is theta-identification with a grid position, her Warlpiri mechanism. From D&W: the bundle is φ-level. From Creamer and Dowty: κ is a capacity feature. From Willie: linking by φ, sortal class and position.
4. The capacity feature κ (from Creamer 1974)
SOURCED Creamer: higher status comes from “capacity for having intent or purpose; … intelligence; … strength, vigor, aggressiveness, or special potency; … usefulness to man, or relatedness to man; and … animation, or capacity for movement”, “summed up in an inherent ‘capacity to act upon’ in relation to other beings” (1974:30). Equal rank means “either noun may be chosen … The subject chosen by the speaker is the object of major interest or emphasis” (p. 32). A newborn ranks below other humans “only as long as the infant retains the helplessness characteristic of the newborn” (p. 33). Personified nouns are treated as equal in status to whatever they interact with (p. 37). Full data on eight ranks.
HYPOTHESISWhy this fits a bundle. κ is a general capacity of the participant, not a per-event entailment count. That is shown by old age, which causes the death and still ranks below the tree ((c28), p. 37, and eight-ranks §6.3). A general capacity is exactly the kind of inherent feature a bundle can carry before any noun is present, and it is assigned by construal. The same word ‘baby’ gets a lower κ when it means a helpless newborn (p. 33), and a personified tree gets a raised κ (p. 37). So κ belongs to the bundle's referent as construed, not to the word. That is what Grant's “not quite a word” predicts: the verb ranks what fills its slots, and a noun's lexical identity does not fix the rank. OPEN κ is verb-sensitive in two places: lightning with ‘strike’ (p. 33), and the mover in group 7 (p. 36). FB-4 has to let Σ(√) adjust κ (movement entailed by the verb raises it). That is a constrained interaction, but it needs to be stated (FB-4′, §7).
5. The configurational objection, steelmanned
SOURCED(book opened; Navajo chapter not available in the copy we could read) Speas 1990 argues that “there are no configurationality parameters” (p. 2). Every language projects arguments by the same lexicon-to-syntax mapping, and Navajo's surface freedom comes from independent, lexically stated properties (pp. 195–198). Ch. 4 (pp. 202–281) covers subject/object asymmetries (§4.2.3, p. 209), null objects and null topics (§4.3.3, p. 217), relative clauses and the Parallel Function Constraint (§4.3.4, pp. 221–237), “On Navajo Nominals as Adjuncts” (§4.4, pp. 237–240), pronominal clitics as infixes (§4.5.2, p. 247) and bi- as an incorporated pronoun (§4.5.3, pp. 260–275). Through Hale & Platero’s negative-polarity ms. (secondhand; earlier labelled Hale & Perkins), her Parallel Function Constraint (p. 232) explains Perkins's Condition C facts without adjunction. Her 1986 abstract says overt nominals “behave like arguments” by standard tests.
Steelman. If nouns were adjuncts outside the clause, Navajo should look like Mohawk: free order, no subject/object asymmetries, no binding into or out of nominals. Instead it shows a fixed default order (Willie (20)), asymmetries (only the external slot raises; inverse definiteness, pp. 76–77), and the relative-clause facts. So the simplest grammar puts the nouns in argument positions, with the prefixes as clitics (or, for bi-, incorporated pronouns) linked to them.
HYPOTHESISWhere FB stands. Most of the steelman is compatible with FB. FB has hierarchical slots, so asymmetries and fixed default linking are expected (the bundles are ordered in Σ, and nouns link by position default). Speas's Warlpiri mechanism, nominals as modifiers theta-identified with a grid position that only a pronominal discharges (pp. 167–170), is FB-3 in her own vocabulary. She applies it to Warlpiri and, on the abstract's evidence, rejects it for Navajo. If the Navajo chapter shows nouns behaving as arguments on a test where Warlpiri KPs behave as modifiers, that would be contradicting data. Until it has been read, FB stands.
HYPOTHESISUpdate 2026-10-14 (8:46pm). The 1986 Navajo chapters contain no such test (datum table). Her final structure (4.124, p. 324) assigns the θ-role to a CHAIN of noun plus clitic. FB assigns it to the bundle and links the noun. No datum in the dissertation tells those two apart.
6. The argument diagnostics under FB
Diagnostic
Configurational prediction
FB prediction
Data so far
Verdict
D6 Condition C
Nouns in A-positions should show Condition C. Perkins's relative clause is explained by the Parallel Function Constraint (Speas p. 232)
The bundle is φ-level (a variable, D&W). It cannot c-command a noun as a binder because the noun is linked, not merged, in the argument position. So no Condition C between a prefix and a noun, as in Perkins's case (Hale handout (17))
Compatible with FB; also compatible with Speas + PFC. 8:46pm: Speas 1986 (4.15a), (4.119)–(4.121) show the same exemption, and her one Condition C exclusion is confounded by rank: leans SUPPORTS FB (Speas 1986). 2026-10-15 (12:46pm, restated 1:07pm): Hale 1999 (10), (11a,b) give an equal-rank case with i ≠ j, which is in tension with FB-5's exemption if assumptions A1–A10 hold; Speas's (4.119) has the opposite judgment. COMPATIBLE, UNDECIDED. LINK-IN, FB-5′, FB-6′ are conditional additions (page, audit)
Obviation-type disjointness
—
D&W: disjoint-reference morphemes are D-level. Prediction: the fourth person / bi- marking “someone different” should show Condition C-like disjointness
The fourth person means “someone different” in all four constructions studied (relatives, clause chains)
SUPPORTS (a D/φ split inside the bundle system)
D7 Extraction
Extraction from argument nouns possible (subject to islands)
Wh-words are adjunct-like and link to a bundle (Willie p. 201). No extraction out of a noun. No long-distance wh
No long-distance questions; no extraction from wh + noun (Willie Ch. 8)
True NP quantifiers with scope from their position
Scope is computed on bundles. A quantifier on a noun takes scope via its bundle, and da- (on the agent bundle) shows this
da- follows ρA in both voices (Willie 82–84). The scope of (58)–(59) is unstated
Partly supports; OPEN (T5c)
Possessor yi- binding
Lexical DPs in A-positions
Binding of possessor yi- by the bundle, not the noun
Unopened thesis argues A-position (abstract)
OPEN. Only its data would count
7. Derivations
F1. (8) shí bi-chidí yílchxǫʼ ‘I (am the one who) ruined his car’ (Willie 1991:56)
Σ Σ(√RUIN) = {β1[ρA], β2[ρTh]} HYPOTHESIS
Merge β2 Comp,V ; β1 Spec,v VPISH base positions
Value β1 = [π1, #sg, κ person, ρA] ; β2 = [π3, κ ?, ρTh] SOURCED (gloss 3s.1s)
Move V-P: β1 → Spec,M/S (portmanteau with mode, Hale ms.:10) HYPOTHESIS
A spell-out yílchxǫ' (complete clause) SOURCED (pp. 51–52)
B shí [π1] matches only β1 ; bi-chidí [π3] matches only β2 SOURCED (ADJ-PHI)
Result: linking by feature match; no position needed.
F2. (c28) Tʼiis są́ biisxį́ ‘The tree died of old age’ (Creamer 1974:37)
Σ {β1[ρ causer], β2[ρ undergoer]} HYPOTHESIS
Value κ(β2: tree) = group 7 > κ(β1: old age) = group 8 SOURCED (ranks, pp. 36–37)
FB-4 higher κ = β2 → external; β2 ≠ ρA-like causer → bi- SOURCED (bi- preferred)
B t'iis links to external β2 (first noun = topic) ; są́ to β1 SOURCED (p. 32 topic rule)
Result: rank is a capacity of the bundle's referent, not an
event-entailment count (old age causes but ranks lower).
F3. FB-4′, the verb's adjustment to κ (group 7, Creamer p. 36)
FB-4′ If Σ(√) entails movement of one participant, raise that bundle's
κ above a stationary co-argument of the same group. HYPOTHESIS
(Creamer: the moving noun "takes higher status than the more
stationary object", p. 36 — SOURCED)
8. Verdict
8:46pm (T25c), revised 9:20pm: Jelinek 1984's Warlpiri tree (28) is pre-VPISH notation (flat S: V, AUX, CPP; no VP or Spec,vP), so it gives no verdict on A3; the 8:46pm wording “supports A3 for that source” is withdrawn. Baker's later n. 16 contemplates PP/vP adjunction for Navajo (against A3). Attachment height for Navajo stays open; restating the PAH in a VPISH/TP/CP structure is T26g — page.
No Navajo datum on the site contradicts FB. It inherits the confirmed parts of Jelinek, Willie and Hale: prefixes as arguments, nouns linked by φ and sortal match, rank computed inside the word. It adds a precise sense for “not quite a word” (a φ-level bundle, D&W). It gives a home to Creamer's capacity-based ranking (κ). It can be stated inside Speas's own configurational machinery, as nominal linking by theta-identification. It survives Speas as far as we can currently check: her general framework and her Warlpiri analysis are compatible with it. But her Navajo-specific arguments (pp. 209–281) and Barss, Hale, Perkins & Speas 1989 have not been read, so “survives Speas” is provisional. 8:46pm update: the 1986 Navajo chapters are now checked, and no datum in them contradicts FB (Speas 1986 on Navajo). FB adds FB-3a (LINK-NEAR) and FB-6 (PAR), restated from Platero and Speas. The 1990 revision and Barss et al. 1989 remain. The tests that could contradict FB are Condition C minimal pairs, extraction out of nominals, and the scope of quantified nouns under the inverse.
9. Next steps
#
Step
Status
FB-a
Get Speas 1990 Ch. 4 (pp. 202–281): Internet Archive lending copy phrasestructurei0000spea, or a library. Check §4.3.4 and §4.4 test by test against FB
PARTLY DONE 2026-10-14 (8:46pm): the 1986 dissertation versions read (page); 1990 Ch. 4 OPEN (T22d)
FB-b
Barss, Hale, Perkins & Speas 1989, “Aspects of Logical Form in Navajo” (Cook & Rice, pp. 317–334): quantifier and binding data
OPEN
FB-c
Apply Preminger's clitic-vs-agreement diagnostics to Navajo prefixes (do they double, and can they fail?)
OPEN
FB-d
Test FB-5: can a 3rd-person prefix be a bound variable under a quantified noun (D&W's φ diagnostic)? Needs data
OPEN (speaker judgments)
FB-e
Read the possessor-yi- binding thesis for its data
OPEN
2026-11-10 · First-principles review 1: what the yi-/bi- examples actually require
2026-11-14: the “On the spine” summary at the top of this page supersedes this section’s framing. Its data, judgments and citations are unchanged.
Restated on the unified structure (2026-11-15). This section’s results now sit on the one structure for Navajo: unified structure, inverse and one-noun rows. Its restatement as derivations is in Results on the spine. Data, judgments and citations below are unchanged.
Speas 1986, Chapter 4, pp. 249–251; Willie 1991, p. 160. Research reset requested by Grant, November 10, 2026 (PT). NX-001–008.
HYPOTHESISPlain-words verdict. The same two nouns can exchange who sees whom when the verb changes from yi- to bi-. With just one noun, the reading is restricted, not freely ambiguous. Subjects starting inside the VP and having theta roles do not explain those restrictions by themselves. Below is a small, explicit candidate account, with its extra assumptions exposed. It needs a rule connecting noun order, unspoken participants and the verb form; it does not need a light determiner. These examples do not decide whether nouns move, whether prefixes are arguments or agreement, or whether light determiners exist. Telicity and the question/negation tests remain OPEN.
Only two starting assumptions
Assumed, by instruction: VPISH (subjects originate within the verbal projection) and theta roles. These are starting commitments, not findings of this review. HYPOTHESIS Everything more specific must be tested: DP or dP structure, light d, stationary versus moving d, null pronouns, head direction, position classes, agreement probes, EPP, Condition B/C, animacy rankings, Platero’s NP-PRO, Speas’s parallelism, the A/B ruleset split, and every earlier ID-, RD-, EH- or FB-rule. “Subject” here names the source’s grammatical relation; it is not automatically a synonym for Agent. “Inverse” does not automatically mean passive.
OPEN A uniquely determined Merge–Agree–Move–spell-out derivation cannot be obtained from these two assumptions alone. Merge and Move below are a way of making a candidate structure explicit, not independent evidence for it. In particular, VPISH alone does not select a v head, a DP layer, a probe, an EPP feature, or the side on which any constituent is pronounced.
The paper’s claim, and the checked data
SOURCED Speas calls (4.6a) SOV and (4.6b) OSV, with grammatical relations “switched” by the object prefix (p. 249). Both participants may be unexpressed as independent nouns ((4.7a,b), p. 249), but the one-noun interpretations are restricted ((4.8a,b), p. 250). She presents Platero’s NP-PRO constraint as an account of that restriction (p. 250), then explicitly says it is inadequate for more complex cases, notably relatives (p. 251). Her larger proposal places nouns in argument positions (p. 201) and assigns a theta role to a noun–clitic CHAIN (p. 324, (4.124)); that is a theoretical analysis, not something proved by these eight entries.
SOURCED Rechecked against the actual page images for this review: Speas pp. 201, 249, 250, 251, 324 and Willie p. 160. The first six IDs are Speas’s; NX-007 and NX-008 are Willie’s, not additional Speas examples. They provide a comparison with a second verb, not a single controlled experiment across speakers. Exact source glosses and reading labels are at the linked index entries; this summary supplies no new gloss or segmentation.
NX-002, Speas (4.6b), p. 249: ashkii atʼééd b-iiłtsą́. — “The girl saw the boy”. The printed word gloss reverses “boy girl” to “girl boy” although the nouns do not change; that inconsistency is retained in the index, not silently repaired.
NX-004, Speas (4.7b), p. 249: b-iiłtsą́. — “S/he saw him/her”, with the source’s qualification “(talking about him/her)”. The two English translations do not establish identical discourse conditions.
NX-005, Speas (4.8a), p. 250: Atʼééd yiiłtsą́. — “Sh/he saw the girl”; explicitly NOT “The girl saw him/her”. “Sh/he” is the source’s spelling.
NX-008, Willie (17b), p. 160: ʼashkii biztsʼǫs — “She was kissed by the boy.”; *“She kissed the boy.” English passive wording is not evidence by itself for a Navajo passive derivation.
SOURCED A printed NOT rejects a reading; it does not mean the reading was absent from the source. A star on an English translation likewise does not star the entire Navajo string. Unstarred means unstarred in that publication, not newly elicited. This review clarifies that distinction in the index.
Extra hypotheses: the full price of a candidate analysis
HYPOTHESIS Call the two semantic participants S and O, with their roles fixed for a given reading. For ‘see’, S is the perceiver and O the perceived participant; for ‘kiss’, S is the kisser and O the kissed participant. These labels are semantic bookkeeping, not claims that seeing is agentive or telic. X and Y below are schematic argument expressions, not Navajo words, and neither is assumed to be a DP.
H1 · Lexical frame and geometry. Model each selected use with two theta-linked expressions. Merge O with V and S at the edge of the same verbal projection: [VP S [V′ O V]]. This particular binary geometry and its head-final pronunciation are hypotheses beyond VPISH. Theta assignment here is V-to-O at the first Merge, and the predicate-to-S at the second. A little-v implementation is possible but not required or assumed.
H2 · Silent realization. A participant expression can have no independently pronounced noun, yet retain its reference and theta role. Calling it pro, an affixal pronoun, or an unpronounced nominal would add a further analysis. The observed noun omissions do not choose among those options.
H3 · Direct/inverse order. Candidate direct order is S–O–V. Candidate inverse order is O–S–V, obtained here by internally merging O at an outer verbal edge above S and pronouncing the upper O occurrence. The lower O occurrence keeps its theta connection; no new theta role is assigned by this Move. Its trigger and A/A′ status are OPEN. A non-movement realization/linking account remains a competitor.
H4 · One-noun restriction. In this limited third-person dataset, if exactly one independent noun is pronounced, it realizes the lower of the two ordered participant expressions: O in direct order, S in inverse order. This fits the printed restrictions. It is an extra stipulation to be tested, not a consequence of theta roles, VPISH or movement. No global NP-PRO rule or Binding Principle is presumed.
H5 · Morphological realization, with optional Agree implementation. Associate the source’s direct and inverse verb forms with those two configurations. A candidate Agree implementation lets a verbal realization head H match the participant features and register their relative order after H3; spell-out then selects the attested direct or inverse form. The existence of H, its probe features, locality and timing are all HYPOTHESIS, not observed here. No feature valuation or intervention contrast in NX-001–008 forces Agree at all; contextual spell-out without Agree is an equally unexcluded implementation. We do not derive whole verb spellings or prefix position classes from this tiny sample.
Merge, theta assignment, Agree, Move, spell-out: entry by entry
HYPOTHESIS The following steps are conditional on H1–H5. To avoid hiding the ordering assumption, the optional Agree implementation of H5 is applied after the candidate inverse Move. Agree copies/values features only; it never itself moves a noun or assigns a theta role. Nothing in these examples chooses that timing over chain-sensitive spell-out. No Q or NEG head is licensed by these affirmative declaratives.
NX-001. Merge perceived girl with SEE (O role); merge perceiver boy inside VP (S role). Move: none across the participants. Optional Agree: H matches the two participants in S-over-O configuration. Spell-out: both expressions pronounced in S–O order, with the source’s direct verb form. H4 is inapplicable because there are two nouns. This reproduces the printed boy-sees-girl reading.
NX-002. Merge perceived boy with SEE; merge perceiver girl inside VP. Move boy/O over girl/S to the outer verbal edge; do not reassign roles. Optional Agree: H registers O above S. Spell-out: upper boy occurrence, then girl, then the attested inverse form; lower boy occurrence unpronounced. This reproduces girl-sees-boy without changing either noun’s meaning.
NX-003. Merge a perceived participant with SEE; merge a perceiver inside VP. H2 leaves both without independent nouns. Move: no inverse crossing in this candidate. Optional Agree: S over O. Spell-out: the attested direct verb only. Reference must be supplied by context; the translation does not fix it.
NX-004. Same two theta-linked participants. Move O above S although neither has a pronounced noun. Optional Agree: O over S. Spell-out: the attested inverse verb only. H3/H5 thus allow a contrast without visible noun reordering. The source’s “talking about him/her” qualification is preserved as a requirement for a future discourse test, not derived by an invented Topic head.
NX-005. Merge girl/O with SEE; merge silent perceiver/S inside VP. Move: none across them. Optional Agree: S over O. H4 licenses girl in the lower O expression. Spell-out: girl and the printed direct form. The rejected girl-as-perceiver reading would put the sole noun in the higher S expression, so H4 excludes it. H1–H3 alone would not exclude it.
NX-006. Merge silent perceived participant/O with SEE; merge girl/S inside VP. Move O above S. Optional Agree: O over S. H4 licenses girl in the lower S expression; spell-out leaves upper O silent and pronounces girl plus the printed inverse form. The rejected girl-as-O reading would place the sole noun in the higher expression, violating H4. It is H4, not theta theory, that supplies this exclusion.
NX-007. Merge boy/O with KISS; merge silent kisser/S inside VP. Move: no crossing. Optional Agree: S over O. H4 licenses the lower O noun. Spell-out: the printed boy noun and direct verb. The source’s starred boy-as-kisser reading fails H4, just as the rejected reading in NX-005 does.
NX-008. Merge silent kissed participant/O with KISS; merge boy/S inside VP. Move O over boy/S. Optional Agree: O over S. H4 licenses lower boy/S; spell-out pronounces boy and the attested inverse verb. The rejected boy-as-kissed-participant reading would realize the upper expression and fails H4. No agent deletion, passive Voice head or suppression of an external theta role is needed for this candidate; whether Navajo has a different passive construction is a separate test.
HYPOTHESISWhat the fit buys us. This candidate reproduces the listed role readings and their exclusions, but H4 was fitted to the one-noun contrasts themselves. That is descriptive coverage, not independent confirmation. The noun-less inverse has no visible crossing; H3 is a structural prediction. A rival can preserve the theta structure and map noun realization/order to the direct/inverse form without syntactic crossing. Binding, reconstruction, scope and question/negation interactions are needed to separate them. Do not count “all readings covered” as proof of movement.
Does light d earn its keep?
HYPOTHESISNo: light d is not needed for NX-001–008. The candidate above has no d node and still states the observed contrasts. Adding one d for each participant, and calling the exponents light determiners, does not remove H3–H5 or predict a new reading here. Keeping d stationary versus moving it with a retained lower copy changes none of these observations. Conversely, omitting d does not explain the one-noun restriction automatically. Light d is idle for this comparison, not disproved in Navajo. Evidence that independently requires its category, selection, scope or distribution could change that verdict. Nor have these entries proved that prefixes are non-arguments or that overt nouns are adjuncts.
Telicity: an actual test, not a hidden premise
OPEN None of NX-001–008 supplies a telic/atelic contrast with the other factors held fixed, or a source diagnostic establishing an endpoint. ‘See’ versus ‘kiss’ also changes the predicate and source; it is not a telicity minimal pair. English past tense, a perfective gloss elsewhere, or an event’s being over does not by itself establish telicity. Therefore these data neither support nor contradict a surface-grammar effect of telicity.
OPEN Next test: find sourced pairs with an independently motivated endpoint contrast, controlling participants, person, noun count/order, discourse, voice and tense/aspect/mode as far as the data permit. Ask whether any surface change tracks telicity rather than those controls. Record supported / contradicted / OPEN for a precisely stated prediction. Search Smith, Perkins & Fernald 2007 and the Navajo Verb Modes resource first, then the HMP inchoative material; none of their prior site interpretations is a premise.
Passive, polar/wh questions and their negatives
OPEN Grant’s diagnostic is a 2×2 (polar versus wh) × (affirmative versus negative), crossed with active versus passive, with inverse tracked separately until its relation to passive is established. There are no questions or negatives in NX-001–008 or in the particular source passages image-checked for this evaluation. The absence is local to this batch, not a claim about either whole dissertation. A passive-looking English translation in NX-008 does not fill a Navajo passive-question cell.
Question cell
Active
Inverse / bi-
Passive / mediopassive
Polar, affirmative
OPEN
OPEN
OPEN
Polar, negative
OPEN
OPEN
OPEN
Wh, affirmative
OPEN
OPEN
OPEN
Wh, negative
OPEN
OPEN
OPEN
OPEN Search the Navajo Questions resource, Willie’s question/focus and negation discussions, Speas’s question examples, Hale & Platero’s negative-polarity manuscript (earlier labelled Hale & Perkins; corrected 2026-11-13, audit pass 7), and the Navajo Verb Modes resource. Willie’s passive/indefinite paradigms supply a voice comparison to seek, not permission to manufacture questions from declaratives. For each attested cell, first copy the full form, source gloss, context and judgment; then show each Merge and theta assignment, optional Agree, every Move and copy pronounced, and the competing placements of Q, the wh expression and NEG. Test which item crosses which, marker attachment, inversion, scope and possible reconstruction. Do not assume a fixed Q–NEG–Voice hierarchy from the desired English translation, or infer wh movement from the presence of a question word. The location and scope of doo … da must be established from attested negative examples. No Q/NEG ordering is forced by this batch, so adding their trees here would be invention, not derivation.
Agreement and disagreement: Speas and our earlier §§10–15
SOURCEDSpeas: retain her printed role contrasts, null-noun possibilities and warning about NP-PRO’s limits (pp. 249–251). HYPOTHESIS Her prefix-switched description fits the observations, but the causal choice “prefix changes the syntax” versus “configuration selects the prefix” is not settled here. We neither refute nor inherit her noun/CHAIN or null-pronominal analysis. Theta roles are assumed, but a universal NP-PRO or Binding Principle is not.
HYPOTHESIS§10: keep the empirical one-noun generalization and underdetermination of stationary d. Withdraw movement-as-established language: DP movement followed only conditional on ID-1–ID-6. This review supplies an explicitly d-free candidate and admits that movement itself still needs evidence.
OPEN§11 and §12: parallelism, rank and the ID-8 head restriction are separate hypotheses to retest against indexed relatives, not licenses inherited by this review. NX-001–008 contains no relative clause and cannot validate them.
OPEN§13, §14 and §15: the overtness-dependent ID-8 revisions, external-head/dialect explanation, EH- and RD-rules, raising/matching options and A/B boundary all return to the test queue. Keep attributions distinct from explanations. This batch cannot establish a dialect difference or relative-head movement. The earlier sections remain a research history, not the premises of the new series.
Predictions for unindexed evidence
OPENOne-noun generalization: H4 predicts the same lower-expression restriction for additional third-person verbs under comparable discourse. An attested contrary reading defeats H4 in that domain; do not rescue it by silently changing a judgment or declaring a new dialect.
OPENMove versus realization: H3 predicts an object with a lower theta connection and a higher inverse occurrence. Seek source-controlled binding, scope or reconstruction differences; accept that an order-only account may fit this batch equally well.
OPENDiscourse: compare the noun-less direct/inverse in documented contexts before inferring a Topic head from NX-004’s parenthetical.
OPENOut-of-sample checks: person combinations, possessed nouns, reflexives, relatives, true passive/mediopassive paradigms, all four question/negation cells and controlled telicity contrasts. H4/H5 are not asserted outside the present third-person declaratives. No unattested Navajo string or speaker judgment is predicted as if already observed.
Next paper: Willie 1991, beginning with Chapter 3’s direct/inverse account and the p. 160 comparison, then the passive/indefinite and question/negation evidence. The box research queue records the wider source inventory and unresolved access gaps. No new example IDs were added in this review: all eight forms were already indexed and have been rechecked.
2026-11-10 (evening) · First-principles review 2: Willie 1991 on direct, inverse and passive
2026-11-14: the “On the spine” summary at the top of this page supersedes this section’s framing. Its data, judgments and citations are unchanged.
Restated on the unified structure (2026-11-15). This section’s results now sit on the one structure for Navajo: unified structure, inverse and passive rows. Its restatement as derivations is in Results on the spine. Data, judgments and citations below are unchanged.
Willie 1991, Chapter 3 (printed pp. 51–101; examples from pp. 57–100) and Chapter 2 §2.3 (pp. 35–39). Every form below was read off rendered page images before indexing. New index entries NX-020–061. Follows review 1.
2026-11-11 audit pass 3. “Confirmed” was softened to “consistent” / “fits” for H3 and H4: fitting more forms does not prove a hypothesis. See the pass 3 report.
HYPOTHESISPlain-words verdict. Willie’s data hold up well, and they test review 1 for real. The one-noun rule from review 1 holds for a new verb (‘kick’), but it is only a default: the verb’s meaning can override it. The inverse’s word order is less fixed than review 1 assumed, and Willie’s yes/no question test shows that the reordered version has a different structure. Her definiteness restriction follows the role (the inverse patient), not the first word. It is stated most simply on a derivation in which that patient has moved. Willie gives no argument, only a translation, for treating Navajo passives (Chapter 2 and pp. 57, 99) as anything other than verbs that have one fewer argument a noun can name. A light determiner is idle again: nothing here needs it, and nothing here counts against it. Telicity: OPEN. Grid: one cell (polar question, affirmative, inverse) is filled with sourced data. All other cells are OPEN.
Starting point and what is not inherited
Assumed, by instruction: VPISH and theta roles, nothing else. HYPOTHESIS Review 1’s H1–H5 are carried forward as hypotheses to be tested, not results. Willie’s own framework is reported as hers and is not adopted. That framework includes pronominal arguments in the verb, nouns as adjuncts coindexed with them, Case on the incorporated pronouns, the animacy hierarchy and her use of “internal/external argument”. Where her terms are quoted (“internal”, “external”, “subject”), they are her grammatical relations, not ours.
Willie’s claims, as hers
SOURCEDOne noun, transitive verb: “When a transitive sentence has only one adjunct, that adjunct is interpreted as coindexed with the object (patient) argument” (principle (17), p. 59), unless the verb’s semantic features “override” it (pp. 61–62).
SOURCEDTwo nouns: the inner noun goes with the internal argument; the outer noun goes “(by default)” with the external argument (principle (20), p. 60). In both voices, the noun immediately before the verb is an adjunct to the “internal” argument, “regardless of whether the theta role of that argument is agent or patient” (p. 72).
SOURCEDInverse: it is “comparable to a passive in that the subject has a Patient theta role”, but transitive (p. 71). In it, the agent is the “internal” argument, with Ergative case, and the patient pronoun has “Accusative case” (p. 72). Yet p. 77 calls the inverse patient “the external argument (the patient, with Absolutive case)”. Both case labels are retained; we do not choose between them. The inverse “always makes the argument with an agent theta role the internal argument” (p. 97).
SOURCEDDefiniteness: “with the Inverse, the external argument … must be definite” (p. 77). Animacy: the agent-higher pattern uses the direct; the patient-higher pattern “must” use the inverse; equal rank is left to discourse ((70), p. 86).
SOURCEDOrder freedom and focus: with bi-, but not yi-, the nouns can be reordered without changing meaning (p. 90). The fronted ʼiiʼniʼ gets there “by focus movement” (p. 91). The yes/no clitic “usually follows the first word”, but it cannot attach anywhere in the focus-moved structure (p. 92).
SOURCEDPassives: the “Agentive Passive” has an indefinite agent that “must be understood as human, but no noun that describes this agent may be adjoined” (p. 36). The impersonal ʼa- prefix “never permits an adjoined coreferent nominal” (p. 38). For a one-argument verb with a patient subject, Willie cites Young & Morgan’s “simple passive” label (p. 57). Example (94) is “a simple passive, an Impersonal”, so its postposition “must be lexical” (p. 99).
The checked data
SOURCED Exact forms with printed example numbers. Word glosses, readings, judgments and printed inconsistencies are given at the linked index entries. Translations are hers; English passive wording is not treated as evidence of a Navajo passive.
Direct vs inverse, two nouns, equal animacy:NX-029, (37a), p. 71: łį́į́ʼ dzaanééz yiztał “The horse kicked the mule.” · NX-030, (37b), p. 71: dzaanééz łį́į́ʼ biztał “The mule was kicked by the horse.”
One complex noun vs two nouns:NX-035, (40), p. 74: Díí ʼashkii yiztał (a) “This (one) kicked the boy.”, (b) “He kicked this boy.”, *(c) “This boy kicked him.” · NX-036, (41), p. 75: Díí ʼashkii biztał (a) “He got kicked by this boy.”, (b) “This (one) got kicked by the boy.”, *(c) “This boy was kicked by him.”
Semantic override of the one-noun rule:NX-027, (22), p. 61: ʼashkii yiyííbį́į́ʼ “The boy picked them. (c.f. berries)” · NX-028, (24), p. 62: ʼashkii yiyííłtaʼ “The boy counted them.” Compare NX-026, (18), p. 59: chidí yiyíílchxǫʼ “He/she ruined the car.”
Unequal animacy, order and focus:NX-045, (73a), p. 90: ʼiiʼni łį́į́ʼ yiyiisxį́ “The lightning killed the horse.” (Willie finds it objectionable; some speakers accept it) · NX-046, (73b), p. 90: łį́į́ʼ ʼiiʼniʼ biisxį́ “The horse was killed by lightning.” · NX-047, (74a), p. 90: iiʼniʼ łį́į́ʼ biisxį́ and NX-048, (74b), p. 90: lííʼ iiʼniʼ biisxí, both “The lightning killed the horse.” · *NX-049, (74c), p. 90: iiʼniʼ łííʼ łaʼ biisxį́ · NX-050, (75a), p. 91: łííʼ iiʼniʼ yiyiisxį́ “The horse killed the lightning.” (only reading) · NX-051, (75b), p. 91: iiʼniʼ łííʼ yiyiisxį́ “The lightning killed the horse.”
SOURCED A star before an index link means the whole Navajo string is starred in print. A star inside a reading rejects that reading only. ??? is Willie’s own mark. Printed spelling differences are not corrected. For example, ‘horse’ appears as łį́į́ʼ, łííʼ and lííʼ, and ‘lightning’ as ʼiiʼni, ʼiiʼniʼ and iiʼniʼ. Nothing below depends on those differences.
Hypotheses tested, revised or added
HYPOTHESIS S and O are semantic bookkeeping for the two theta-linked participants of a transitive verb: S is the kicker, kisser, killer or climber, and O is the one kicked, kissed, killed or climbed. G is a postposition’s object. None of these is assumed to be a DP.
H1–H2 (from review 1), unchanged. [VP S [V′ O V]]; a participant may have no pronounced noun and still keep its role.
H3 (from review 1), consistent in shape, still unforced. Inverse = O internally merged above S at an outer verbal edge, with the upper O pronounced. Checked against NX-029/030 (‘kick’) as well as ‘see’ and ‘kiss’.
H4 → H4′ (revised). Review 1’s one-noun rule realized a single noun as the lower of the two ordered participants: O in the direct, S in the inverse. It fits the out-of-sample forms NX-031/032 and NX-035/036 (‘kick’). It is contradicted as a strict rule by NX-027/028, where the lone noun with the direct form is the agent. H4′: H4 is a default for linking a noun to a participant. Theta compatibility outranks it: a noun cannot be linked to a participant whose role it cannot bear, such as berries as pickers. Only this last part follows from the theta-role axiom. The ranking “compatibility over H4” is an added hypothesis. NX-035(c)/NX-036(c) show that H4 still decides when both linkings are semantically possible.
H6 (new) · Definiteness on the moved O. The inverse Move of O (H3) is restricted to definite O. A rival, H6-lin, says “the first noun of an inverse clause must be definite”.
H7 (new) · Focus fronting. In the inverse, a further internal Merge can front S above the raised O, with only the upper S occurrence pronounced. This is Willie’s focus movement, restated. Its trigger is OPEN; Willie links its availability to animacy presuppositions (p. 91).
H8 (new) · Polar Q. Two candidates. Q-pf: Q is an enclitic placed after the first pronounced word, with no syntactic Move. Q-edge: Q is merged above the clause and needs a host constituent at its edge, so the host is the highest constituent there. The focus-fronted S of H7 cannot also be that host, and no other constituent can be the host while S is fronted.
H9 (new) · yi- needs a definite co-argument. A third-person object form yi-, on a verb or a postposition, needs a distinct, definite, third-person co-argument above it. Otherwise the b- form appears. The inverse reading of b- arises only when such a co-argument exists.
H10a / H10b (new) · Passive. H10a: the passive or impersonal verb merges no agent at all. H10b: it merges an agent participant that is indefinite and cannot be linked to a noun.
Derivations, step by step
HYPOTHESIS Each step is Merge with theta assignment, then Agree (optional, as in review 1’s H5, and never forced here), then Move, then which copy is pronounced. Agree never moves anything or assigns a role.
NX-029 (direct, two nouns). Merge mule with KICK, giving O its kicked role. Merge horse at the VP edge, giving S the kicker role (VPISH). No Move. Optional Agree: S over O. Spell-out: horse, mule, then the direct form, S–O–V. This gives “The horse kicked the mule.”
NX-030 (inverse, two nouns). The printed reading makes mule the kicked one, so merge mule with KICK as O and horse at the VP edge as S. Move O (mule) above S (H3). Optional Agree: O over S. Spell-out: upper mule, horse, then the inverse form; the lower mule is silent. This matches Willie’s description that the noun nearest the bi- form is the agent (p. 72).
NX-031 / NX-032 (one noun). In NX-031, mule is linked to the lower participant in the direct, which is O, giving “It kicked the mule.” In NX-032, the lower participant after H3 is S, so horse is the kicker. These follow from H4 and are not fitted to ‘kick’, so they are an out-of-sample success for H4.
NX-035 / NX-036 (complex noun). If Díí ʼashkii is one expression, H4 links it to the lower participant: O in the direct (35b) and S in the inverse (36a). The starred (c) readings link it to the higher participant, which H4 excludes. If díí and ʼashkii are two expressions, the two-noun orders of NX-029/030 give (35a) and (36b). All six printed judgments are derived. The analysis requires no extra hypotheses other than H1–H4. It does not assume a DP layer.
NX-027 (override). Merge silent berries-referent as O; merge boy as S. No Move; direct form. H4 would link the lone noun ʼashkii to O, but O’s role (thing picked, round objects in Willie’s gloss) is not one a boy can bear here. Theta compatibility therefore blocks that link, and H4′ links the noun to S. Spell-out: boy, then the direct form. So H4 alone is contradicted; H4′ derives it.
NX-040 / NX-041 / *NX-042 (definiteness). In all three, merge boy as O and girl as S, then Move O above S (H3), and spell out O–S–inverse. In NX-041, the indefinite marker sits with S, which did not move, and the form is fine. In NX-042 it sits with O, which moved, and the string is starred. H6 and H6-lin both fit these three.
*NX-049 separates H6 from H6-lin. Merge horse plus łaʼ as O and lightning as S. Move O above S (H3: inverse), then front S above O (H7: focus). Spell-out: lightning, horse łaʼ, then the inverse form. The first noun, lightning, is definite, so H6-lin wrongly predicts acceptance. The moved O is indefinite, so H6 correctly predicts the star, and so does Willie’s own account (“despite the word order”, p. 90). The derivation, not the surface position, carries the restriction. This favors H6 or Willie’s role-based statement over any linear rule. It does not distinguish H6 from Willie, because both single out the inverse patient.
NX-046 vs NX-047 (same meaning, reversed nouns). NX-046: Merge horse as O and lightning as S, Move O above S, and spell out horse, lightning, inverse. NX-047: the same steps, then also front S (H7). Spell-out: lightning, horse, inverse, with the lower copies silent. Without H7, H3 alone would wrongly predict horse-as-killer for NX-047, because review 1’s order mapping reads the first of two nouns in the inverse as O. So review 1’s H3 needs H7 for this pair. Without NX-052/053, H7 would just be a word for the data.
NX-050 (direct, rigid). Merge lightning as O and horse as S. No Move, direct form: “The horse killed the lightning”, with no rescue by plausibility (p. 91). With two nouns, the direct gives no H7-type reordering here. Why the direct lacks H7 is OPEN. Willie’s answer is pragmatic (no lower-ranked agent is presupposed, p. 91); our derivation has no syntactic reason for it.
NX-053 (polar question, inverse). Merge horse as O and lightning as S. Move O above S (H3). Inverse spell-out of the verb. Merge Q above the clause. Under Q-pf, -ísh is placed after the first word, horse. Under Q-edge, the raised O is the highest constituent and hosts Q, with no further Move needed if O is already at the edge. Both give łííʼ-ísh iiʼniʼ biisxį́. Q takes scope over the proposition (a yes/no question); the translation shows no narrow focus.
*NX-052 (polar question with focus fronting). Same steps as NX-047, plus Q. Q-pf puts -ish after the first word, iiʼniʼ, and so wrongly predicts the printed string is fine. Willie adds that the clitic “cannot attach anywhere in this structure” (p. 92). This is her prose; no other placement is printed, and we invent none. If that holds, putting -ish after a later word cannot rescue Q-pf either. Q-edge with H8’s condition derives the star whatever the host. Result: a pure first-word rule is contradicted, given Willie’s statement. Polar Q and inverse focus-fronting cannot co-occur, so neither can be ordered above the other. No Q > Focus or Focus > Q hierarchy follows from this pair.
NX-024 / NX-026 (“simple passive” vs transitive). NX-024: Merge the ruined thing as the only argument of RUIN, so it gets the patient role. Under VPISH it is VP-internal and is the only participant there. H10a merges nothing else; H10b merges a silent indefinite causer. Spell-out: yíchxǫʼ. Willie’s remark that only one noun is allowed fits both (p. 57). NX-026: two participants, as in NX-031. The noun is linked to O by H4. The verb forms differ, but we do not segment them.
NX-020 / NX-021 (agentive passive). Merge the second- or first-person patient as O. Merge an agent participant as S: the agent “must be understood as human” (p. 36), which is semantic evidence that an agent role is assigned. This favors H10b over H10a for this construction. No noun may be linked to that S, a property that has to be stated. The order of the patient and agent prefixes is reported by Willie (p. 38); we do not derive it.
NX-054 / *NX-055 / NX-056 / ???NX-057 (postposition object). NX-056: Merge tree as G of the postposition. Merge a silent definite third-person climber as S. H9 is met, so the y- form appears, giving “He climbed the tree.” ???NX-057: the same verb with b-. With a definite third-person co-argument present, H9 makes b- the inverse. H4 then links the lone noun to the lower participant, S, so the tree is the climber, which is Willie’s nonsense reading. NX-054: under H10a there is no co-argument; under H10b the co-argument is indefinite. Either way H9 is not met, so the form is b- with no inverse reading, and *NX-055 is excluded. This is why H9 must say definite: if the indefinite climber counted, H10b with an unrestricted H9 would wrongly predict that NX-055 is fine.
*NX-059 vs NX-060 (H9 under pressure). In NX-059, ball (S) is a definite third person, so H9 permits yiiʼ, but the string is starred. In NX-060, with a woman as S, yiiʼ is fine. H9 as stated is contradicted. Willie’s answer is that an animate subject makes the postposition “grammatical” (p. 98); inanimate subjects leave it “lexical”, with bi- only. Adding “animate” to H9 is a fitted repair, so it is OPEN, not supported. ???NX-061 then follows from H9 plus animacy and H4, like NX-057.
Does light d earn its keep?
GRANT'S CONJECTURE Light d takes a DP argument, as v takes VP. Only head d joins the verb word; d stays in place while the DP moves where needed.
HYPOTHESISIdle here; no counter-evidence. Every derivation above runs without a d node, and adding one changes none of them. (i) Definiteness (H6) is carried by the moved O itself, not by anything in the verb word. The restricting words léiʼ/łaʼ are separate words inside the noun expression. (ii) Movement of the noun expressions (H3, H7) is consistent with “d stays, DP moves”, but just as consistent with no d. (iii) The prefixes that never allow a noun (ʼadi- agentive passive, ʼa- impersonal) would, under the conjecture, be d heads that take no DP. That is a stipulation, but every rival needs an equal one. Pronominal arguments need “this pronoun takes no adjunct”, and agreement accounts need “no DP with these features”. So these data cannot decide among them. Data that could decide: whether the indefinite prefixes and their nouns disagree in scope or binding in ways that track a d head inside the verb word. None is printed in these chapters.
Telicity test
OPEN Not supported and not contradicted. No pair in these pages differs only in endpoint. Kick, kiss, kill, bite, ruin, climb, pick and count appear with no controlled telic/atelic contrast. The printed glosses “kicked” vs “kicks” (NX-036) and “3s.ruin PERF” (p. 58) are tense/viewpoint glosses, and the standing method says viewpoint is not a telicity test. No telicity tag has been assigned.
SOURCED The filled cell is the only printed question in Chapter 3. Willie’s general statement about where -ish goes (p. 92) does not fill the active cell: no active question is printed there, and we do not build one. No negation (doo … da) and no passive question occur in these pages. The passive column is labeled with Willie’s terms (“Agentive Passive”, “simple passive”, “Impersonal”), not our analysis. HYPOTHESIS The derived result is in steps 10–11. Polar Q is compatible with the inverse’s raised O as first constituent, and incompatible with H7 focus-fronting. A pure first-word placement rule fails. Which projection is higher cannot be determined, because the two never co-occur. Next places to look: Willie Chapter 8 (wh-constructions and focus), and the Navajo Questions resource for passives and negatives.
Agreement and disagreement with Willie
HYPOTHESISAgree: her principles (17)/(20) and the override match H4′ step for step. Her role-based definiteness statement survives NX-049, where a linear rule fails. Her focus-movement analysis of NX-047 is supported by her own question test (NX-052/053), so it is not just a label.
HYPOTHESISNot inherited: (a) Nouns as adjuncts versus nouns in argument positions: nothing here forces either, since H1–H4′ need only linking. (b) Her Case labels are inconsistent within the chapter (Accusative p. 72 vs Absolutive p. 77), and nothing here needs Case. (c) Her claim that the inverse re-maps roles, making the agent “internal”, competes with H3, where roles are assigned as in the direct and O moves. VPISH plus theta roles cannot choose between them without a further mapping principle, which is not assumed. Quantifier scope in her (58)–(60), pp. 79–80, could separate them, but has not been image-checked for this review.
HYPOTHESISDisagree, mildly: calling (12) and (94) “passives” is a label borrowed from Young & Morgan or from the English translation. With theta roles alone, NX-024 is just a one-participant verb (H10a or H10b). Only the agentive passive (NX-020/021) has evidence of a suppressed agent role.
OPENSpeaker variation: NX-045 is accepted by some speakers but not by Willie, yet NX-051 has the same order and verb and is printed unstarred. We record this and do not attribute it to a dialect.
What this changes in review 1
HYPOTHESIS H4 has been tested out of sample. It passes for ‘kick’ and fails as a strict rule (NX-027/028). It is replaced by H4′.
HYPOTHESIS H3 stands, but the inverse’s word order is not a direct readout of roles. With unequal animacy, H7 can reorder the nouns, and the question test detects that reordering.
OPEN Review 1’s “Move vs realization” question is still open. H6 and H8 are most simply stated on movement, but a realization account that refers to “the inverse patient” covers the same facts.
OPEN predictions
OPENH7 with equal animacy: if focus fronting needs an animacy presupposition (Willie p. 91), equal-rank inverses like NX-030 should not reorder without a change in meaning. No such pair is printed here.
OPENH6 scope: a quantified inverse O should take wide scope if it has moved and does not reconstruct. Check Willie (58)–(60), pp. 79–80, against the page images.
OPENH8: Q-edge predicts that any construction with a fronted focus constituent blocks -ish, in the direct too. Look for printed direct-voice focus plus question examples (Willie Chapter 8).
OPENH9 + animacy: predicts that yi- on a postposition object needs an animate, definite third-person co-argument. Test with inanimate subjects of transitive verbs and with ʼadi- passives (indefinite human agent). Under H9, the passive should take b- only, because its agent is indefinite.
OPENH10a vs H10b for the simple passive: seek source data in which a purpose or instrument phrase needs an implicit agent with yíchxǫʼ-type verbs.
Next in the queue: Willie Chapter 8 (wh-constructions and focus), to fill the question cells, and her (58)–(60) on quantifier scope. Then Hale & Platero 1974.
2026-11-11 · First-principles review 3: Willie 1991 on wh-questions, the question clitic and the inverse
2026-11-14: the “On the spine” summary at the top of this page supersedes this section’s framing. Its data, judgments and citations are unchanged.
Restated on the unified structure (2026-11-15). This section’s results now sit on the one structure for Navajo: unified structure, question rows. Its restatement as derivations is in Results on the spine. Data, judgments and citations below are unchanged.
Willie 1991, Chapter 8 “Questions and Focus” (printed pp. 199–219; examples from pp. 201–215). Every form below was read off rendered page images before indexing. New index entries NX-062–088. Follows review 2; hypothesis numbers H1–H10 refer to reviews 1–2.
HYPOTHESISPlain-words verdict. Navajo question words sit wherever an ordinary noun with the same role would sit, in both the yi- and the bi- form. So the rules for nouns from reviews 1–2 carry over to them unchanged, and that is an out-of-sample success for those rules. But question words are not simply frozen in place. They can be fronted when meaning makes the roles clear, and the question clitic =sh can leave its question word and attach to the first word, except when that first word is the patient raised by the bi- inverse. This one contrast between direct and inverse forms (NX-085 vs *NX-086) shows that inverse movement has already happened when the clitic is placed. It is the first sourced result in this project about the order of operations between voice and questions. A light determiner is idle again: nothing here needs it, and nothing here counts against it. Telicity: OPEN. Grid: the wh-affirmative cells are now filled for both the direct and the inverse. Every negative cell and every passive question cell is still OPEN.
Starting point
Assumed, by instruction: VPISH and theta roles, nothing else. HYPOTHESIS H1–H10 from reviews 1–2 are tested here as hypotheses, not taken as results. Willie’s framework is reported as hers and is not adopted. In that framework the pronouns in the verb are the arguments, nouns and wh-words are adjuncts, there is no wh-movement except at LF, and an animacy hierarchy applies. S and O are again semantic bookkeeping for the two participants of a transitive verb (kisser/kissed, seer/seen). G is the goal or postposition object, and T is the thing given.
Willie’s claims, as hers
SOURCEDNo wh-movement: “Wh-words in Navajo always appear as adjuncts in A-bar positions, and thus cannot undergo Wh- movement”, and “the positions of Wh- words at S-structure … is the same as that of other adjoined nominals” (p. 199). She assumes wh-movement “takes place at LF” (p. 200).
SOURCEDTwo kinds: nominal wh-words (háísh “who”, haʼátʼíísh “what”) question an argument and exclude a coreferent noun. Adverbial wh-words (where, when) do not (pp. 201–202).
SOURCEDVoice: a wh-word “appears in the same adjunct position that a coindexed NP would occupy”, and that is “determined by the Direct/Inverse voice alternation” (p. 203). A wh-word linked to the inverse subject is “presuppositional”, like “which one” (pp. 204–205).
SOURCEDFocus movement: an inanimate wh-object can come first in a direct clause, as “a kind of focus movement of the Wh- word”. This is possible only when animacy keeps agent and patient apart, and not with equal rank (pp. 206–207).
SOURCEDThe clitic:=sh “overtly marks focus”. Without it, and with a modal clitic, a wh-word is an indefinite (p. 210). =sh “can be moved away from the Wh- word to a position following the first word”, but not in inverse constructions, which she attributes to the inverse already placing “focus on the patient as subject” (p. 211).
SOURCEDOne wh per clause: “It is impossible to have both an adverbial and an argumental wh- word in the same clause” (p. 215). Her long-distance claims (pp. 212–218) are not indexed or evaluated in this review.
The checked data
SOURCED Exact forms with printed example numbers. Word glosses, readings and printed inconsistencies are given at the linked entries. The translations are Willie’s and are deliberately “echo question” in style (p. 204). English passive wording translates the inverse; it is not a Navajo passive.
Ditransitive:NX-076, (11a), p. 208: John Mary haʼátʼíísh yeiníʼą́ “John gave Mary what?” · NX-077, (11b), p. 208: John Mary haʼátʼíísh beiníʼą́ “John was given what by Mary?”
Postposition object:NX-078, (12a), p. 209: John Mary yichʼįʼ yáłtiʼ “John is talking to Mary.” · NX-079, (12b), p. 209: John Mary bichʼįʼ yáłtiʼ “John is being talked to by Mary.” · NX-080, (13a), p. 209: John háísh yichʼįʼ yáłtiʼ “John is talking to who?” · NX-081, (13b), p. 209: John háísh bichʼįʼ yáłtiʼ “John is being talked to by who?”
SOURCED A star before an index link means the whole string is starred in print. A star inside a reading rejects only that reading. Printed oddities are kept and noted at the entries, not corrected. For example, NX-074’s verb reads yiviiłtsą́ in the scan where NX-072 has yiyiiłtsą́. NX-075 is introduced as “(4a), repeated” but differs from NX-062. Willie’s prose on p. 205 refers the “which one” reading to (8b), but that translation is printed under (8a), NX-068. Example (10) on p. 207 repeats the one-noun “berries” pattern of NX-027 (Chapter 3) and is not indexed again.
Hypotheses tested, revised or added
H1–H2, H3 and H4′ (reviews 1–2): carried over, now tested on wh-words. [VP S [V′ O V]] (VPISH). H3: the inverse is O internally merged above S, with the upper copy pronounced. H4′: a lone noun is linked to the lower participant by default, unless theta compatibility rules that out.
H3-loc (refinement). The phrase that H3 raises is the closest internal participant to S: G, not T, in a ditransitive. This needs an extra hypothesis that G is merged above T. Willie’s rival is that the inverse switches theta roles (p. 208).
H6′ (revision of H6). The O raised by H3 must be definite or presuppositional (a “which one” reading).
H9′ (revision of H9). Review 2’s H9 required a definite third-person co-argument for yi-. H9′ drops “definite”: yi- needs a distinct third-person co-argument above it.
H11 (new) · A wh-word is an indefinite plus Q.háí (printed without =sh in NX-084 and NX-085), and the parts of haʼátʼíísh and háa=di left once =sh is set aside, are indefinites. That segmentation is ours. The clitic =sh is (or spells out) a question element Q, which is merged with the phrase containing the indefinite. Willie’s observations support the split (p. 210, NX-084, NX-085). That the split is syntactic is our hypothesis.
H12 (new) · Q can be displaced.H12-move: Q alone can be internally merged at the clause edge, and the upper copy is then pronounced after the first constituent there. The lower copy, next to the wh-word, is silent. H12-pf: the relocation is purely phonological, with no syntactic Move.
H13 (new) · What blocks displacement in the inverse.H13-int (intervention): displaced Q cannot be linked back to its wh-word across a phrase that H3 has raised above that wh-word. H13-host (a version of Willie’s focus account): an H3-raised O cannot host =sh unless the question is about O itself (O is the wh-word, or the question is polar).
H14 (new) · Fronting a wh-O in the direct. A wh-O can be internally merged above S in a yi- clause, with the upper copy pronounced. This is the H7 operation extended to the direct. The condition on it is OPEN. Willie’s condition is that the roles must remain recoverable from animacy (pp. 206–207).
Theta biuniqueness (made explicit). Each theta role goes to exactly one phrase. This is the standard reading of the theta-role axiom, and it is listed so that nothing hides inside it.
Derivations, step by step
HYPOTHESIS Steps: Merge with theta assignment, then Agree (optional, as in reviews 1–2, and never needed below), then Move, then which copy is pronounced, then what is spelled out inside the verb word (the inflected verb, which we do not segment) and what is spelled out outside it (nouns, wh-words, the clitic). “Q” is H11’s element.
NX-062 / *NX-064 / NX-065 (one participant). NX-062: Merge the indefinite háí with WORK; it gets the worker role, VP-internal under VPISH. Merge Q with it (H11). No Move. Spell-out: háísh outside the verb word, then the verb word. *NX-064: háísh and shizhéʼé are two separate phrases, and there is only one worker role. By theta biuniqueness, one of them gets no role, so the derivation fails. Why they cannot form a single phrase is OPEN. In Willie’s (18)–(20), not indexed here, a noun and a wh-word do form one phrase, but there the wh-word comes after the noun. NX-065: háadish “where” takes no theta role from WORK. It is merged as a modifier, and shizhéʼé takes the worker role, so nothing clashes. No extra hypotheses are needed beyond biuniqueness and H11.
NX-066 / NX-067 (who, direct). NX-066: Merge John with KISS, which gives O its role. Merge háí+Q at the VP edge, which gives S its role (VPISH). No Move. Spell-out: háísh, John, then the direct verb. This is the same S–O–V order as two nouns (NX-029, review 2). NX-067: the same, but the indefinite+Q is O and John is S, giving John, háísh, verb. No step moves the wh-word. Overt wh-fronting is not obligatory, so nothing in these derivations needs a Q probe that attracts a wh-phrase to the clause edge.
NX-068 / NX-069 (who, inverse). NX-068: Merge háí+Q as O and John as S. Move O above S (H3). Pronounce the upper copy. Spell-out: háísh, John, inverse verb. So H3 moves a wh-phrase, together with its Q, exactly as it moves a noun. H6 would require the moved O to be definite, but a wh-word is not definite. The printed reading, “Which one (of you)…”, is presuppositional, so H6 is revised to H6′. That revision is fitted to this example, but it also predicts that a non-presuppositional reading should push the speaker to the direct form, which is what Willie reports (p. 205). NX-069: Merge John as O and háí+Q as S. Move O (John) above S. Spell-out: John, háísh, inverse verb. The wh-word stays where it was merged.
NX-070 / NX-071 / NX-082 / NX-083 (lone wh-word). With one overt phrase, H4′ links it to the lower participant: O in the direct (“He kissed who?”, “He is talking to who?”) and S in the inverse after H3 (“He was kissed by who?”, “…talked to by whom?”). All four match. H4′ was built on nouns, so this is an out-of-sample success for it.
NX-072 / *NX-073 (what, unequal animacy). NX-072: Merge ‘what’+Q as O and John as S. No Move; direct. Spell-out: John, haʼátʼíísh, verb. *NX-073: the same Merge steps, then H3 raises the inanimate O above the human S. Review 2 already found the inverse barred when the agent outranks the patient (NX-044, reading 4), so nothing specific to questions is needed. This animacy condition is still a stipulation (OPEN).
NX-074 vs NX-075 (fronting in the direct). NX-074: Merge ‘what’+Q as O and John as S. In the direct, internally merge O above S (H14). Pronounce the upper copy; the lower copy is silent. Spell-out: haʼátʼíísh, John, direct verb, meaning “What did John see?” Without a Move step, VPISH plus H4′ would read the first phrase as S, and theta compatibility would then be violated (a “what” cannot be the seer here). So some displacement is forced. NX-075: the same derivation with two humans would give the same string as base order with “who” as S. The printed facts are that only the base-order reading exists: *“John saw who?”. So H14 is unavailable when both participants are human. Willie’s recoverability condition fits, but in our terms it is a filter on the output, not a consequence of VPISH or theta roles (OPEN). Contrast with plain nouns: in review 2, NX-050 (direct, horse/lightning) allowed no reordering even when plausibility favored one. So H14 may be specific to wh-phrases. The verbs and animacy gaps differ, however, so this is not a controlled pair (OPEN). Willie’s only non-interrogative illustration of “comparable focus movement” (p. 207) is a one-noun string. It shows the H4′ override, not reordering.
NX-076 / NX-077 (ditransitive). Merge ‘what’+Q as T with GIVE, Mary as G above T (extra hypothesis, H3-loc), and John as S at the edge. NX-076: no Move. Spell-out: John, Mary, haʼátʼíísh, direct verb, meaning “John gave Mary what?” NX-077: the printed reading makes John the recipient, so merge John as G and Mary as S. Move G above S (H3). Spell-out: John, Mary, haʼátʼíísh, inverse verb. The wh-word T stays next to the verb in both. H3-loc predicts that the inverse can never raise T over G and S. Willie’s role-switch account makes the same prediction for these strings, so the two are not yet distinguished (OPEN).
NX-080 / NX-081 (postposition object). NX-080: Merge háí+Q as G of the postposition. Merge that phrase inside VP and John as S. No Move. Spell-out: John, háísh, postposition with yi-, verb. NX-081: Merge John as G and háí+Q as S. Move G out of the postposition phrase above S (H3), leaving the postposition, which is spelled out with bi-, in place. Spell-out: John, háísh, bichʼįʼ, verb. This gives Willie’s observation that the wh-word “intervenes” between the postposition and its noun (p. 209). Extra hypothesis: H3 can extract a noun from a grammatical postposition phrase (HYPOTHESIS). Review 2 needed the same for NX-056/057.
NX-085 (displaced clitic, direct). Merge háí+Q as O and díí as S. No voice Move. H12-move: internally merge Q at the clause edge. The highest constituent there is díí, so Q is pronounced after it and its lower copy is silent. Spell-out: Díí=sh, háí, direct verb. The question is still about the O (“kicked who?”), so Q keeps its link to háí. H12-pf gives the same string.
*NX-086 (displaced clitic, inverse). Merge díí as O and háí+Q as S. Move O above S (H3). Then Q would be placed at the edge, after the raised díí. Neither H12-move nor H12-pf alone excludes this, and neither does review 2’s H8 (Q-pf or Q-edge). So H12 needs H13. Under H13-int, the raised O lies between Q’s edge position and háí’s position, which blocks the link. Under H13-host, the raised O is not what the question is about, so it cannot host =sh. Both derive the star.
Willie’s focus account checked against NX-069 / NX-081. If the inverse already uses up the clause’s one focus (p. 211), then any wh-question in the inverse about S should fail. It does not: in NX-069 and NX-081, háísh is S of an inverse and the strings are fine. So a plain one-focus-per-clause version is contradicted by her own examples. Only a version restricted to the host of =sh (H13-host) survives.
Order of operations (result). In NX-086 the blocker is the output of H3. If Q placement were evaluated before H3 moved O, díí would still be below háí, there would be nothing in the way, and the string should be fine. So H3 inverse movement must precede (or feed) the placement of a displaced Q, or the condition must be checked on the final structure. NX-068 shows the converse is free: H3 can move a phrase that contains Q. These are our derivations from sourced judgments, so the result is labeled HYPOTHESIS. It is the first ordering in this project between voice and question formation with sourced support.
NX-084 (indefinite). Merge háí as O, with the probability clitic in place of Q, and John as S. No Move; direct. The result is a statement with an indefinite: “John (probably) kicked someone.” Together with NX-085, where háí without =sh on it is still interpreted as the questioned phrase, this supports H11: question force comes from Q, not from the indefinite itself.
*NX-087 / *NX-088 (two wh-words). Nothing in VPISH, theta roles or H11–H14 excludes these. “Where” takes no theta role, so biuniqueness does not apply (compare NX-065). háa=di is printed without =sh, so we cannot tell whether the problem is two Qs, two wh-phrases, or an indefinite “somewhere” inside a question. Willie’s one-wh statement is recorded and not derived (OPEN).
H9 rechecked (a correction to review 2). H9 said yi- needs a definite third-person co-argument above it. In NX-066 and NX-075, that co-argument is S háísh, which is not presuppositional (Willie: the direct is used when no presupposition is intended, p. 205), yet yi- appears. On recheck, review 2’s own NX-038 (“A boy kissed the girl.”) also has yi- with an indefinite S. The definiteness clause of H9 is contradicted. H9′ drops it. Review 2 used that clause to exclude NX-055 (an impersonal) on the assumption of H10b. Under H9′ that account works only if the impersonal has no third-person co-argument (H10a, or an implicit agent that is not third person). That is now OPEN again.
Does light d earn its keep?
GRANT'S CONJECTURE Light d takes a DP argument, as v takes VP. Only head d joins the verb word; d stays in place while the DP moves where needed.
HYPOTHESISIdle again; no counter-evidence. Each case where it might have done work:
(i) NX-081:bi- stays on the postposition while the noun ends up far from it. This is exactly “d stays, DP moves”, but it also fits Agree on the postposition and Willie’s pronoun-plus-adjunct analysis. It does not decide between them.
(ii) *NX-064: a d with a single DP complement gives one phrase per participant. Theta biuniqueness already gives that, so d adds nothing.
(iii) Definiteness: the inverse needs a definite or presuppositional raised O (H6′). Yet the verb-word marker is the same bi- for a presuppositional wh-word (NX-068) and for a name (NX-069’s John). So whatever carries definiteness is outside the verb word. This counts against putting definiteness on d, but not against light d as such.
(iv) Q: the question clitic is always outside the verb word, which every rival also predicts.
A decisive test would need a verb-word element that changes with the features of the outside phrase (wh vs name, indefinite vs definite) in a way that tracks a d head. No such alternation is printed in this chapter.
Telicity test
OPEN Not supported and not contradicted. The verbs are work, kiss, see, give, talk to and kick. No pair differs only in having an endpoint, and no question/non-question pair holds the verb constant while changing telicity. Glosses like “3-working” vs “3.3.kissed” mark viewpoint and tense, and viewpoint is not a telicity test. No telicity tag has been assigned. A possible test, still OPEN: does a wh-question about T in an endpoint verb (give, NX-076) allow the same fronting (H14) as with an activity verb? That would need a printed pair with matched animacy.
Question × polarity × voice grid (cumulative after review 3)
SOURCED Chapter 8 has no doo … da, no polar question and no passive or impersonal verb. A search of the extracted chapter text finds doo only inside yidooloh. The negative and passive cells stay OPEN; they are not NONE. Where to look next: the Navajo Questions resource (polar and content pages, and its negation links), Willie & Jelinek 2000 (57)–(60), and Willie Chapter 4 (negative declaratives on p. 119; not questions).
HYPOTHESISWhat the contrasts between cells force:
(a) Wh-affirmative, direct vs inverse: the wh-word can stay in its merged position in both voices (NX-067, NX-069). No voice requires overt wh-fronting.
(b) Within the inverse, polar vs wh: the H3-raised O can host polar -ísh (NX-053) but cannot host a displaced wh =sh (*NX-086). So polar Q and displaced wh-Q have different derivations. A single “Q after the first word” rule is contradicted again; review 2’s Q-pf already failed on NX-052. Polar Q fits being merged at the edge (H8 Q-edge), and wh-Q fits arriving there from below (H12-move with H13).
(c) Order of operations: inverse Move happens before displaced-Q placement (derivation 12).
(d) Not yet testable: any Q > NEG or NEG > Q order, and any interaction between questions and passive voice.
Agreement and disagreement with Willie
HYPOTHESISAgree: wh-words take the position a noun with the same role would take, in both voices (every pair above). The inverse wh-word is presuppositional (NX-068). The direct/inverse clitic contrast (NX-085/086) is real, and it is the most informative pair in the chapter.
HYPOTHESISDisagree: “no wh-movement” is too strong as a description of the surface. On her own analysis, NX-074 is focus movement of the wh-word, and NX-068 is a wh-word in the inverse’s raised position. The supportable claim is that overt wh-fronting is not obligatory. Whether there is also covert LF movement cannot be tested with these data. Her focus explanation of *NX-086 is too strong unless it is restricted to the host of =sh (derivation 11).
HYPOTHESISNot inherited: treating wh-words as adjuncts versus as phrases with theta roles. Every derivation above runs with the wh-phrase as the bearer of the role. Nothing here forces the adjunct view, and nothing refutes it.
What this changes in reviews 1–2
HYPOTHESIS H3 and H4′ pass an out-of-sample test with wh-words. H3 extends to goals (NX-077) and postposition objects (NX-081), refined as H3-loc.
HYPOTHESIS H6 becomes H6′ (definite or presuppositional). H9 loses its definiteness clause (NX-066, NX-075, and review 2’s own NX-038), so review 2’s account of NX-055 is OPEN again.
HYPOTHESIS H7-type fronting is not limited to the inverse: a wh-O can front in the direct (H14, NX-074). Review 2’s question of “why the direct lacks H7” now applies only to plain nouns (NX-050).
HYPOTHESIS H8 Q-pf is contradicted a second time (NX-086).
OPEN predictions
OPENH13-int vs H13-host, test 1 (negation): intervention accounts treat negation as a typical blocker. So H13-int predicts that a displaced =sh (as in NX-085) should fail when doo stands between the first word and the wh-word, but be fine with =sh left on the wh-word. H13-host predicts no such effect. This directly targets the wh-negative cells. Search Navajo Questions and Young & Morgan for printed examples; construct nothing.
OPENH13 test 2 (direct with a fronted noun): H13-int predicts that displacing =sh across any fronted phrase fails, in the direct too. H13-host predicts that only the inverse’s raised O matters.
OPENH14 controlled pair: the same verb with the same animacy gap, a plain-noun O vs a wh-O, fronted in the direct. If only the wh-O can front, H14 is specific to wh-phrases.
OPENPassive questions: an agentive-passive (ʼadi-) or impersonal verb with a wh-word questioning the patient, and with a displaced =sh. H13-int predicts that a passive’s patient, if it raises, blocks displacement as the inverse O does. If it does not raise, there is no block.
OPENOne wh per clause (NX-087/088): is háa=di, printed without =sh, an indefinite? Check Willie (26a), p. 214, hádáaʼshąʼ, which carries a clitic, and the Navajo Questions content page.
OPENH3-loc: the inverse should never raise T over G in a ditransitive. Look for a printed counterexample.
Next in the queue: the Navajo Questions resource (#4) for polar-active, negative and passive question cells, then Willie Chapter 3 (58)–(69) on scope.
2026-11-11 · Questions, negation and inverse: the matched-test audit
2026-11-14: the “On the spine” summary at the top of this page supersedes this section’s framing. Its data, judgments and citations are unchanged.
Restated on the unified structure (2026-11-15). This section’s results now sit on the one structure for Navajo: unified structure, §2. Its restatement as derivations is in Results on the spine. Data, judgments and citations below are unchanged.
SOURCED Canonical inverse polar and WH questions exist, but Willie’s non-WH clitic-host contrast is voice-sensitive. Negative inverse and matrix-passive question cells remain OPEN. See the eight-cell grid, LF-checked rival derivations and verb-movement tests. No unique headedness boundary is established.
2026-11-11 · First-principles review 4: the Navajo Questions database on negative, alternative and passive-glossed questions
2026-11-14: the “On the spine” summary at the top of this page supersedes this section’s framing. Its data, judgments and citations are unchanged.
Restated on the unified structure (2026-11-15). This section’s results now sit on the one structure for Navajo: unified structure, question and negation rows. Its restatement as derivations is in Results on the spine. Data, judgments and citations below are unchanged.
Source: Fernald et al., Navajo Questions: grammar and examples, version 3.2022 (Navajo Language Academy): the yes/no, content and alternative-question pages and the searchable example database behind them. We queried the database on 2026-11-11 (about 1 pm PT) and checked every form below against the saved permalink HTML. New index entries are NX-203–217. The numbered page examples are already indexed as NX-114–116 and NX-119–139, and the grammar-draft negatives as NX-144–161 and NX-184, for the headedness audit. They are cited here, not indexed again. This review follows review 3, and H-numbers continue from there.
2026-11-11 audit pass 3. Five GRANT’S CONJECTURE badges were styled as HYPOTHESIS; they now use the conjecture style. The trees here passed the new c-command, HMC and LF-scope checks, and doo…da remains one clause-wrapping negation. See the pass 3 report.
2026-11-13 audit pass 7. Grant’s instruction (2026-11-13): VPISH is not a word-order account; it frames the verb’s arguments so that syntax and semantics line up. In ordinary tensed clauses, V→v→T and subject raising Spec,vP→Spec,TP (on to Spec,CP when needed) are the working default, not GRANT’S CONJECTURE to be tested against a no-movement rival. This review was posted before that instruction. Its derivations, labels and verdicts on verb movement are rewritten here to use the default: the former no-movement rival is kept only as labelled contrast, each derivation says what else has to move, and examples that seemed to resist the default are kept as tests of it (OPEN). No Navajo form, gloss, judgment or citation changed. See the 2026-11-13 note and the pass 7 report.
HYPOTHESISPlain-words verdict. Navajo negation doo … da (or tʼáadoo … da) wraps one clause and means one “not”. In “cold or not?” it wraps only the second half. Yes/no and either/or markers can sit on doo or inside the frame. In every negative why-question, though, the question word and its marker sit outside the frame, to its left, and scope over it. Questions with a verb the source glosses as passive work like ordinary questions with the doer left unsaid. None of these strings shows the verb moving, and none rules that out.
Starting point
Assumed, by instruction: VPISH (the subject argument first merges in Spec,vP) and theta roles; since audit pass 7 (2026-11-13), also the V→v→T working default stated below. HYPOTHESIS Everything else is tested as a hypothesis, not taken as a result. That includes H1–H14 from reviews 1–3, light d, NegP/C/T head positions and head direction, Agree. Since audit pass 7 (2026-11-13), V→v→T and subject raising Spec,vP→Spec,TP (on to Spec,CP when needed) are the working default in tensed clauses, assumed rather than tested: VPISH frames where the arguments start and get their roles; it is not a word-order account. An example that seems to resist the default is kept and flagged OPEN. Under the default, adverbs like “often” are pied-piped with VP, and only sentential (not constituent) negation counts in verb-movement tests. Every derivation ends in a logical form (LF), and a rival fails if its LF contradicts a reported reading. The database’s analyses appear only as labeled contrast.
The source’s claims, as theirs
SOURCEDYes/no questions.daʼ occurs only at the beginning (NX-114 vs *NX-120–122). -ísh occurs once per question (*NX-126). On the first phrase it gives a “neutral” question. Elsewhere it focuses its phrase (Yes/No page §2, NX-127–131).
SOURCEDContent questions. Content questions use the enclitic -sh, which “is not the same as the yes/no question enclitic -ísh”. Question particles “frequently appear just after the h-word … although they can appear in other places” (Content page, below (4)–(5)).
SOURCEDAlternative questions. They “contain doodaiiʼ (meaning ‘or’) and -ísh or daatsʼí”, with doodago also possible. The “two conjoined constituents can be marked by -ísh or by daatsʼí”, and “there is a presupposition that exactly one of the alternatives is true” (Alternative Questions page).
SOURCEDGlosses. The glossing page lists NEG “negative” and Q “question particle or enclitic”. It has no entry for “pass”, but individual database entries do gloss verbs “pass”. We keep that label as the source’s and do not test it as a category.
What was searched (non-exhaustive)
SOURCED We pulled all 20 question-expression lists (yes/no and content), the negation and “or” particle lists (doo, da, doo … da, tʼah doo … da, doo ndi … da, doodaiiʼ, doodago) and the buy/sell verb list, and ran about 65 text and gloss searches (NEG, neg, pass, doo, …). We parsed 798 distinct entries. The database caps any list at 50 entries, and the -ísh and -sh lists hit that cap, so an empty result is OPEN, not NONE. Questions with negation or a “pass” gloss were then kept for checking.
The checked data
SOURCED Exact forms, with database entry numbers. Glosses, speakers and printed inconsistencies are at the linked index entries.
Polar, affirmative, active:NX-203Daʼ Ron saad neiiłkaah? “Is Ron studying words?” · NX-204Daʼ Ron saad nináánéíłkaah yaʼ? “Ron’s investigating words again?”
Alternative “X or not X” (a negative disjunct):NX-205Nidlóóhísh doodaiiʼ dooísh nidlóoh da? “Are you cold or not?” · NX-206Nił hǫ́zhǫ́ǫ́sh doodaiiʼ doósh nił hózhǫ́ǫ da? “Are you happy or not happy?” · NX-207Nidlóóhísh doodaiiʼ doo daatsʼí nidlóoh da? · *NX-208Nidlóóhísh doodaiiʼ doo shaʼshin nidlóoh da? · NX-209Nidlóóhísh ʼéí doodago doo daatsʼí nidlóoh da?
WH, negative:NX-210Haʼátʼéegoshąʼ tʼáadoo ʼazeeʼálʼį́į́góó nisíníyáa da? “Why didn’t you go to the hospital?” (extends NX-138) · NX-211Haa léíʼ yitʼéehgo nichidí yichxǫʼígíí doo baa níniʼ da? “Why are you not worried about your car that is ruined?”
WH with a negative word, not a negative clause:NX-212Háish dooda ní? “Who said no?”
WH with a verb glossed “pass”:NX-213Háadi lá dibé nidahaniih? “Where are sheep being sold?” · NX-215Naʼazhǫǫshgi díkwíí shį́į́ shaa woobįįh? “How much will I lose at the hoop and pole game?” (the source’s note on this entry reads “my things will be won (by someone else)”)
Declarative controls, not questions:NX-214… Díísh jį́įdi ʼéiyá kindi nidahaniih. “Now, corn is bought at the store.” (same verb form as NX-213) · NX-216… naaltsoos tʼáadoo niigį́į da. “… the mail had not been delivered.” (negative, glossed “pass”)
Candidate only:NX-217… nihichʼįʼ nidaʼiilyé, yaʼ? “… we will be paid, right?” This entry glosses the verb “1pl-payment.I”, not “pass”, so it does not fill a polar × passive cell.
SOURCED Two strings repeat earlier entries from a different source: database entry 694 = NX-145, Doo bił bééhózin da yaʼ?, and database entry 2165 = NX-152, Díídóʼ doo ákótʼée daásh łí?. They are not indexed again. OPENConflict between sources: the grammar draft stars daʼ with the tag yaʼ (*NX-147, Daʼ dibé nee hólǫ́, yaʼ?, printed with a comma). Yet NX-204 has both, recorded unstarred for two speakers (entries 845 and 846, no comma, a different verb). We keep both reports and do not reconcile them.
Hypotheses tested, revised or added
N1 · one negation per frame (new).doo … da (and tʼáadoo … da) contributes exactly one ¬. Its two pieces stand in a dependency; they are not two operators. Which piece is the head is OPEN: doo in left Spec,NegP with final Neg⁰ da, or the reverse concord, as in the headedness rivals.
N2 · scope by containment (new). ¬ takes scope over exactly the material between doo and da and what that material c-commands. It does not scope over material pronounced outside the frame.
ALT · alternative marking (new). In an alternative question, each alternative carries a marker (-ísh or daatsʼí, per the source). A single question operator, C[alt], above the disjunction asks which alternative is true. So the -ísh on doo in NX-205 marks the negative alternative; it is not the clause-typing Q.
H11, H12, H13-int, H13-host (review 3). These are tested again below against negation, a modal and passive-glossed verbs.
H15 · WH over NEG (new). A reason or manner wh-adjunct together with its Q is merged (or ends up) above NegP, and it takes scope over ¬.
H16 · passive-glossed verbs (new).H16a: a silent implicit-agent DP is merged in Spec,vP (VPISH) and existentially closed. H16b: no agent is merged in syntax, and the verb’s meaning adds ∃x Agent. The theme merges as the complement of V in both.
Working default since audit pass 7 (when first posted: GRANT’S CONJECTURE, tested, not assumed): V→v→T, and the Spec,vP DP raises to Spec,TP. Sentential negation is kept apart from constituent and lexical negation.
Derivations, step by step
HYPOTHESIS Each derivation is given as Merge with theta assignment, then Agree (stated, or “none needed”), then Move, then which copy is pronounced, then what is spelled out inside the verb word versus outside it, and finally the LF. Extra assumptions are marked. “Q” is the question element. The inflected verb is not segmented.
NX-203 (polar, affirmative, active). Merge saad “words” as the complement of V “study”, which assigns it Theme. Merge v. Merge Ron in left Spec,vP, where it gets Agent (VPISH). Working default: V→v→T, and Ron moves to Spec,TP, leaving a silent lower copy. With a final T⁰ (HYPOTHESIS) this gives Ron saad V, with nothing else moving. Merge C with daʼ. Extra assumption (HYPOTHESIS): daʼ spells out a head-initial C⁰ or sits in Spec,CP as an operator. Its initial-only position is SOURCED (*NX-120–122). Agree: none needed. Spell-out: daʼ, Ron and saad are outside the verb word, and the person marking for both participants is inside it. LF: ASK_POL(∃e[study(e) ∧ Agent(e,Ron) ∧ Theme(e,words)]). Result: the polar question keeps the declarative order S O V; nothing beyond the default has to move. The cell is filled, and nothing in it resists the default.
NX-204 (daʼ + yaʼ). This has the same Merge and theta steps as NX-203 (with the source’s verb “investigate … again”). If yaʼ selected a declarative and daʼ typed the clause as interrogative, the two would compete, and the string should fail. It is printed unstarred. Extra assumption (HYPOTHESIS): daʼ is at the left edge of CP, and yaʼ is a right-adjoined discourse tag above it, so they do not compete for one position. LF: ASK_POLbiased(∃e[investigate(e) ∧ again(e) ∧ Agent(e,Ron) ∧ Theme(e,words)]). Whether the bias comes from yaʼ is OPEN.
NX-205 (alternative “cold or not”).Disjunct 1: merge the experiencer (second person, no separate pronoun printed) in Spec,vP of “be cold”. Spell-out: the verb word, with -ísh written on it (ALT marker). Disjunct 2: the same vP. Merge NegP over TP, with doo in left Spec,NegP and final Neg⁰ da (N1; head assignment HYPOTHESIS). Write -ísh on doo (ALT marker, by Agree with C[alt]; HYPOTHESIS). Join the two disjuncts with doodaiiʼ. The category of the coordinator is OPEN and is not drawn. Merge C[alt] above the coordination. Move (working default): V→v→T and the experiencer moves to Spec,TP inside each disjunct; nothing else moves. LF: ASK_ALT({p, ¬p}), where p = ∃s[cold(s) ∧ Exp(s,you)]. Rival LFs that fail: (i) If doo and da were two ¬s, disjunct 2 would be ¬¬p = p, giving “cold or cold?”. The reported “Are you cold or not?” contradicts this, so N1 holds. (ii) If ¬ scoped over the whole disjunction, ¬(p ∨ …), we would get no “or not” alternative. This is also contradicted, so N2 holds. The surviving LF puts ¬ inside disjunct 2 and below C[alt], which is exactly where doo … da sits in the c-command structure.
Analysis A: Working default: V→v→T, subject Spec,vP→Spec,TP, under NegP Hypothesis
(1) The negative disjunct of an alternative question: dooísh nidlóoh da (NX-205) HypothesisNX-205, Navajo Questions database entry 804: Nidlóóhísh doodaiiʼ dooísh nidlóoh da? 'Are you cold or not?'. Only the second disjunct is drawn. Leaves doo, nidlóoh, da are printed tokens of that string; the split of dooísh into doo + Q is our analysis. AdvP, Neg⁰ and C⁰[Q] are our category assignments. The experiencer DP is a placeholder: no separate pronoun is printed.Bracket notation (copy)
DP (id S-high; landing site of Move from subject): experiencer: you
T′
vP
DP (id subject; struck copy; moved to S-high): experiencer: you
v′
v⁰ (head; id v; head-moves to T)
VP
V′
V⁰ (head; id V; head-moves to v): nidlóoh
T⁰ (head; id T; hosts the “verb word” (V + v + T))
Neg⁰ (head; id Neg): da
C⁰ (head; id C; [Q, alternative]; probe for Agree Q-host with host)
NX-206 (“happy or not happy”). The participant is the object of the postposition nił “2-with”. Merge 2nd person as the complement of P, then merge the PP inside VP with V “be happy”. Whether this clause has any Spec,vP DP at all is OPEN: VPISH fixes where a subject starts, not that every clause has one. In disjunct 1, the PP and verb carry -ísh on the verb. In disjunct 2, doósh is followed by the PP nił, the verb and da. So the PP+V unit sits inside the negative frame. LF: ASK_ALT({q, ¬q}), where q = ∃s[happy(s) ∧ Exp(s,you)]. HYPOTHESIS Under the working default, a PP + verb unit under NEG comes out either way: the PP stays in VP while V raises to a final T⁰, or the remnant VP is pied-piped. This pair does not decide between those two (OPEN).
NX-207 / *NX-208 / NX-209 (a “perhaps” word after doo). The Merge steps are the same as for NX-205, but with no -ísh in disjunct 2. In its place, a “perhaps” word is merged between doo and the verb. daatsʼí is fine (NX-207, NX-209); shaʼshin is starred (NX-208). ALT-1 (HYPOTHESIS): only -ísh and daatsʼí are ALT markers (the source’s own list), so in *NX-208 the second alternative is unmarked. ALT-2:shaʼshin cannot take scope under ¬. The tag-question data do not help here. The grammar draft rejects the tag yaʼ with daatsʼí and with shaʼshin alike (*NX-149, *NX-151), so the tag treats the two words the same way. ALT-1 versus ALT-2 stays OPEN. Either way, an ALT/Q-type marker can be pronounced inside the negative frame, to the right of doo. LF for NX-207 is the same as for NX-205. Under ALT-1, daatsʼí marks an alternative and is not a modal on ¬p.
The key test from review 3: H13-int and negation. H13-int predicts that a displaced content =sh fails if doo stands between its host and the wh-word. (a) In NX-138, NX-210 and NX-211 the wh-word and its Q are left of the negative frame, and the Q stays on the wh-word. There is no displacement, so these are compatible with H13-int and H13-host alike. (b) We found no printed wh-question, starred or unstarred, in which content =sh is displaced to a host on the far side of doo. So H13-int’s negation prediction remains OPEN (not NONE: the lists are capped). (c) Polar and alternative markers do occur inside the frame: on doo in NX-205/206 (and NX-153), and after doo in NX-207/209 and NX-184. Result: if H13-int were a general intervention condition on every Q-type enclitic, doo would be an intervener for these, and it is not. That general version is contradicted. A version restricted to content =sh, which the source distinguishes from -ísh, survives untested. H13-host is unaffected.
NX-210 (WHY over tʼáadoo … da). Merge the goal PP ʼazeeʼálʼį́į́góó “hospital-to” inside VP with V “go”, which assigns Goal. Merge the 2nd-person mover in Spec,vP (VPISH). Merge NegP over TP: tʼáadoo in Spec, da in Neg⁰ (N1). Merge the reason phrase haʼátʼéego+Q (shąʼ) above NegP (H15). Base-merged at the CP edge or moved there from a lower adjunct site: both are HYPOTHESIS, and the string does not choose. Agree: C[Q, wh] with the reason phrase. Spell-out outside the verb word: wh+shąʼ, tʼáadoo, PP, then the verb word, then da. LF: ?r. REASON(r, ¬∃e[go(e) ∧ Mover(e,you) ∧ Goal(e,hospital)]). Rival that fails: ¬ over WHY, “it is not for reason r that you went to the hospital”, presupposes that you went, which contradicts “Why didn’t you go”. So WHY > ¬ is forced by the reading, not imposed, and it matches c-command (wh left of and above the frame). Working default: with V→v→T and a final Neg⁰, the verb ends up right of the PP and left of da, as printed, with nothing else moving. (Labelled contrast: no movement, or V→Neg, gives the same string.)
NX-211 (a nominal outside the frame). Merge the relative phrase “your car that is ruined” (overt nominalizer, source-glossed NOM) as the complement of P -aa “about”. Merge PP + V “be on one’s mind / worry” inside VP and the 2nd-person experiencer in Spec,vP. Merge NegP. The DP is pronounced left of doo, but its P baa (written with b-) is inside the frame. Option (i): move the DP out of the PP across NegP and pronounce the upper copy. That is the same kind of extraction from a PP that H3-loc needed for NX-081. Option (ii): base-merge the DP high, linked to the b- on the P. Both are HYPOTHESIS. Merge the why-phrase on the left. The source’s alignment of its three gloss words is unclear, so no internal analysis of it is offered. LF: ?r. REASON(r, ¬∃s[worry(s) ∧ Exp(s,you) ∧ Theme(s, ιx[car(x) ∧ yours(x) ∧ ruined(x)])]). The ruin of the car is presupposed, not negated, as in the translation. A DP outside the frame is outside ¬ (N2), so the c-command structure and the LF agree. This is sentential negation of the whole “worry” clause, not “not the car”.
NX-212 (lexical “no”: constituent, not sentential). Merge dooda “no” as the complement of V “say”, where it gets Content/Theme, the thing said. Merge háí+Q in Spec,vP as Agent. There is no da after the verb, so there is no frame. The source glosses the verb as fourth person, which is OPEN in a question about the agent. LF: ?x.∃e[say(e) ∧ Agent(e,x) ∧ Content(e,“no”)]. Rival that fails: sentential negation, ?x.¬∃e[say(e) ∧ Agent(e,x)] “Who didn’t speak?”, contradicts the translation “Who said no?”. GRANT’S CONJECTURE This is a Navajo case of a negative element inside the VP that does not negate the clause, like English “kisses not Mary”. By Grant’s split it does not count in any verb-movement test.
NX-213 (WH with a verb glossed “pass”), with NX-214 as control. Merge dibé “sheep” as the complement of V “sell/buy”, which assigns Theme. Then either H16a (merge a silent implicit agent in Spec,vP) or H16b (merge no agent in syntax). Left-adjoin háadi “where” to VP. Merge C[Q, wh] with lá. In both panels below, V→v→T and the silent agent moves to Spec,TP (the working default). In Panel A, háadi also moves to Spec,CP and lá is an initial C⁰. In Panel B, háadi stays where it was merged and lá is written after its host. Agree: C with the wh-phrase. The plural marking (source “pl”) might agree with the theme or with the implicit agents, and which one is OPEN. Spell-out outside the verb word: háadi lá dibé; inside it, the plural, “pass” and the stem. LF: ?λl.∃x∃e[sell(e) ∧ Agent(e,x) ∧ Theme(e,sheep) ∧ Loc(e,l)]. Where takes widest scope, and the agent is existential below it, never questioned and never named. Control: NX-214 puts the non-wh locative kindi “store-at” in the same slot before the same verb form, so the wh-word again sits where a non-wh phrase of the same role sits. H3/H4′’s “wh where nouns sit” therefore extends to a passive-glossed clause. The theme is silent in NX-214, though, so the relative order of locative and theme is not controlled (OPEN).
Analysis A: Working default (passive: implicit agent to Spec,TP) plus wh to Spec,CP Hypothesis
Analysis B: Working default (V→v→T, implicit agent to Spec,TP); wh-PP in situ, lá written on its host Hypothesis
(2) A passive-glossed wh-question: Háadi lá dibé nidahaniih? (NX-213) HypothesisNX-213, Navajo Questions database entry 2199, source gloss where-at Q sheep pl-3-pass-sell.CI, 'Where are sheep being sold?'. Leaves are printed tokens. The silent implicit-agent DP, C⁰ and the VP-adjunct site of háadi are our hypotheses. A rival with NO agent in syntax (argument-reducing passive) is discussed in prose; it is not drawn because the site's tree checks require a left DP in Spec,vP.Bracket notation (copy)
PP (id wh-high; landing site of Move from wh): háadi
C′
C⁰ (head; id C; [Q, wh]; probe for Agree Q/wh with wh): lá
TP
DP (id S-high; landing site of Move from agent): implicit agent (silent)
T′
vP
DP (id agent; struck copy; moved to S-high): implicit agent (silent)
v′
v⁰ (head; id v; head-moves to T)
VP
PP (id wh; struck copy; moved to wh-high; goal of Agree Q/wh): háadi
VP
V′
DP (id theme): dibé
V⁰ (head; id V; head-moves to v): nidahaniih
T⁰ (head; id T; hosts the “verb word” (V + v + T))
Analysis B
CP
C′
TP
DP (id S-high; landing site of Move from agent): implicit agent (silent)
T′
vP
DP (id agent; struck copy; moved to S-high): implicit agent (silent)
v′
v⁰ (head; id v; head-moves to T)
VP
PP (id wh; goal of Agree Q/wh (in situ)): háadi lá
VP
V′
DP (id theme): dibé
V⁰ (head; id V; head-moves to v): nidahaniih
T⁰ (head; id T; hosts the “verb word” (V + v + T))
C⁰ (head; id C; [Q, wh]; probe for Agree Q/wh (in situ) with wh)
Not drawn: H16b with no Spec,vP DP. It gives the same string and, with ∃x supplied lexically, the same truth conditions. The site’s tree checks require a left Spec,vP DP, so the H16b panel is shown only in prose.
NX-215 (“how much will I lose”). Merge díkwíí “how much” as Theme of “win”. Merge the 1st person as the complement of P -aa (shaa), so the loser is a P object. The winner is an implicit agent. The source’s lexical note, “(by someone else)”, supports an agent entailment, which both H16a and H16b supply, so it does not choose between them. LF: ?n.∃x∃e[win(e) ∧ Agent(e,x) ∧ Theme(e, my things of amount n) ∧ From(e,me)]. Tension with H11: review 3 used Willie’s report that a wh-word with a modal clitic and no =sh is an indefinite (NX-084). Here díkwíí shį́į́ “how-much perhaps”, with no =sh and no lá, is still a question. Either H11’s “modal ⇒ indefinite” clause is too strong, or díkwíí, which the source says is not an h-word, behaves differently. OPEN, but H11 as stated is no longer safe.
NX-216 (negative passive declarative, a building block only). Merge naaltsoos “mail” as Theme. The agent is implicit (H16a/b). Merge NegP (tʼáadoo … da). The theme is pronounced left of tʼáadoo. Either it moved above NegP (HYPOTHESIS), or NegP is lower than its landing site. LF: ¬∃e∃y[deliver(e) ∧ Agent(e,y) ∧ Theme(e, ιx mail(x))], with “the mail” definite and presupposed outside ¬, as in the translation. This fits a theme above the frame (N2). It is not a question and fills no question cell.
Does light d earn its keep?
HYPOTHESISFair test. A light-d account would put a silent d on every argument nominal and could locate definiteness or genericity there. The test data: bare dibé is generic plural in NX-213; bare naaltsoos is definite in NX-216 (“the mail”), and it is also outside ¬. The relative in NX-211 ends in overt -ígíí, glossed NOM by the source. Result: idle. Each scope fact follows from where the nominal is pronounced relative to the negative frame (N2), with or without d. The overt -ígíí is D-like material we can see, not evidence for a silent d on bare nouns. No datum here needs light d, and none counts against it.
Telicity test
OPEN No minimal pair in these data differs only in telicity. Variation in mode labels across speakers (the source glosses “cold” .X in one entry and .I in another, NX-205) is a difference in glossing, not a contrast in form. Imperfective or perfective mode is not a telicity diagnostic. No telic/atelic surface effect is supported or contradicted by this batch.
Question × polarity × voice grid (cumulative after review 4)
(a) Polar/alternative versus wh under negation. Polar and alternative markers can sit inside the frame or on doo (NX-205–209, NX-153, NX-184). In all three negative wh-questions, the wh-word and its Q are outside the frame and to its left (NX-138, NX-210, NX-211). Because nothing starred shows a wh-word inside the frame, “wh must be outside” is observed, not forced (OPEN).
(b) Readings that force scope. One ¬ per frame (NX-205 “or not”). ¬ contained in its disjunct (NX-205). WHY > ¬ (NX-210). A definite nominal outside the frame lies outside ¬ (NX-211, NX-216). In every case the scope order the reading requires matches c-command.
(c) Sentential versus lexical negation.dooda as the thing said (NX-212) is not sentential negation and is set aside for verb-movement tests.
(d) Passive-glossed wh-questions question a location or an amount, never the doer, and have no by-phrase. Q takes scope over the existential agent.
(e) Order of operations. No new ordering result. Review 3’s “inverse Move before displaced-Q placement” is untouched, because there are still no negative or passive inverse questions.
(f) HYPOTHESISWorking default, verdict for this batch: every derivation above goes through with V→v→T and the subject in Spec,TP (the figures now draw it in every panel). Nothing here resists it, and with final T⁰ nothing else has to move except the wh-phrase in panel A of the passive-glossed figure.
Where we part from the source’s descriptions (labeled contrast)
HYPOTHESISFits:daʼ initial only. Alternatives are marked by -ísh or daatsʼí: NX-205–209 all fit, and *NX-208 is the predicted failure if shaʼshin is not a marker. -sh and -ísh are distinct, and our narrowing of H13-int depends on that distinction.
HYPOTHESISGoes beyond the source: the source does not say where negation sits relative to question particles. The containment and scope results (N1, N2, H15) are ours.
OPENConflicting reports:daʼ + yaʼ: unstarred in NX-204, starred in NX-147. The two strings differ in the verb and in the comma, so the cause is OPEN.
What this changes in reviews 1–3
HYPOTHESIS H13-int is narrowed to content =sh. Its general version is contradicted by polar and alternative markers inside the negative frame.
HYPOTHESIS H11’s “a modal clitic makes the wh-word indefinite” clause is under pressure from NX-215 (OPEN).
HYPOTHESIS H3/H4′’s “wh where nouns sit” extends to a passive-glossed clause (NX-213 versus NX-214).
HYPOTHESIS Light d is still idle, and telicity is still OPEN.
OPEN predictions
OPENH13-int, negation test (still the key one): a content =sh displaced to a first word that is separated from the wh-word by doo. Search Young & Morgan 1987 and the grammar draft’s content-question chapter for printed examples. Construct nothing.
OPENALT-1 versus ALT-2: an alternative question whose negative disjunct has no marker at all. ALT-1 predicts it is starred.
OPENPlural in nidahaniih: does it agree with the theme or with the implicit agents? Look for a singular theme with the same verb.
OPENPolar × passive, negative × passive questions, and negative inverse questions: all three remain empty. NX-216 (negative passive declarative) and NX-217 (a candidate without a “pass” gloss) are the nearest material.
OPENH11:díkwíí shį́į́ (interrogative, NX-215) versus háí-shį́į́ (indefinite, NX-084). Is the difference in the wh-word, the speaker or the source?
2026-11-11 (evening) · First-principles review 5: Hale & Platero on negative polarity — doo … da, h-word indefinites and the displaced -í-
2026-11-14: the “On the spine” summary at the top of this page supersedes this section’s framing. Its data, judgments and citations are unchanged.
Restated on the unified structure (2026-11-15). This section’s results now sit on the one structure for Navajo: unified structure, sentential-negation row. Its restatement as derivations is in Results on the spine. Data, judgments and citations below are unchanged.
Source: Ken Hale & Paul Platero, “Negative polarity expressions in Navajo”, MIT manuscript, undated (hale014.pdf, 17 pp., from MIT’s Ken Hale papers page). Judgments are credited to Linda Platero (ms. fn. 1). The published version (in Fernald & Platero, eds., The Athabaskan Languages, Oxford University Press, 2000, pp. 73–91) was not opened, so the manuscript is cited by PDF page (it prints no page numbers). Every form below was checked against 200-dpi renders of those pages. New index entries are NX-218–235. This review follows review 4, and H-numbers continue from there. Queue targets that could not be used: Willie & Jelinek 2000 (57)–(60) were opened (printed pp. 272–273, PDF pp. 151–152 of the Jelinek manuscript collection), but they are affirmative direct/inverse wh-questions with no negation, and the same pattern is already indexed from Willie 1991 (review 3). Young & Morgan 1987 is not on hand. The grammar draft’s negation and content-question chapters are on the box from the headedness work (NX-140–161, NX-184), but they were not searched again in this run.
2026-11-13 audit pass 7. Grant’s instruction (2026-11-13): VPISH is not a word-order account; it frames the verb’s arguments so that syntax and semantics line up. In ordinary tensed clauses, V→v→T and subject raising Spec,vP→Spec,TP (on to Spec,CP when needed) are the working default, not GRANT’S CONJECTURE to be tested against a no-movement rival. This review was posted before that instruction. Its derivations, labels and verdicts on verb movement are rewritten here to use the default: the former no-movement rival is kept only as labelled contrast, each derivation says what else has to move, and examples that seemed to resist the default are kept as tests of it (OPEN). No Navajo form, gloss, judgment or citation changed. See the 2026-11-13 note and the pass 7 report.
HYPOTHESISPlain-words verdict. In every example in this paper, doo … da is one sentence-wide “not”. The extra -da on an h-word (“who”, “what”, “where”) is agreement with that “not”, not a second one. It turns the h-word into “any-”, and it has to sit to the right of doo. When the h-word goes missing and a bare -í- shows up on the verb, that is real movement. It obeys islands, and the missing piece must be the first thing after doo. Nothing here needs a silent light d. Every derivation works with the V→v→T working default, and nothing here resists it. The paper has no negative questions, passives or inverse examples that we could use, so the key negative-question test is still OPEN.
Starting point
Assumed, by instruction: VPISH (the subject argument first merges in Spec,vP) and theta roles; since audit pass 7 (2026-11-13), also the V→v→T working default stated below. HYPOTHESIS Everything else is tested: N1/N2/ALT and H1–H16 from reviews 1–4, light d, the NegP head and specifier assignments, Agree. Since audit pass 7 (2026-11-13), V→v→T and subject raising Spec,vP→Spec,TP (on to Spec,CP when needed) are the working default in tensed clauses, assumed rather than tested: VPISH frames where the arguments start and get their roles; it is not a word-order account. An example that seems to resist the default is kept and flagged OPEN. Adverbs are pied-piped with VP, and only sentential negation counts in verb-movement tests. Every derivation ends in a logical form (LF), and a rival fails if its LF contradicts a reading the source reports. The authors’ own analyses (rightward movement in their (12), (16), (26); the parallelism and polysynthesis discussion of pp. 13–16) appear only as labeled contrast.
The source’s claims, as theirs
SOURCEDTwo NPI constructions. (i) doo + h-word + -da, with -da again at the end of the clause (their (1)). (ii) No h-word before the verb, but -í- (“PRN”) written “between the verb and the enclitic -da, the negative scope marker” (their (7), ms. p. 4).
SOURCEDOne set of h-words. The same h-words serve as question words with -lá (their (2)), positive indefinites with a sh… enclitic (their (3), see below) and NPIs with -da (ms. p. 1).
SOURCEDWhere doo goes. Its “more or less neutral” place is just before the verb. It “can also appear initially or following the subject”, but “in the polarity construction … the negative particle must precede the indefinite” (ms. p. 3).
SOURCEDThe gap. The gap of construction (ii) “directly follows the negative particle” (ms. p. 4, with their (9)/(10)). Long displacement is possible out of direct-discourse complements, which have no overt complementizer (their (13)–(15)). It is impossible out of relatives, factives and -go adverbial clauses (their (19d)–(22d)). With an overt h-word in those clauses, the (c)-sentences are “slightly degraded” and the h-word is read as a plain indefinite (ms. p. 10).
SOURCEDPossessor gaps. PRN can correspond to the possessor inside a nominal (their (23)–(25)). The authors argue this bears on whether overt nominals are adjuncts (ms. pp. 12–13). We report that argument here as theirs; we do not test it.
The checked data
SOURCED Exact forms with example numbers and manuscript (= PDF) pages. Source glosses, the printed slips (a missing stem gloss in (1b), (25a)’s gloss repeated under (25b), two spellings of the “fall into mud” verb in (22)) and the crops are at the linked index entries.
Overt NPI:NX-218 (1a), p. 1: Doo háí-da níyáa-da., translated “No one has arrived.” · NX-219 (1b), p. 1: Shi-zhéʼé doo haʼátʼíí-da nayiisniiʼ-da., translated “My father has not bought anything.” · NX-225 (6), p. 3: Doo shi-zhéʼé háá-góó-da deeyáa-da., translated “My father is not going anywhere.” · NX-234 (25b), p. 12: Doo háí-da bi-yeʼ Naʼnízhoozhí-góó deeyáa-da., translated “No one’s son is going to Gallup.”
Question and affirmative controls:NX-220 (2a), p. 1: Háí-lá níyá., translated “Who has arrived?” · NX-221 (2b), p. 1: Shi-zhéʼé haʼátʼíí-lá nayiisniiʼ?, translated “What did my father buy?” · NX-222 (4b), p. 2: Shi-zhéʼé béégashii nayiisniiʼ., translated “My father bought a cow.”
Plain negation:NX-223 (5a), p. 3: Si-tsilí doo níyáa-da., translated “My younger brother did not arrive.” · NX-224 (5b), p. 3: Shi-zhéʼé béégashii doo nayiisniiʼ-da., translated “My father did not buy a cow.”
Displaced PRN:NX-226 (7a), p. 3: Doo níyá(h)í-da., translated “No one has arrived.” · NX-227 (7b), p. 3: Shi-zhéʼé doo nayiisniiʼ-í-da., translated “My father did not buy anything.” · NX-230 (11b), p. 5: Awééʼ doo si-dá(h)-í-gi-da., translated “The baby isn’t sitting anywhere.” · NX-235 (25c), pp. 12–13: Doo ____ bi-yeʼ Naʼnízhoozhí-góó deeyá(h)-í-da., translated “No one’s son is going to Gallup.”
The minimal pair:NX-228 (9), p. 4: Doo béégashii yizloh-í-da., translated “Nobody roped the cow.” The source says it “cannot have the meaning according to which the cow didn’t rope anything”. · NX-229 (10), p. 4: Béégashii doo yizloh-í-da., translated “The cow didn’t rope anything.”
Long displacement and an island:NX-231 (15a), p. 7: Shi-zhéʼé doo ji-deeyá shóʼní(n)-í-góó-da, translated “My father doesn’t think I’m going anywhere.” · NX-232 (22c), p. 10: Doo háí-da hastlʼishítlizh-go hadeeshghaazh-da., translated “I didn’t shout out when someone fell into the mud.” (“slightly degraded”) · *NX-233 (22d), p. 10: Doo ____ hashtlʼishítlizh-go hadeeshghaazh-í-da., starred, no translation.
OPENNot indexed: damaged print. In this PDF the fonts are not embedded, and two Navajo letters come out as substitutes: ł prints as “¬” and nasal hooks print as “²”. This affects (1d), (2d), (3a–d), (4c–d), (5c–d), (7d), (11a), (11c), (13), (14), (17)–(21), (23), (24) and (27)–(30). We do not restore those letters by guesswork, so these examples are used only for what their prose and glosses say. (Some undamaged examples, such as (1c), (2c), (4a), (22a–b) and (25a), are also left out of the index because nothing below depends on them.) That includes the positive indefinites in (3), where the sh… enclitic itself is damaged, and the inverse factive (29b).
Hypotheses tested, revised or added
N1 (review 4), tested again.doo … da contributes exactly one ¬.
H17 · NPI concord (new). The -da on an h-word does not add a ¬. It spells out a feature the h-word gets by Agree with the clause’s Neg⁰. Neg⁰ is the probe. The h-word must be in Neg⁰’s c-command domain, so it must be pronounced to the right of doo. Rival H17-c:doo háí-da is a constituent, a negative quantifier like “nobody”.
H18 · how PRN gets onto the verb (new).H18a: syntactic Move of the PRN phrase (with its postposition, if any) to a position inside Neg’s c-command domain. Its pronunciation as an enclitic on the verb word, inside -da, is then a PF step. H18b: no syntactic Move; PF alone relocates -í-. In trees with specifiers on the left and final heads, no phrase can be pronounced between V and a final Neg⁰ through syntax alone. So every analysis here, the working default included, needs some PF placement. The question is only whether syntax moves PRN first.
H19 · where the PRN gap may be (new).H19-lin: the source’s generalization, that the gap is the first thing after doo. H19-int: PRN cannot be moved across a closer argument DP.
N2 (review 4), tested with an indefinite. ¬ scopes only over material pronounced inside the frame.
Working default since audit pass 7 (when first posted: GRANT’S CONJECTURE, tested, not assumed): V→v→T, and the Spec,vP DP raises to Spec,TP (or Spec,CP when needed). Sentential and constituent negation are kept apart.
Derivations, step by step
HYPOTHESIS Each derivation gives Merge with theta assignment, Agree (or “none”), Move, which copy is pronounced, what is inside the verb word versus outside it, and finally the LF. Extra assumptions are marked. Brackets put specifiers on the left. Heads are final except where a derivation says otherwise. The inflected verb is not segmented beyond the source’s hyphens. For the “go/arrive” verbs we write the mover’s role as Theme. Whether it merges in Spec,vP or as the complement of V is not decided by these data, so it is OPEN.
NX-218 (overt NPI subject). Merge V “arrive” and its one argument háí, which gets Theme. Merge v, then T. Merge NegP above TP, with doo in left Spec,NegP and da as final Neg⁰ (N1; head assignment HYPOTHESIS). Agree (H17): Neg⁰ values [neg] on háí, spelled -da. Move (working default): V→v→T, and háí-da to Spec,TP, still below doo. Spell-out: doo, then háí-da, then the verb word, then -da. LF: ¬∃x∃e[arrive(e) ∧ Theme(e,x)]. Rivals that fail: (i) ∃x¬∃e[…] (“someone didn’t arrive”) contradicts “No one has arrived”. So the h-word must be under ¬, and it is, since doo c-commands it. (ii) H17-c read as two negations, ¬[nobody arrived] = “someone arrived”, also contradicts the translation. So the inner -da adds no ¬ (N1 holds again). Consequence of the working default: since the subject reaches Spec,TP, the NPI subject is still to the right of doo. So NegP must be above TP. That placement is forced by the default, not assumed (its head assignment stays HYPOTHESIS).
NX-225 versus H17-c (sentential or constituent?). Same steps as NX-218, but the goal phrase háá-góó-da “where-AL-DA” merges in VP, and shi-zhéʼé “my father” is the mover. Spell-out puts the overt subject betweendoo and the NPI. If doo + h-word + -da were one constituent (H17-c), doo could not be separated from the h-word. The source prints the separated order as possible. Result: H17-c is contradicted, and doo is clause-level. LF: ¬∃l∃e[go(e) ∧ Theme(e,my_father) ∧ Goal(e,l)]. Sentential versus constituent negation: every doo example in this paper also has the final -da on the verb word and one clause-wide ¬. So all of them are sentential and usable in verb-movement tests. The paper has no constituent negation like English “kisses not Mary”, so for doo … da that half of the split is OPEN. The only non-sentential case on the site is still lexical dooda “no” (NX-212).
NX-226 / NX-227 (displaced PRN, H18). NX-227: merge PRN as the complement of V “buy” (Theme). Merge shi-zhéʼé in Spec,vP (Agent; VPISH). Merge T and NegP. Agree (H17): Neg⁰ with PRN. H18a: PRN moves to a position c-commanded by doo, and its lower copy is silent. At PF, -í- is written on the verb word, inside -da. The subject is pronounced left of doo (see 6). LF: ¬∃y∃e[buy(e) ∧ Agent(e,my_father) ∧ Theme(e,y)]. NX-226 is the same with PRN as the only argument, of “arrive”. Rival that fails without an extra step: the source compares PRN movement with wh-movement to Spec of C. If PRN landed in a Spec,CP abovedoo and were read there, the LF would be ∃x¬… (“someone didn’t arrive”), which contradicts “No one has arrived”. So that rival needs an added assumption (HYPOTHESIS): the LF reads the lower copy. A landing site under doo needs no such step.
NX-230 (a postposition travels with PRN). Merge PRN as the complement of P -gi (LOC). Merge the PP in VP with V “sit”, which assigns Loc. Merge awééʼ “baby” in Spec,vP (Theme). The P is written after PRN on the verb word: si-dá(h)-í-gi-da. So what is displaced is the PP, a phrase, not a lone head (as the authors also argue). LF: ¬∃l∃s[sit(s) ∧ Theme(s,the_baby) ∧ Loc(s,l)]. This counts for “displacement is phrasal” under H18a. Under H18b, PF would have to move a phrase-sized PP, which is an added assumption.
NX-228 / NX-229 (the minimal pair, H19). Our editorial Leipzig gloss (HYPOTHESIS; it follows the source’s hyphens, and reading “3.P.3” as object.perfective.subject is ours): NX-228 NEG cow 3OBJ.PFV.3SBJ.rope-PRN-DA; NX-229 cow NEG 3OBJ.PFV.3SBJ.rope-PRN-DA. NX-228: merge béégashii as the complement of V “rope” (Theme). Merge PRN in Spec,vP (Agent). NegP above. PRN moves (H18a) and is written on the verb word. LF: ¬∃x∃e[rope(e) ∧ Agent(e,x) ∧ Theme(e,the_cow)]. The excluded reading needs the opposite Merge: béégashii in Spec,vP and PRN in VP. With doo above the subject, the string would be the same, Doo béégashii [gap] yizloh-í-da, because the gap is silent. Nothing in VPISH, theta roles, N1 or H17 excludes it, so an extra condition is needed. H19-lin (gap first after doo) excludes it. H19-int also excludes it: the object PRN would cross the closer subject DP. NX-229:béégashii merges in Spec,vP (Agent) and is pronounced left of doo, and the PRN object is the first thing after doo. LF: ¬∃y∃e[rope(e) ∧ Agent(e,the_cow) ∧ Theme(e,y)]. Why an extra condition, and why it is about PRN: with an overt NPI, an overt subject can stand between doo and the NPI (NX-225). So the restriction belongs to the displaced form, not to NPI licensing. Effect on H4′: review 2’s one-noun default (a lone noun is the lower participant, O) fits NX-228 but not NX-229, where the lone noun is the agent. Two repairs are open. Either PRN counts as a participant expression, so these are not one-noun clauses and H4′ does not apply. Or position outside doo overrides the default. The data do not choose (OPEN).
NX-223 / NX-224 (subjects and objects left of doo, N2 and the working default). NX-223: merge si-tsilí “my younger brother” (Theme of “arrive”), then NegP. The subject is pronounced left of doo. Under the working default, NegP is above TP (forced by 1), so the subject must move again, from Spec,TP to a position above NegP (Spec,CP or a topic position). The default allows this “when needed”; this second move is what else has to move (HYPOTHESIS), and what triggers it here is OPEN. Labelled contrast (no subject movement): the subject would stay in Spec,vP, and doo would need a second, lower merge site, below Spec,vP. The source does say doo’s position varies. LF: ¬∃e[arrive(e) ∧ Theme(e,my_younger_brother)]. A definite takes the same truth conditions inside or outside ¬, so this example cannot test scope. NX-224: the object béégashii, translated indefinite (“a cow”), is also left of doo. N2 then predicts ∃y[cow(y) ∧ ¬∃e[buy(e) ∧ Agent(e,my_father) ∧ Theme(e,y)]] (“there is a cow he didn’t buy”). The neutral reading is ¬∃y∃e[…cow(y)…]. The English translation fits both, so the test is OPEN, and N2 is under pressure. Its support in review 4 came only from definites (NX-211, NX-216), which cannot show scope. A rival, N2′ (HYPOTHESIS), puts ¬ at final Neg⁰ da. On N2′, ¬ scopes over the whole clause wherever doo is, and doo only has to precede any NPI (as stated by the source). N2′ derives NX-218–235 equally well. What would decide between them: a printed scope judgment for an indefinite or quantified nominal left of doo.
NX-231 (long displacement). Embedded clause: merge the goal PP (PRN + P -góó) in VP with V “go”, and the fourth-person mover (prefix ji-, no overt nominal) in Spec,vP. No overt C is printed. Matrix: V “regard/think” takes that clause. shi-zhéʼé is the matrix experiencer. NegP is above the matrix TP. H18a: the PP moves out of the complement into the matrix, under doo, and is written on the matrix verb word. LF: ¬∃s[think(s) ∧ Exp(s,my_father) ∧ Theme(s, ∃l go(I,l))]. The translation keeps “anywhere” inside the embedded clause, so the PP is read in or near its lower copy. Whether a wider-scope reading (“no place is such that he thinks I am going there”) is also available is not reported (OPEN). Check on H19-int: the embedded subject is a closer argument than the goal PP, and PRN crosses it. So strict H19-int is contradicted here, unless it is restricted to pronounced DPs (then it says nothing beyond H19-lin) or to DP-over-DP (the moved phrase here is a PP). H19-lin survives every example.
NX-232 / *NX-233 (island). Merge an adverbial clause headed by overt -go (source: COMP), whose subject is the faller, inside the matrix clause of “shout” (Agent: 1st person). NegP is above the matrix TP. *NX-233: PRN would have to move out of the -go clause. The gap is still the first thing after doo, so H19-lin is satisfied, yet the string is starred. So a structural boundary blocks the displacement, which counts for H18a (syntactic Move) over H18b (PF only). H18b survives only if it copies the island into PF, which is an added assumption. NX-232 has the overt h-word inside the -go clause. LF (from the translation): ¬∃t[shout(I,t) ∧ ∃x fall_into_mud(x,t)]. The h-word loses its NPI reading across the same boundary, as the source reports, so H17’s Agree also stops at the overt C. Its scope relative to ¬ beyond what the translation shows is OPEN.
NX-234 / NX-235 (possessor NPI and possessor gap). Merge bi-yeʼ “3-son” with its possessor (háí in NX-234, PRN in NX-235) inside the nominal. The nominal merges in Spec,vP as the mover of “go”, and the goal Naʼnízhoozhí-góó merges in VP. NegP is above. In NX-235 PRN leaves the nominal (H18a) and is written on the verb word. The gap is first after doo, so H19-lin holds. LF: ¬∃x∃e[go(e) ∧ Theme(e, son_of(x)) ∧ Goal(e,Gallup)]. OPENPossible test of the working default, not run: if the subject nominal has already moved to Spec,TP before PRN leaves it, this is extraction from a moved phrase. Many languages bar that (“freezing”). No source here tests freezing in Navajo. Copy theory also lets PRN leave the lower copy first. So NX-235 does not decide.
NX-220 / NX-221 / NX-222 (questions and control). The object h-word haʼátʼíí-lá (NX-221) is pronounced exactly where the object noun béégashii is in NX-222, after the subject and before the same verb. Merge it as the complement of “buy” (Theme), and the question particle -lá with it. Agree: C[Q] with the h-word (HYPOTHESIS). No Move is forced. LF: ?y.∃e[buy(e) ∧ Agent(e,my_father) ∧ Theme(e,y)]. The same h-words appear as NPIs under doo (NX-218, NX-219), so the h-word’s force comes from what it combines with: -lá gives a question, and -da under Neg gives “any-”. This supports H11 (a wh-word is an indefinite plus Q) out of sample. The positive-indefinite third member of that set, (3), is printed with damaged letters and is not indexed.
Does light d earn its keep?
HYPOTHESISFair test. This paper is the best case yet for a determiner-like element. The authors identify PRN -í- with “the morphological base of certain pronouns, determiners, and question words”: their (8) lists sh-í “1st singular”, b-í “3rd”, há-í “who” and é-í “that” (ms. p. 4). A light-d analysis would say that PRN is a bare d (D⁰ with no noun) and that háí is a d with an h- restrictor. Result: the data do not need it, and they do not count against it. (a) Everything above needs only a pronominal phrase that can move and carry a postposition along. Its category (D or N) never matters to any derivation or LF. (b) The shared -í- is overt, D-like material. It is not evidence for a silent d on bare nouns such as béégashii, which is what light d claims. (c) A possessor gap inside a nominal (NX-235) is compatible with a DP layer and with no DP layer alike. So light d stays idle. What is new is a concrete place to look: if -í- is a D, nominals with an overt -í- base should pattern apart from bare nouns in some scope or extraction test. No such contrast is in hand (OPEN).
Telicity test
OPENPrediction tested: the negative-polarity forms differ with the verb’s telicity. In the indexed data, the same pattern, V-(h)-í-(P)-da, appears with verbs translated as telic events (“arrive” NX-226, “buy” NX-227, “rope” NX-228/229) and with a state (“be sitting” NX-230). The visible alternations are the parenthesized (h) and the long vowel before -da (níyáa-da, deeyáa-da). They go with a vowel-final stem: (h) appears with telic níyá and with stative sidá alike. So we found no telicity effect. But the telic/atelic classification comes from the English translations (ours, HYPOTHESIS), and no pair differs only in telicity. The result is therefore neither supported nor contradicted: OPEN.
Question × polarity × voice grid (cumulative after review 5)
(a) No new negative or passive question cell. The paper’s questions are all affirmative and active. Its negatives are declaratives (NX-218–235), and its one inverse example (29b) is glyph-damaged. Footnote 2’s inverse note prints no Navajo form.
(b) The h-word is constant across cells. The same háí and haʼátʼíí sit in the same argument positions as questions (NX-220/221) and as NPIs (NX-218/219). A future negative wh-question with an h-word inside doo … da would therefore have two possible readings for the h-word, “who” or “anyone”. That is a concrete thing to look for in negative questions.
(c) H13-int (review 3): still no printed wh-question with content =sh displaced across doo, so it stays OPEN. Related, but not the same: in this paper, an h-based element (NPI or PRN) is never pronounced left of doo (source statement, p. 3). Both PRN displacement and long displacement stay on the right side of doo. That follows from the NPI having to be under ¬ (H17), so it says nothing about whether doo blocks a question clitic.
(d) HYPOTHESISWorking default, verdict for this batch: every derivation goes through with it. What else has to move: NegP is above TP (from NX-218/226/228), and a subject left of doo makes a second move above NegP (NX-223/229; trigger OPEN). The verb word plus PRN plus -da comes out as printed. Labelled contrast: a no-movement analysis would instead need doo to have two merge sites.
Where we part from the source’s analysis (labeled contrast)
HYPOTHESISAgrees with our derivations: PRN displacement is a phrasal movement that obeys islands (NX-230, *NX-233). The gap is first after doo (NX-228/229).
HYPOTHESISNot adopted: the authors draw PRN moving rightward (their (12), (16), (26)). Our trees allow only left specifiers, so the right-hand position must come from PF placement on the verb word (H18a), with Move itself going up and to the left. Their comparison with wh-movement to Spec of C is adopted only if the LF reads a copy below doo (derivation 3).
OPENTheir wider argument about polysynthesis and whether nominals are adjuncts (pp. 12–16) rests on the glyph-damaged (23)–(24) and (27)–(30), plus NX-235. We do not test it here.
What this changes in reviews 1–4
HYPOTHESIS N1 confirmed out of sample. The inner -da on an h-word is concord (H17), not a second ¬ (NX-218, NX-225).
HYPOTHESIS N2 is weakened to OPEN. Its support was definites only. The rival N2′ (scope from final da) fits everything, and NX-224 is the place to decide.
HYPOTHESIS H4′ needs a repair or a scope limit (NX-229).
HYPOTHESIS H11 gains out-of-sample support (NX-218–221).
HYPOTHESIS Light d is still idle, and telicity is still OPEN.
OPEN predictions
OPENN2 versus N2′: a printed scope judgment for an indefinite or quantified nominal pronounced left of doo, like NX-224.
OPENConstituent negation with doo … da: a printed case where negation applies to one nominal only (the analogue of “kisses not Mary”). None is in this paper.
OPENH13-int, negation test: a content =sh displaced across doo (Young & Morgan 1987; grammar draft content-question chapter). Construct nothing.
OPENNegative wh-questions with an h-word inside the frame: is the h-word read as “who” or as “anyone” (grid note b)?
OPENThe glyph-damaged examples: the published 2000 chapter (pp. 73–91) would give clean forms for (3), (11a, c), (13)–(14), (17)–(21), (23)–(24) and (27)–(30), including the inverse factive (29b).
OPENFreezing: does Navajo bar extraction from a moved subject? If it does, NX-235 becomes a test of Grant’s obligatory Spec,TP raising.
2026-11-12 · First-principles review 6: the grammar draft’s negation chapter and Willie’s negative impersonal — where doo may sit, what -í- is, and constituent negation with hanii
2026-11-14: the “On the spine” summary at the top of this page supersedes this section’s framing. Its data, judgments and citations are unchanged.
Restated on the unified structure (2026-11-15). This section’s results now sit on the one structure for Navajo: unified structure, negation rows. Its restatement as derivations is in Results on the spine. Data, judgments and citations below are unchanged.
Sources: (1) MaryAnn Willie, Navajo Pronouns and Obviation (University of Arizona dissertation, 1991), ch. 4 §4.2.5 “The Impersonal Fourth”, printed p. 119 (PDF p. 125), examples (15)–(16). (2) Ellavina T. Perkins & Theodore B. Fernald, A Grammar of Navajo, working draft (revised 2014), Chapter 6 “Negation” (chapter PDF), printed pp. 88–102 (PDF pp. 9–23). The chapter PDF uses an old Navajo font: its text layer is garbled, and it even encodes the same nasal vowel in two different ways. So every form below was read from page renders at 200 dpi, with zooms of up to 800 dpi for diacritics. The text layer only helped us find examples. The draft restarts its example numbers inside the chapter, so each number is given with its section. New index entries are NX-236–260. This follows review 5, and H-numbers continue from there. Also searched, with nothing new found: the draft’s content-question chapter (ch. 17). Its text layer has doo only in a prose answer passage, so it has no content question with =sh across doo. Willie ch. 4 has only one negative example, (15).
2026-11-13 audit pass 7. Grant’s instruction (2026-11-13): VPISH is not a word-order account; it frames the verb’s arguments so that syntax and semantics line up. In ordinary tensed clauses, V→v→T and subject raising Spec,vP→Spec,TP (on to Spec,CP when needed) are the working default, not GRANT’S CONJECTURE to be tested against a no-movement rival. This review was posted before that instruction. Its derivations, labels and verdicts on verb movement are rewritten here to use the default: the former no-movement rival is kept only as labelled contrast, each derivation says what else has to move, and examples that seemed to resist the default are kept as tests of it (OPEN). No Navajo form, gloss, judgment or citation changed. See the 2026-11-13 note and the pass 7 report.
HYPOTHESISPlain-words verdict. Which clause is negated is fixed by where da is, not by where doo is. doo can move around inside its clause without changing the meaning. The verb-final -í- is a real argument: add one noun too many and the sentence is out, which is what theta roles alone predict. Navajo uses a separate word, hanii, to negate a single phrase, which bears out Grant’s split between sentential and constituent negation. One starred example, where hanii cannot follow “verb + object”, is what the V→v→T working default predicts: the verb has left the VP, so “verb + object” is not one phrase. An analysis with no verb movement (labelled contrast) would need an extra rule. It is one example, it depends on an added assumption (H20), and its two follow-up tests are OPEN. Light d is not needed. Telicity is still OPEN. No negative passive or negative inverse question turned up.
Starting point
Assumed, by instruction: VPISH (the subject argument first merges in Spec,vP) and theta roles; since audit pass 7 (2026-11-13), also the V→v→T working default stated below. HYPOTHESIS Everything else is tested: N1/N2/N2′ and H1–H19 from reviews 1–5, light d, the NegP head and specifier assignments, Agree. Since audit pass 7 (2026-11-13), V→v→T and subject raising Spec,vP→Spec,TP (on to Spec,CP when needed) are the working default in tensed clauses, assumed rather than tested: VPISH frames where the arguments start and get their roles; it is not a word-order account. An example that seems to resist the default is kept and flagged OPEN. Adverbs are pied-piped with VP, and only sentential negation counts in verb-movement tests. Every derivation ends in a logical form (LF), and a rival fails if its LF contradicts a reading the source reports. The draft’s own analyses (the “negative frame around the verb phrase”, incorporation of -í- following Hale & Platero 2000) appear only as labeled contrast.
The sources’ claims, as theirs
SOURCEDWillie, the impersonal fourth person. Impersonal sentences “instruct or advise the hearer about acceptable behavior”, and they “often contain a value judgment or a negative” (pp. 118–119).
SOURCEDWhich clause is negated. “The part of a complex sentence that is negated by doo…da is the clause in which these particles appear.” The pairs “differ syntactically on the placement of da”, and “the placement of da is crucial in determining meaning” (pp. 88–89). doo may instead come “immediately before the main verb”, and this placement “tends to be dispreferred” (p. 89).
SOURCEDConstituent negation. “To do this, the particle hanii is placed immediately after the constituent to be negated” (p. 92). Any noun phrase to its left can be its focus, and there is a tendency for it to follow the first constituent. It “cannot appear after the final verb in a simple clause” (p. 95). Answers like (4)–(6) have a verb as its focus (p. 93). The frame doo hanii … da is a separate expression; in a question it gives “a ‘why not’ question” (p. 97).
SOURCEDIndefinites. The h-words form indefinites with -shįį, -da or -dashįį. The h…da forms are negative polarity items and are “unacceptable in simple affirmative sentences” (pp. 98, 101). The shįį forms are “not true positive polarity items”, since they occur inside doo…da (pp. 98, 102).
SOURCEDThe incorporated -í-. It “appears immediately after the verb stem” in place of a free-standing indefinite; a postposition comes along with it. When doo is clause-initial, -í- is read as the subject; after a noun phrase, as the object (pp. 99–100, attributed to Hale & Platero 2000). “Speakers judge examples like the following to have too many arguments in them” (p. 100).
The checked data
SOURCED Exact forms, with example numbers and printed pages. Source glosses, printing slips and crops are at the linked index entries.
Negative impersonal (Willie):NX-236 (15), p. 119: Doo hatis njigháa da, translated “One does not step over people.” · NX-237 (16), p. 119: ʼabíní=go hojitaał=go yáʼátʼééh, translated “It is good for one to sing in the morning.”
Where doo sits in a complex sentence:NX-238 (40b), p. 89: Jáan doo mósí diyeeshxééł nízin da. · NX-239 (42), p. 89: Jáan mósí diyeeshxééł doo nízin da.. Both are translated “John doesn’t want to kill the cat”, and the second is “dispreferred”. · NX-240 (41a), p. 89: Nátʼoh doo neididoołnihígíí yínálniih., translated “He remembers not to buy cigarettes.” · NX-241 (41b), p. 89: Nátʼoh neididoołnihígíí doo yínálniih da., translated “He doesn’t remember to buy cigarettes.” Compare the earlier-indexed pair NX-180/NX-181 (38a/b, p. 88): only da moves, and the meaning changes from “John expects Mary not to buy a car” to “John doesn’t expect Mary to buy a car”.
Constituent negation:NX-243 (13), p. 95: ʼAzeeʼíłʼíní shicheii ʼáshįįh łikan yitsʼéidiníʼą́. (affirmative baseline) · NX-244 (14): ʼAzeeʼíłʼíní hanii shicheii ʼáshįįh łikan yitsʼéidiníʼą́, hataałii gaʼ., translated “It is not the doctor who prohibited my grandfather (maternal) from using sugar, it was the medicine man.” · NX-245 (16): ʼAzeeʼíłʼíní shicheii ʼáshįįh łikan hanii yitsʼéidiníʼą́, ʼáshįįh gaʼ., translated “It is not sugar that the doctor prohibited …, it was salt.” · *NX-246 (17), pp. 95–96: *Hérii dibé nayiisniiʼ hanii., starred, with the intended reading “It is not buying sheep that Harry did.” · NX-242 (4 A), p. 93: ʼAshxosh hanii; shinááʼ yeeʼ hanályįįh., translated “I am not sleeping; I was just resting my eyes.” (an answer to “What are you doing, are you sleeping?”).
A negative question with no wh-word:NX-247 (28), p. 97: Doo hanii kótʼéego ʼánílééh da?, translated “Why don’t you make it like this?”
Indefinites:NX-248 (30), p. 98: Doo háiida atʼééd yiztsʼǫs da., translated “No one kissed the girl.” · NX-249 (31): Háíshįį atʼééd yiztsʼǫs., translated “Someone kissed the girl.” · NX-259 §5.1 (2b), p. 101: Atʼééd háíshįį yiztsʼǫs., translated “The girl kissed someone.” · NX-250 fn. 8 (i), p. 98 (= §5.1 (4), p. 102): Atʼééd doo háíshįį yiztsʼǫsda., translated “The girl didn’t kiss anyone.” · *NX-260 §5.1 (3), p. 101: *Háíida atʼééd yiztsʼǫs., starred.
Incorporated -í-:NX-251 (38b), p. 99: Doo atʼééd yiztsʼǫsída., translated “No one kissed the girl.” · NX-252 (39b): Atʼééd doo yiztsʼǫsída., translated “The girl didn’t kiss anyone.” · NX-255 (44a), p. 100: Łééchąąʼí doo yishxashída., translated “The dog didn’t bite anything.” · NX-256 (44b): Doo łééchąąʼí yishxashída., translated “Nothing bit the dog.” · NX-253 (40b), p. 100: Hastiin doo oołbąsígóóda., translated “The man is not driving to anywhere.” · NX-254 (42), p. 100: Kingóó doo oołbąsída., translated “No one is driving to town.” · *NX-257 (45), p. 101: *Ashkii atʼééd doo yiztsʼǫsída. · *NX-258 (46): *Atʼééd doo háíida yiztsʼǫsída., both starred as having “too many arguments”.
OPENNot indexed. Some examples on these pages print with substitute glyphs even in the render, for example (35)–(36) on p. 88, so we left them out rather than restore letters by guesswork. Other undamaged examples, such as (27), (32a–c), (39a), (40a), (41a–b) of §4.2, (43a–b) of §4.2 (already NX-182/183) and the hanii examples (18)–(26), are also left out because nothing below depends on them.
Hypotheses tested, revised or added
N2 versus N2′ (reviews 4–5), tested on clause scope. N2: ¬ scopes over exactly the material between doo and da. N2′: ¬ sits at the final da and covers the clause da closes. doo is free inside that clause but must precede any NPI.
H17 (review 5), NPI concord. Tested for the first time against a starred string with no negation.
H18a/H19-lin (review 5), and a new rival H18c.H18c:-í- is not an argument. It is a polarity affix on the verb, and the missing argument is just an unpronounced pronoun.
H20 · hanii takes a phrase (new).hanii merges right after a maximal projection (XP), and that XP is its focus. ¬ applies to the focus; the rest of the clause is presupposed.
Working default (V→v→T, subject Spec,vP→Spec,TP, on to Spec,CP when needed): assumed in every tensed clause, not tested against a no-movement rival (audit pass 7, 2026-11-13). Each derivation says what else has to move.
Derivations, step by step
HYPOTHESIS Each derivation gives Merge with theta assignment, Agree (or “none”), Move, which copy is pronounced, what is inside the verb word versus outside it, and finally the LF. Extra assumptions are marked. Brackets put specifiers on the left. Heads are final except where a derivation says otherwise. NegP is placed as in review 5: doo in Spec,NegP, final Neg⁰ da, NegP above TP, as the working default requires (head assignment still HYPOTHESIS). The inflected verb is not segmented beyond the source’s hyphens.
NX-236 (negative impersonal). Merge P “over” with its fourth-person object, which the source glosses as the ha- of hatis and translates “people”. The PP is the path argument of V “walk/step” (theta: Path, an assumption about this verb). Merge v. The fourth-person subject merges in Spec,vP (Agent). It has no overt nominal, and its person shows only in the verb word. Merge T, then NegP. Agree: the verb’s subject marking matches the fourth-person subject (HYPOTHESIS). Working default: V→v→T, and the silent subject goes to Spec,TP. doo is still first, so the order is doo + PP + verb word + da. Nothing else has to move: with a final T⁰ the verb word comes right before da (HYPOTHESIS). LF: GEN x[human(x)] ¬∃e[walk(e) ∧ Agent(e,x) ∧ Path(e, over(people))]. Placing the generic operator above ¬ is our reading of the translation and of Willie’s “instruction” use (HYPOTHESIS). It is silent, so it cannot test N2. Result: the impersonal is an active clause with an Agent, so it is not a passive and fills no passive cell. The negation is sentential (final da), so the example could be used for verb-movement tests, and it fits the default with nothing else moving.
NX-238 / NX-239 versus NX-180 / NX-181 (which clause ¬ applies to). Merge V “kill” with mósí (Theme) and a first-person Agent in Spec,vP. The source glosses the verb 3-1, so the complement seems to be the wanter’s own words, “I will kill the cat” (HYPOTHESIS, as for NX-231). That clause is the Theme of matrix V “want”. Jáan merges in matrix Spec,vP (Experiencer). Merge matrix T and NegP. LF for both: ¬∃s[want(s) ∧ Exp(s,john) ∧ Theme(s, ∃e[kill(e) ∧ Agent(e,john) ∧ Theme(e,the_cat)])]. NX-238 pronounces doo before the complement. NX-239 pronounces it after the complement, right before nízin. What the four sentences show: moving da (NX-180 → NX-181) changes which verb is negated. Moving doo (NX-238 → NX-239) changes nothing in the printed meaning. N2′ derives this directly. N2 handles NX-180, NX-181 and NX-238, but not NX-239: there, the complement is pronounced outside the frame and yet is read inside ¬. N2 survives only if the complement moves out of the frame and the LF reads its lower copy (reconstruction, HYPOTHESIS). Since a clause carries no quantifier of its own, this is a cost for N2, not a contradiction. So N2′ is favored. The quantifier test (an indefinite left of doo with a printed scope judgment) is still OPEN.HYPOTHESISWhat else moves under the working default: NegP is above TP, so both Jáan and the complement in NX-239 must move above NegP. The complement’s move is a scrambling step (HYPOTHESIS), with the lower copy read at LF. (Labelled contrast: an analysis with no movement would instead need doo to merge between the complement and V, a third merge site besides the clause-initial one and the one after the subject in review 5.) The “dispreferred” judgment might reflect the cost of the extra Move, but that link is untested.
NX-240 / NX-241 (embedded versus matrix negation). NX-240: doo is inside the -ígíí complement, and da is absent before -ígíí (source, fn. 6). LF: remember(he, ¬∃e[buy(e) ∧ Agent(e,he) ∧ Theme(e,cigarettes)]). NX-241: doo … da is in the matrix clause, and the whole complement, including its object nátʼoh, is pronounced left of doo. LF: ¬remember(he, ∃e[buy(e) ∧ Agent(e,he) ∧ Theme(e,cigarettes)]). The pattern is the same as NX-239: what doo precedes does not decide what is negated. The bare noun nátʼoh is read under ¬ here, but it may be a kind term, which has no scope. A reading like “there are cigarettes he doesn’t remember to buy” is neither printed nor excluded. So this is not the decisive indefinite test.
NX-251 / NX-252 and NX-255 / NX-256 (incorporated -í-, H19-lin out of sample). NX-256: merge łééchąąʼí as the complement of V “bite” (Theme), PRN in Spec,vP (Agent), then T and NegP. Agree (H17): Neg⁰ with PRN. PRN moves to a position under doo (H18a) and is written on the verb word inside da. LF: ¬∃x∃e[bite(e) ∧ Agent(e,x) ∧ Theme(e,the_dog)]. NX-255: łééchąąʼí merges in Spec,vP (Agent) and is pronounced left of doo; PRN is the Theme. LF: ¬∃y∃e[bite(e) ∧ Agent(e,the_dog) ∧ Theme(e,y)]. The kiss pair NX-251/252 is the same. The reverse readings, again, are not excluded by VPISH, theta roles or H17. H19-lin (the gap is the first thing after doo) excludes them, so it now holds for three verbs (rope, kiss, bite) and for “see” (NX-182/183). It also holds when the hidden agent is non-human (“nothing bit the dog”). H4′: the lone noun is the agent in NX-252 and NX-255, as in review 5’s NX-229, so the one-noun default again needs restricting to clauses without a displaced PRN.
NX-254 (a goal phrase in front of doo). Merge kingóó “town-to” in VP with V “drive”, and PRN in Spec,vP. PRN is read as the subject, so its gap must be the first thing after doo (H19-lin). Base order has Spec,vP before VP, so doo must come before Spec,vP. But kingóó is pronounced before doo. If kingóó is an argument of V (a Goal), it has moved out of VP across the subject position, on every analysis, with or without the working default. If it is an adjunct, it can be merged above NegP with no movement. Theta roles alone do not tell us which, so this is OPEN. LF: ¬∃x∃e[drive(e) ∧ Agent(e,x) ∧ Goal(e,town)]. NX-253 shows the postposition traveling with PRN, as in review 5’s NX-230: ¬∃l∃e[drive(e) ∧ Agent(e,the_man) ∧ Goal(e,l)].
*NX-257 / *NX-258 (too many arguments: theta roles decide). “Kiss” assigns two roles, Agent and Theme (source gloss 3-3). In *NX-257 there are three argument expressions: ashkii, atʼééd and PRN. In *NX-258 there are also three: atʼééd, háíida and PRN. By the theta axiom, every argument needs a role and every role exactly one argument, so one expression is left without a role and the sentence is out. This follows from the axioms alone, provided PRN is itself an argument. So H18c is contradicted: if -í- were only a polarity affix, *NX-257 should be fine with the meaning “The boy didn’t kiss the girl (at all)”. *NX-258 adds a second point. PRN cannot double an overt h-word the way the h-word’s -da agrees with Neg (H17). So PRN is an argument, not a concord copy.
NX-248 / *NX-260 / NX-249 / NX-259 / NX-250 (indefinites). NX-248 is derived like review 5’s NX-218: ¬∃x∃e[kiss(e) ∧ Agent(e,x) ∧ Theme(e,the_girl)]. *NX-260 has the same h-word with -da but no doo … da. There is no Neg⁰ to probe it, so H17’s Agree cannot apply and the -da form is unlicensed. This is the first starred string that bears on H17, and H17 predicts it. NX-249/NX-259: the same háíshįį is the Agent when first and the Theme when after atʼééd, so the first noun is the subject (review 2’s H3/H4 pattern). LF: ∃x[kiss(x,the_girl)] and ∃y[kiss(the_girl,y)]. With review 5’s háí-lá (NX-220) and NX-248’s háiida, the clean forms now show one h-base with three companions: -lá gives a question, -da an NPI and -shįį a plain indefinite. H11 is supported further. (Review 5 had only a glyph-damaged form of the third.) NX-250: háíshįį inside doo … da is read under ¬: ¬∃y∃e[kiss(e) ∧ Agent(e,the_girl) ∧ Theme(e,y)]. That fits both N2 and N2′. Whether a wide reading (“there is someone the girl didn’t kiss”) is also available is not reported (OPEN).
NX-244 / NX-245 / NX-242 (constituent negation, H20). NX-244: build the clause of NX-243. The doctor is in Spec,vP (Agent); shicheii and ʼáshįįh łikan are internal arguments. The roles of the “take away from” verb follow the source gloss 3-away.from-3-3 (assigning those roles is our HYPOTHESIS). hanii follows the subject DP, its focus. LF: presupposed ∃x[prohibit(x, my_grandfather, sugar)]; asserted ¬[ιx.prohibit(x, my_grandfather, sugar) = the_doctor], with gaʼ supplying the correction. NX-245 is the same with the object DP as focus. NX-242: hanii follows the whole verb word ʼAshxosh, which carries a first-person subject. The focus is the whole clause, with alternatives such as “I am resting my eyes”: ¬[what I am doing = sleep]. No da appears in any hanii sentence, and the scope of ¬ is the focus alone. So Navajo marks Grant’s split between sentential and constituent negation with different words. That supports counting only doo … da in verb-movement tests. Every doo … da example in this chapter is clause-level. doo hanii … da (NX-247) is a separate fixed frame (source), not constituent negation.
*NX-246 (hanii after “verb + object”): a test of the working default. The intended focus is “buying sheep”, the verb with its object and without Hérii. HYPOTHESISWorking default, V→v→T:[TP Hérii [T′ [vP t [VP dibé t_V] t_v] [T nayiisniiʼ]]]. The string dibé nayiisniiʼ is T′, which contains T but not Spec,TP, so it is not a maximal projection. Under H20 it cannot be hanii’s focus, so the intended reading is underivable, which matches the star. In NX-242, by contrast, the verb word with its silent subject is a whole TP, so it can be a focus, which matches the acceptable answer. Labelled contrast, no verb movement: V stays in VP, and T’s and v’s marking joins it without movement. Then [VP dibé nayiisniiʼ] is a maximal projection pronounced exactly as dibé nayiisniiʼ. Under H20 it should be a possible focus, so this contrast analysis predicts the starred reading to be fine. To match the star, it needs an added rule, for example that a VP cannot be a hanii focus. Result: the star is what the working default predicts; the no-movement contrast needs an extra rule. This is the first example in the project that bears on verb movement this way. But it is one starred string, it relies on H20, and two things are untested. (a) The default predicts that the same string should allow a whole-clause focus (“it isn’t that Harry bought sheep”); the source does not say whether it does. If that reading is also out, the star has some other cause and the example no longer bears on the default (OPEN). (b) The default predicts that “buying sheep” can be focused by putting hanii after the object, on the VP left behind when V moves, the same position hanii has in NX-245. No such sentence with that reading is printed. This is our prediction, not data, and it must be checked with speakers or another source. We construct no Navajo string for it.
NX-247 (a “why not” question without a wh-word). Merge V “make”, with a second-person Agent in Spec,vP and the manner phrase kótʼéego “this way” in VP. Then T, NegP (doo … da), and a C that makes the clause a question, which we assume because the sentence is interrogative (silent; HYPOTHESIS). LF from the translation: ?r[cause(r, ¬∃e[make(e) ∧ Agent(e,you) ∧ Manner(e,this_way)])]. Not understood: nothing pronounced supplies “why”. Neither hanii nor a silent reason operator can be derived from these data, so the source of the “why” is OPEN. For comparison, NX-138 has an overt “why” with tʼáadoo … da.
Does light d earn its keep?
HYPOTHESISFair test. A light-d analysis would put a silent D⁰ on bare nouns like atʼééd, łééchąąʼí and ashkii, which are translated as definites. It would also treat the h-word base (háí, which contains review 5’s -í-) as a D. Result: the data do not need it, and they do not count against it. (a) The theta count that rules out *NX-257 and *NX-258 counts argument expressions. A silent D on each noun changes nothing in that count. (b) The definite readings come for free in every LF above, and no contrast turns on the presence of a D. (c) Willie’s impersonal has no nominals at all; its arguments are marked only inside words. (d) The overt -í- base is D-like, but it is overt, so it is not evidence for a silent d. Light d stays idle.
Telicity test
OPENPrediction tested: where -í- and its postposition are written depends on the verb’s telicity. In the indexed data the pattern is the same with verbs translated as punctual events (“kiss” NX-251/252, “bite” NX-255/256, and “see” NX-182/183) and with “be driving (to)” (NX-253/254). The source’s labels (P, Pf, I, Pg) mark mode and viewpoint, not telicity. The two glosses of the same stem, “3-drive.Pg” in (32) and “3-drive.I” in (40)–(42), show these labels are not even used consistently. No pair differs only in telicity. So we found no telicity effect, but the test is not clean: neither supported nor contradicted, OPEN. Willie’s (15)–(16) are a negative and an affirmative impersonal with different verbs, so they are not a telicity pair either.
Question × polarity × voice grid (cumulative after review 6)
OPEN (declaratives NX-216 and, impersonal, NX-236 only)
Wh, affirmative
Review 3 set; NX-212; NX-220, 221
Review 3 set
NX-213, 215
Wh, negative
NX-210, 211; NX-138; new:NX-247 (“why not” reading, no wh-word)
OPEN
OPEN
HYPOTHESISWhat this batch adds and forces:
(a) One new negative question, of a new kind. NX-247 is formally a negative clause with doo hanii … da and no wh-word or -ísh, but it is read as “why not”. We place it in the wh-negative cell by its reading only. What supplies the question force is OPEN (derivation 10).
(b) No passive or inverse question. The impersonal (NX-236) is a fourth-person active clause with an Agent (derivation 1), so it fills no passive question cell. It is a negative declarative.
(c) H13-int (review 3) is still OPEN. Neither ch. 6 nor the content-question chapter prints a content question with =sh displaced across doo.
(d) Ordering facts that are forced: the clause ¬ applies to is the one closed by da (derivations 2–3). The PRN gap must be first after doo (derivation 4). If kingóó is an argument, material before doo got there by Move (derivation 5). The order of doo relative to a complement clause is optional, with one order dispreferred (NX-238/239).
(e) HYPOTHESISWorking default, verdict for this batch: every derivation goes through with it. *NX-246 versus NX-242 (derivation 9) is what it predicts, conditional on H20 and on the two untested predictions there (OPEN). What else has to move: NegP is above TP, so whatever precedes doo has moved above NegP (NX-239, NX-241, NX-254; HYPOTHESIS). Labelled contrast: a no-movement analysis would need a third doo site instead.
Where we part from the sources’ analyses (labeled contrast)
HYPOTHESISAgrees with our derivations: the draft’s report that speakers find *NX-257/258 to have “too many arguments” is what the theta axiom predicts once PRN is an argument (derivation 6). Its rule that the reading of -í- depends on where doo is is H19-lin (derivation 4).
HYPOTHESISNot adopted: the draft says “the negative frame appears around the verb phrase” (p. 88). We do not take over a VP label for the frame. Subjects and complements are regularly to the left of doo, and under the working default the frame surrounds TP minus whatever moved above NegP. The draft’s term “incorporation” for -í- is not adopted as head movement. Review 5’s phrasal Move plus PF placement (H18a) still covers the postposition traveling along (NX-253).
HYPOTHESISWillie’s point that the impersonal avoids addressing or accusing anyone is a fact about how it is used. Nothing in our derivation depends on it.
What this changes in reviews 1–5
HYPOTHESIS N2′ is favored over N2 for which clause is negated (NX-238/239 with NX-180/181). N2 needs reconstruction for NX-239 and NX-241. The quantifier version of the test is still OPEN.
HYPOTHESIS H17 now predicts a starred string (*NX-260), and H11 has its third clean form (NX-249).
HYPOTHESIS H19-lin holds for three more verbs (NX-251/252, 255/256). H4′ again needs restricting (NX-252, NX-255).
HYPOTHESIS PRN -í- is an argument, by the theta axiom (*NX-257/258). Rival H18c is contradicted.
GRANT’S CONJECTURE Navajo’s split between sentential and constituent negation is lexical (doo … da versus hanii). *NX-246 is the first conditional test of the V→v→T working default, and it fits (OPEN until tests (a) and (b) are run).
HYPOTHESIS Light d is still idle, and telicity is still OPEN.
OPEN predictions
OPENThe hanii test of the working default: (a) Can the string of NX-246 have a whole-clause focus? (b) Can “buying sheep” be focused with hanii after the object? That form is our prediction, not a printed example. Construct nothing for the site, and ask speakers or find a source.
OPENN2 versus N2′ with a quantifier: we still need a printed scope judgment for an indefinite or quantified nominal left of doo.
OPENIs kingóó an argument? If it is a Goal argument, NX-254 forces Move across the subject position on every analysis.
OPENWhere the “why” of doo hanii … da questions comes from (NX-247; Young & Morgan 1987:350 is the draft’s source).
OPENStill missing from the grid: a polar passive question, a negative inverse question and a negative passive question. H13-int (content =sh across doo) is also still untested.
2026-11-12 · First-principles review 7: Willie on quantifier scope and the distributive da- — one string with two scope readings, and which argument da- counts
2026-11-14: the “On the spine” summary at the top of this page supersedes this section’s framing. Its data, judgments and citations are unchanged.
Restated on the unified structure (2026-11-15). This section’s results now sit on the one structure for Navajo: unified structure, distributive and scope rows. Its restatement as derivations is in Results on the spine. Data, judgments and citations below are unchanged.
Source: MaryAnn Willie, Navajo Pronouns and Obviation (University of Arizona dissertation, 1991), ch. 3, the pages just before §3.3.2 and §3.3.2 “The distributive and Inverse voice” itself, printed pp. 78–85 (PDF pp. 84–91; PDF page = printed page + 6), examples (54)–(69). Every form below was read from 200-dpi page renders and 400-dpi crops, with diacritics checked on enlarged crops. The PDF text layer helped only to find examples. New index entries are NX-261–283. This follows review 6, and H-numbers continue from there. (54)–(57) on pp. 78–79 are indexed with (58)–(69) because the scope examples are built on them. Also checked, with nothing new found: these pages have no doo … da, no hanii, no question and no adverb. A text-layer search of the ch. 8 remainder (printed pp. 210–216, examples (18)–(32)) found doo only inside the verb yidooloh, so the ch. 8 remainder has no negation either.
2026-11-13 audit pass 7. Grant’s instruction (2026-11-13): VPISH is not a word-order account; it frames the verb’s arguments so that syntax and semantics line up. In ordinary tensed clauses, V→v→T and subject raising Spec,vP→Spec,TP (on to Spec,CP when needed) are the working default, not GRANT’S CONJECTURE to be tested against a no-movement rival. This review was posted before that instruction. Its derivations, labels and verdicts on verb movement are rewritten here to use the default: the former no-movement rival is kept only as labelled contrast, each derivation says what else has to move, and examples that seemed to resist the default are kept as tests of it (OPEN). No Navajo form, gloss, judgment or citation changed. See the 2026-11-13 note and the pass 7 report.
HYPOTHESISPlain-words verdict. One printed sentence, “every girl kissed a boy” with da- (NX-267), means either “one boy” or “many boys”. With the subject starting above the object, the “one boy” reading can only come from an extra step: the indefinite moves up unseen, or a scope operator sits at the top of the clause. Pronounced order alone does not fix scope. The plural prefix da- needs a plural agent wherever the agent is pronounced, which follows from VPISH if da- looks at the nearest argument. In the inverse it also needs a plural or pointed-out patient, and we do not yet understand that part. Light d is not needed. Telicity is still OPEN. No question or negation cell is filled. These pages have no negation, so the N2/N2′ quantifier test and the hanii test of the working default are still OPEN.
Starting point
Assumed, by instruction: VPISH (the subject argument first merges in Spec,vP) and theta roles; since audit pass 7 (2026-11-13), also the V→v→T working default stated below. HYPOTHESIS Everything else is tested, including H1–H20 from reviews 1–6, light d, Agree, the label and site of any head for da-, and covert movement. Since audit pass 7 (2026-11-13), V→v→T and subject raising Spec,vP→Spec,TP (on to Spec,CP when needed) are the working default in tensed clauses, assumed rather than tested: VPISH frames where the arguments start and get their roles; it is not a word-order account. An example that seems to resist the default is kept and flagged OPEN. Adverbs are pied-piped with VP, and only sentential negation counts in verb-movement tests. Every derivation ends in a logical form (LF) whose scope order matches c-command. A rival fails if its LF contradicts a reading the source reports. Willie’s own analysis (nouns as adjuncts to pronouns in the verb, number as a property of the pronouns) appears only as labeled contrast.
The source’s claims, as hers
SOURCEDPaucal and definiteness. “Note that the external patient argument is interpreted as definite. The paucal plural (less than three) can appear with either argument in either the Direct or Inverse” (p. 79). “(With more than two, the distributive must be used.)” (p. 78).
SOURCEDQuantifier. “When the universal quantifier has scope over the external argument, the voice alternation provides different interpretations of the quantifier” (p. 79). “With greater plurals, the distributive da- must be used” (p. 80).
SOURCEDDistributive, direct. Distributive marking “is acceptable when the noun modifying the agent pronoun is plural, but excluded when it is not, as in (63d). It appears that the agent falls under the scope of the distributive marker”, and the patient’s number does not seem to matter (p. 82).
SOURCEDDistributive, inverse. “With common nouns, da- plus bi- requires that both internal and external arguments be plural” (p. 83). “However, if the external argument is made maximally referential with demonstratives, then the sentence becomes acceptable” (p. 83). Then, on p. 84: “bi- requires the external argument to be definite, but when the Distributive da- is present, it can license a plural external argument, which can escape the requirement on definiteness. With dabi-, the internal argument must be plural, and the external argument may optionally be interpreted as plural.” For nouns with no plural form, “In (67a), the agent must be interpreted as plural. In (67b), both must be interpreted as plural” (p. 85). The p. 83 wording (“both … be plural”) and the p. 84 wording (“optionally”) are not the same; we record both.
SOURCEDHer conclusion. Number “would not be possible” to associate with different arguments by voice “if number were a property of the nominals, rather than the pronouns”. In the inverse, the patient “must be either definite or plural if the sentence is to be acceptable” (p. 85).
The checked data
SOURCED Exact forms, with example numbers and printed pages. Willie prints no gloss line for (54)–(60), (63d), (65b–c) or (66). Source glosses, where printed, and the crops are at the linked index entries. In Willie’s terms, “Direct” is the yi- form and “Inverse” the bi- form. Her English passives translate the bi- form, so they do not make these sentences passives.
Paucal plural, direct versus inverse:NX-261 (54), p. 78: John ʼatʼééké yiztsʼǫs, “John kissed the girls.” · NX-262 (55): John ʼatʼééké biztsʼǫs, “John was kissed by two girls.” · NX-263 (56), p. 79: ʼatʼééké John yiztsʼǫs, “Two girls kissed John.” · NX-264 (57): ʼatʼééké John biztsʼǫs, “Those two girls were kissed by John.”
Universal plus indefinite:NX-265 (58), p. 79: tʼáá ʼałtso ʼatʼééké ʼashkii łaʼ biztsʼǫs, “All the girls were kissed by a boy.” · NX-266 (59): tʼáá ʼałtso ʼatʼééké ʼashkii łaʼ yiztsʼǫs, “Each of the girls kissed a boy.” · NX-267 (60), p. 80: tʼáá ʼałtso ʼatʼééké ʼashkii łaʼ deiztsʼǫs, printed with two readings, (a) “Every girl kissed a boy. (one boy)” and (b) “Every girl kissed a boy. (many boys)”.
Singular baseline:NX-268 (61), p. 81: ʼashkii ʼatʼééd yooʼį́, “The boy sees the girl.” · NX-269 (62): ʼatʼééd ʼashkii booʼį́, “The girl is seen by the boy.”
da- in the direct:NX-270 (63a), p. 81: dayiiłtsą́, “They saw them/it/him/her.” · NX-271 (63b), p. 82: ʼashiiké ʼatʼééké dayiiłtsą́, “The boys saw the girls.” · NX-272 (63c): ʼashiiké ʼatʼééd dayiiłtsą́, “The boys saw the girl.” · *NX-273 (63d): ʼashkii ʼatʼééké dayiiłtsą́, starred, intended “(The boy saw the girls.)”
da- in the inverse:NX-274 (64a), p. 83: dabiiłtsą́, “They/he (was)were seen by them/*him.” · NX-275 (64b): shizhéʼé ʼashiiké dabiiłtsą́, “My father was seen by the boys.” · NX-276 (65a): ʼashiiké ʼatʼééké dabiiłtsą́, “The boys were seen by the girls.” · *NX-277 (65b): ʼashiiké ʼatʼééd dabiiłtsą́, intended “(The boys were seen by the girl.)” · *NX-278 (65c): ʼatʼééd ʼashiiké dabiiłtsą́, intended “(The girl was seen by the boys.)” · NX-279 (66), p. 84: ʼéidí ʼatʼééd ʼashiiké dabiiłtsą́, “That girl (over there) was seen by the boys.”
A noun with no plural form:NX-280 (67a), p. 84: mąʼii tłʼízí deishxash, “The coyotes bit the goat(s).” · NX-281 (67b): tłʼízí mąʼii dabishxash, “The goats were bitten by the coyotes.” · NX-282 (68), p. 85: mąʼii tłʼízí yitah daʼashxash, “The coyote (went among and) bit the goats.” · NX-283 (69): ʼéidí tłʼízí mąʼii dabishxash, “That goat there was bitten by the coyotes.”
OPENNot indexed: (52)–(53), p. 78 (“when”-clauses with direct and inverse, already outside this question) and the prose statement (70), p. 86, which is not a Navajo example.
Hypotheses tested, revised or added
H3 and H6′ (reviews 1–3). H3: the inverse is O internally merged above S, with the upper copy pronounced. H6′: the O moved by H3 must be definite or presuppositional. Tested here on a numeral plural and a universal.
H6-scope (review 2’s OPEN prediction). A quantified inverse O takes wide scope if it has moved and does not reconstruct. Review 2 asked for (58)–(60) to be checked against the page images for this.
SR · scope rigidity (rival, stated so it can be tested). A quantifier’s scope is read only from a position where one of its copies sits. There is no unseen movement and no special mechanism for indefinites.
H22a / H22b · wide scope for a łaʼ indefinite (new).H22a: the indefinite moves to a position above the clause’s other quantifier, and only its lower copy is pronounced (covert Move). H22b:łaʼ introduces a choice-function variable that is bound by an existential operator merged at the top of the clause. The noun itself does not move.
H21a · da- counts the nearest argument (new).da- is a head (call it Dist⁰; the label is a placeholder) merged directly above vP. It Agrees with the closest argument it c-commands, which must be plural. With VPISH, the closest argument in an unmoved clause is the agent in Spec,vP.
H21b · the inverse patient is also counted (new, partial). In the inverse, H3’s moved O passes through a left (outer) specifier of vP. That position is also inside the reach of Dist⁰, so the moved O is counted too, unless it is a demonstrative, possessed or pronominal phrase. That exception is a stipulation. We do not know why it holds.
Working default (V→v→T, subject Spec,vP→Spec,TP, on to Spec,CP when needed): assumed in every tensed clause, not tested against a no-movement rival (audit pass 7, 2026-11-13). Each derivation says what else has to move.
Derivations, step by step
HYPOTHESIS Each derivation gives Merge with theta assignment, Agree (or “none”), Move, which copy is pronounced, and finally the LF. Brackets put every specifier on the left. Heads are final, as in reviews 5–6. Extra assumptions are marked. The verb word is not segmented beyond what the source glosses. Position classes are not used, so the order of prefixes inside the verb word is not taken as evidence for any head order.
NX-263 / NX-264 (the same nouns, yi- versus bi-).NX-263: Merge V “kiss” with John (Theme). Merge v. ʼatʼééké merges in Spec,vP (Agent). Merge T. Agree: none needed for anything below. Under the working default, V→v→T and the agent moves to Spec,TP: [TP ʼatʼééké [T′ [vP ʼatʼééké [v′ [VP John V] v]] yiztsʼǫs]], with the lower copies silent. LF, from the translation “Two girls kissed John”: ∃X[girls(X) ∧ |X|=2 ∧ ∃e[kiss(e) ∧ Agent(e,X) ∧ Theme(e,john)]]. NX-264: the roles are reversed. ʼatʼééké is the Theme in VP and John the Agent in Spec,vP. H3 moves the Theme to a left specifier above the agent, and the upper copy is pronounced; the verb is spelled out with bi-. LF, from “Those two girls were kissed by John”: ∃e[kiss(e) ∧ Agent(e,john) ∧ Theme(e, ιX[girls(X) ∧ |X|=2])]. Result: the same noun, in the same first position, is indefinite as the direct agent and definite as the inverse patient. This is H6/H6′ passing out of sample on a numeral plural. The linear rival H6-lin (“the first noun of an inverse clause is definite”) also fits this pair. Review 2’s *NX-049 still separates the two.
NX-266 (universal agent, direct). Merge V “kiss” with ʼashkii łaʼ (Theme). Merge v, and tʼáá ʼałtso ʼatʼééké in Spec,vP (Agent). Merge T. Under the working default, V→v→T and the agent goes to Spec,TP. The universal c-commands the indefinite in every copy. LF, from “Each of the girls kissed a boy”: ∀x[girl(x) → ∃y[boy(y) ∧ ∃e[kiss(e) ∧ Agent(e,x) ∧ Theme(e,y)]]]. The scope order matches c-command. Whether a “one boy” reading is also possible is not printed for this string.OPENNot understood: the source says greater plurals need da-, yet (59) has “all the girls” without da- and is printed unstarred. Perhaps the universal licenses the form without da-, or the group is small here. Nothing printed decides it.
NX-267 (universal agent, direct, with da-): two LFs for one string. Merge V “kiss” with ʼashkii łaʼ (Theme), then v, then tʼáá ʼałtso ʼatʼééké in Spec,vP (Agent). Merge Dist⁰ (da-) above vP. Agree (H21a): Dist⁰ finds the closest argument, the agent, which is plural, so the derivation converges. Merge T. Under the working default, V→v→Dist→T, and the agent moves to Spec,TP: [TP tʼáá ʼałtso ʼatʼééké [T′ [DistP [vP tʼáá ʼałtso ʼatʼééké [v′ [VP ʼashkii łaʼ V] v]] Dist] deiztsʼǫs]] (lower copies silent). Reading (b), “many boys”:∀x[girl(x) → ∃y[boy(y) ∧ ∃e[kiss(e) ∧ Agent(e,x) ∧ Theme(e,y)]]]. This matches surface c-command. Reading (a), “one boy”:∃y[boy(y) ∧ ∀x[girl(x) → ∃e[kiss(e) ∧ Agent(e,x) ∧ Theme(e,y)]]]. Here the existential must c-command the universal. By VPISH and the theta roles, the agent is merged above the theme, so there is no copy of the universal below the indefinite. Reconstruction cannot give reading (a). Result: SR is contradicted by NX-267 (a). One extra step is needed. Under H22a, ʼashkii łaʼ moves to a position above the universal in Spec,TP, and its lower copy is the one pronounced. Under H22b, an existential operator over a choice function is merged at the top of the clause: ∃f[CH(f) ∧ ∀x[girl(x) → ∃e[kiss(e) ∧ Agent(e,x) ∧ Theme(e, f(boy))]]]. Both keep scope equal to c-command, by the moved phrase under H22a and by the operator under H22b. Telling them apart needs an indefinite inside an island with a reported reading. None is printed here, so H22a vs H22b is OPEN. Also shown:da- does not fix the indefinite’s scope, since both readings are printed with it. With a final T⁰, nothing else has to move for the surface order; the extra step is for scope only.
NX-265 (universal patient, inverse) and H6-scope. Merge V “kiss” with tʼáá ʼałtso ʼatʼééké (Theme), then v, then ʼashkii łaʼ in Spec,vP (Agent). H3 moves the Theme to a left specifier above the agent; the upper copy is pronounced and the verb is spelled out with bi-. Under the working default, V→T and the agent goes to Spec,TP, so the moved Theme must land above TP (the label of that landing site is OPEN). H6′ is met, because a universal over “the girls” presupposes its set. The two LFs that the structure allows are: reading the upper copy of the Theme, ∀x[girl(x) → ∃y[boy(y) ∧ ∃e[kiss(e) ∧ Agent(e,y) ∧ Theme(e,x)]]]; and reading its lower copy (reconstruction), ∃y[boy(y) ∧ ∀x[girl(x) → ∃e[kiss(e) ∧ Agent(e,y) ∧ Theme(e,x)]]]. The source prints one translation, “All the girls were kissed by a boy”, and no reading label. The English sentence is ambiguous, and we do not treat that ambiguity as a Navajo judgment. Willie says the two voices give “different interpretations of the quantifier” (“all” in (58), “each” in (59)), but she does not say which scope readings each has. OPENH6-scope is not confirmed and not contradicted. What would decide it: if (58) had only the “one boy” reading, H3 would need obligatory reconstruction, and Willie’s role-remapping claim (labeled contrast in review 2: the patient merged high, with no lower copy) would need the agent to move covertly. If (58) had the “each girl” reading, both accounts derive it from the surface. So even a printed reading would favor H3 only weakly.
NX-270–273 (da- in the direct).NX-271/272: Merge V “see” with ʼatʼééké or ʼatʼééd (Theme), then v, then ʼashiiké in Spec,vP (Agent). Merge Dist⁰. Agree (H21a): the closest argument is the plural agent, so the derivation converges. The theme is not checked, so its number is free (63b, 63c). Merge T; under the working default V→v→Dist→T and the agent goes to Spec,TP. LF for NX-272: DIST(the_boys)(λx.∃e[see(e) ∧ Agent(e,x) ∧ Theme(e,the_girl)]). Whether da- really distributes here or only marks a plural agent is not shown by the translations (OPEN). *NX-273: the same steps with ʼashkii (singular) as the agent. Agree finds a singular, so the derivation crashes. The star is derived.NX-270: no nominals. The translation gives a plural agent and leaves the theme’s number open, exactly as H21a predicts. NX-280 (“the coyotes bit the goat(s)”) is the same pattern with a noun that has no plural form: the agent must be read plural, and the theme need not be.
NX-274–279, NX-281, NX-283 (da- in the inverse): the agent part. Merge V “see” with the patient (Theme), v, and the agent in Spec,vP. Merge Dist⁰, which Agrees with the agent (on H21b it also Agrees with the patient; step 7). The patient ends up pronounced above the agent (H3). The agent must be plural: *NX-277 (singular ʼatʼééd as agent) is starred, and NX-274 rules out “*him”. H21a holds with the agent in second position. It holds in first position in the direct (*NX-273) as well, so the restriction follows the agent role, not the agent’s position in the string. That is what VPISH predicts if Dist⁰ looks at the nearest argument.
The same sentences: the patient part, and where H21a fails. *NX-278 has a plural agent and is still starred, because its patient is a bare singular. H21a does not predict this: in the direct, a bare singular theme is fine with da- (NX-272), and in the inverse without da-, the same bare singular patient is fine (NX-269, “The girl is seen by the boy”). Compare the rules: “the first noun must be plural” derives *NX-273 and *NX-278 but wrongly allows *NX-277. “The agent must be plural” (H21a) derives *NX-273 and *NX-277 but wrongly allows *NX-278. So neither a linear rule nor the agent rule alone is enough.H21b adds: H3 moves the patient through a left specifier of vP before Dist⁰ is merged, so Dist⁰ sees two arguments at the vP edge and both must be plural. This needs one Agree relation with two goals (multiple Agree, an added HYPOTHESIS). Plain closest-goal Agree would check only the patient, which is now closer, and would leave *NX-277’s singular agent unchecked. [TP … [DistP [vP patient [vP agent [v′ [VP patient V] v]]] Dist] …]. This derives *NX-278 and NX-276, and also NX-281, where the bare noun tłʼízí “must be interpreted as plural” in the inverse but not in the direct (NX-280). What H21b does not explain: why a demonstrative (NX-279, NX-283), possessed (NX-275) or silent pronoun patient (NX-274, “he”) escapes. A silent pronoun agent does not escape (NX-274, “*him”), so the exception cannot simply be “referential phrases are never counted”. We list the exception as a stipulation and mark it OPEN. LF for NX-279: ∃e[see(e) ∧ Agent(e,the_boys) ∧ Theme(e, that_girl)], with DIST over the agent as in step 5. Bearing on review 1’s “Move versus realization” question: the inverse patient behaves differently from the direct theme (*NX-278 versus NX-272). If it sits in a different position (H3 with H21b), that difference follows. If the inverse were only a different spelling of the same structure, the rule would have to mention the inverse by name. That is mild support for a moved patient, not proof.
NX-282 (a singular agent with a plural-marked verb). Merge P yitah “among (them)” with tłʼízí. The PP joins the verb phrase of “bite”. mąʼii merges in Spec,vP (Agent). The verb word begins da- and is glossed “PL.bit”, yet the agent is translated singular, “the coyote”. H21a is contradicted here unless da- in this sentence is merged lower, inside the verb phrase, where the closest argument is the plural object of yitah. Nothing printed tells us where it is merged, or what role the PP bears (argument or adjunct). OPEN LF from the translation: ∃e[bite(e) ∧ Agent(e,the_coyote) ∧ among(e,the_goats)], with the “went among” part of the translation not represented (not understood).
Does light d earn its keep?
HYPOTHESISFair test. A light-d analysis would put a silent D⁰ on bare nouns such as ʼatʼééd and tłʼízí, and might let that d carry number or referentiality. Result: the data do not need it, and they do not count against it. (a) Every contrast here involves an overt word: the universal tʼáá ʼałtso, the indefinite łaʼ, the demonstrative ʼéidí, the possessor in shizhéʼé. (b) The hard contrast, NX-269 versus *NX-278, has the same bare noun ʼatʼééd in both. A silent d on that noun would be the same in both, so it cannot explain why one is starred. (c) H22b’s choice function attaches to the overt łaʼ, not to a silent d. Light d stays idle.
Telicity test
OPENPrediction tested: whether da- and the scope readings depend on the verb’s telicity. The same da- pattern appears with “kiss” (NX-267), “see” (NX-270–279) and “bite” (NX-280–283). The present “sees” (yooʼį́, NX-268) versus past “saw” (dayiiłtsą́) is a difference of stem and viewpoint, which is not a telicity test. No pair differs only in telicity: neither supported nor contradicted, OPEN.
Question × polarity × voice grid (cumulative after review 7)
SOURCED Every example in this batch is an affirmative declarative, direct or inverse. No cell changes. The bi- sentences with English passive translations go in no passive cell.
OPEN (declaratives NX-216 and, impersonal, NX-236 only)
Wh, affirmative
Review 3 set; NX-212; NX-220, 221
Review 3 set
NX-213, 215
Wh, negative
NX-210, 211; NX-138; NX-247 (“why not” reading, no wh-word)
OPEN
OPEN
HYPOTHESISWhat this batch adds and forces:
(a) Forced: one string (NX-267) has two LFs with opposite quantifier order. If syntax is to match the logic, the two readings need two different structures: one with an extra step (H22a Move, or the H22b operator) and one without. Scope cannot be read off pronounced order.
(b) Forced: the plural requirement of da- follows the agent role in both voices (*NX-273, *NX-277). In the inverse it also reaches the moved patient (*NX-278), with an exception for demonstrative, possessed and pronoun patients that we do not understand.
(c) N2 versus N2′ (reviews 4–6): still OPEN, because there is no doo on these pages. But NX-267 sharpens the test. An indefinite insidedoo … da that takes scope over ¬ would decide nothing, since H22 makes that possible under both. The deciding case is an indefinite pronounced left of doo and read inside ¬. N2′ allows that directly, and N2 needs reconstruction.
(d) HYPOTHESISWorking default, verdict for this batch: every derivation goes through with it, and nothing here resists it. Raising the subject to Spec,TP keeps the universal above the indefinite, and the “one boy” reading needs one extra step (H22a or H22b), which is what else has to happen; it is about scope, not order. There is no negation, adverb or hanii on these pages, so H20 and the review-6 predictions are untested.
Where we part from Willie’s analysis (labeled contrast)
HYPOTHESISAgrees with our derivations: her statement that “the agent falls under the scope of the distributive marker” (p. 82) is H21a. We derive it from VPISH plus a closest-argument condition, rather than stating it per role.
HYPOTHESISNot adopted: her conclusion that number belongs to the pronouns, not the nouns (p. 85). Every derivation above works with the nouns in argument positions, so nothing here forces nouns to be adjuncts. Her p. 84 account (da- lets a plural patient “escape the requirement on definiteness”) does not cover *NX-278: its bare singular patient is translated definite elsewhere (NX-269), so by that account it should be fine.
HYPOTHESISHer quantifier claim (“different interpretations of the quantifier”, p. 79) is recorded, but she prints no scope readings for (58)–(59), so we cannot test it as a scope claim.
What this changes in reviews 1–6
HYPOTHESIS H6/H6′ pass out of sample on a numeral plural (NX-263/264) and a universal (NX-265). H6-scope is still OPEN, because no reading is printed for (58).
HYPOTHESIS New: SR is contradicted (NX-267a). H22a vs H22b is OPEN. H21a (da- counts the nearest argument, the agent) holds in both voices but is contradicted by NX-282 unless da- can merge lower. H21b is partial.
HYPOTHESIS Review 1’s “Move versus realization” question gets mild support for Move: the inverse patient is not counted like the direct theme (*NX-278 vs NX-272).
Working default (V→v→T, subject to Spec,TP): every derivation uses it; nothing here resists it.
HYPOTHESIS Light d is still idle, and telicity is still OPEN.
OPEN predictions
OPENH6-scope: which readings does (58) have, one boy, different boys, or both? Ask speakers or find a source that prints them. We construct no Navajo string.
OPENH22a vs H22b: look for a łaʼ indefinite inside a relative clause or other island, with a printed wide-scope reading.
OPENThe H21b exception: why demonstrative, possessed and pronoun patients escape the plural requirement, while a pronoun agent does not.
OPENNX-282: where da- is merged with yitah, and whether the PP is an argument.
OPENN2 versus N2′ with a quantifier: a printed indefinite left of doo with its reading (see grid note c). Also the hanii test of the working default from review 6. Next in the queue: Willie ch. 6, and the grammar draft’s ch. 6 §§6–10.
2026-11-12 evening · First-principles review 8: the rest of the grammar draft’s negation chapter — phrases to the left of doo that are still negated, NPIs that hanii cannot license, and negative yes/no questions
2026-11-14: the “On the spine” summary at the top of this page supersedes this section’s framing. Its data, judgments and citations are unchanged.
Restated on the unified structure (2026-11-15). This section’s results now sit on the one structure for Navajo: unified structure, negation rows. Its restatement as derivations is in Results on the spine. Data, judgments and citations below are unchanged.
Source: Ellavina T. Perkins & Theodore B. Fernald, A Grammar of Navajo, working draft (revised 2014), Chapter 6 “Negation” (chapter PDF), printed pp. 103–122 (PDF pp. 24–43; printed page = PDF page + 79): the end of §5 (§5.2 incorporated indefinites, §5.3 the minimizer łáʼí ndi, §5.4 ndóʼ) and §§6–10 (negative imperatives, coordination, conditionals, double negation, time, causation, answer particles). As in review 6, the PDF’s text layer uses an old Navajo font, so every form below was read from 200- and 400-dpi page renders, with 800-dpi zooms for diacritics; the text layer was only a finding aid. Where the two disagree, the render decides (for example, the acute in ʼádeiilʼį́į sits on the first į in the render). New index entries are NX-284–309. H-numbers continue from review 7. Three strings on these pages were already indexed and are cited, not re-indexed: §5.2 (14) = NX-182, §5.4 (5) = NX-184, and §5.4 (6) = NX-185. OPEN For NX-185, the page render prints Ndóʼísh yiyą́ą́?, with an acute on both vowels, but the index entry has yiyą́ą. We flag this and do not change the older entry here. Our queue had said this range contains doo hanii … da as §10. That was wrong: §10 is about answer particles, and doo hanii … da is §3.6, already covered by review 6 (NX-247).
2026-11-13 audit pass 7. Grant’s instruction (2026-11-13): VPISH is not a word-order account; it frames the verb’s arguments so that syntax and semantics line up. In ordinary tensed clauses, V→v→T and subject raising Spec,vP→Spec,TP (on to Spec,CP when needed) are the working default, not GRANT’S CONJECTURE to be tested against a no-movement rival. This review was posted before that instruction. Its derivations, labels and verdicts on verb movement are rewritten here to use the default: the former no-movement rival is kept only as labelled contrast, each derivation says what else has to move, and examples that seemed to resist the default are kept as tests of it (OPEN). No Navajo form, gloss, judgment or citation changed. See the 2026-11-13 note and the pass 7 report.
HYPOTHESISPlain-words verdict. A phrase can be said beforedoo and still be inside the negation. The clearest case is “not even one dollar”, which can only be read under the negation, so at least for these NPIs, what doo and da enclose does not decide what is negated. The simplest account is that da is the negation and doo is a marker that sits lower in the same clause. One negation cannot reach across two joined clauses. hanii cannot license NPIs, so Grant’s split between sentential and constituent negation shows up in the grammar, not just in the meaning. A negative yes/no question asks about the negative proposition, and the question marker -ísh can be written on either side of the negative word. Light d is not needed. Telicity is still OPEN. No passive or inverse question turned up.
Starting point
Assumed, by instruction: VPISH (the subject argument first merges in Spec,vP) and theta roles; since audit pass 7 (2026-11-13), also the V→v→T working default stated below. HYPOTHESIS Everything else is tested: N1/N2/N2′, H1–H22 from reviews 1–7, light d, NegP head and specifier assignments, Agree. Since audit pass 7 (2026-11-13), V→v→T and subject raising Spec,vP→Spec,TP (on to Spec,CP when needed) are the working default in tensed clauses, assumed rather than tested: VPISH frames where the arguments start and get their roles; it is not a word-order account. An example that seems to resist the default is kept and flagged OPEN. Adverbs are pied-piped with VP, and only sentential negation counts in verb-movement tests. Every derivation ends in a logical form (LF) whose scope order matches c-command, and a rival fails if its LF contradicts a reading the source reports. The draft’s own analyses (the “negative frame”, “incorporation” of -í, its account of minimizers) appear only as labeled contrast.
The source’s claims, as its own
SOURCEDIncorporated -í. It “is ungrammatical without a negative frame” and “in the scope of constituent negation hanii”. “The argument that incorporates can originate inside a complement clause”, and in (19) “what is negated … is the wanting” (pp. 103–104, the last point credited to Hale & Platero 2000).
SOURCEDThe minimizer łáʼí ndi. It is “a fairly strict negative polarity item that is normally acceptable only within the negative frame”. It is ungrammatical in yes-no questions, in the restriction of a universal, and in conditional antecedents. Its one exception is “modifying an argument of the negative verb ʼádin”. Swapping in łáʼ often gives a grammatical counterpart, and łáʼ “is not a positive polarity item” (pp. 104–106).
SOURCEDndóʼ. It “appears immediately following doo in the negative frame”, it can host -ísh, and it is “unacceptable except within the negative frame doo…da”. That rules it out even with ʼádin and with hanii (pp. 106–108).
SOURCEDCoordination. “It is not possible to use one negation frame around two clauses joined by a conjunction.” In (28) and (30), “negation has scope over the conjunction”. In (31), “Negation has scope over doodaiiʼ even though doodaiiʼ has dislocated to the end of the sentence” (pp. 112–116).
SOURCEDNegative imperatives. One type uses tʼáadoo with an imperfective, second-person verb “that has the enclitic -í” (p. 110). The section credits Reichard (1951).
SOURCEDAnswers. After a negative yes-no question, “answering with the affirmative signifies an acknowledgment that the negative declarative sentence corresponding to the question is true” (p. 121).
The checked data
SOURCED Exact forms, with example numbers and printed pages. Source glosses, printing slips and crops are at the linked index entries.
A phrase to the left of doo (the quantifier test):NX-287 §5.3 (20), p. 104: Béeso łáʼí ndi doo naashʼáa da., translated “I haven’t so much as one dollar.” · NX-288 (21), p. 105: Díí chidí bikeeʼ łáʼí ndi doo ditsʼid da lá., translated “Not a one of these tires is any good.” · *NX-289 (22): Béeso łáʼí ndi naashʼá., starred · NX-296 §5.4 (2), p. 107: Béeso doo ndóʼ naashʼáa da., translated “I don’t have any money.” · NX-295 §5.4 (1), p. 106: Chidí bito doo ndóʼ holǫ́ǫ da., translated “There’s not even any gasoline.” · NX-302 §6.2.1 (2), p. 112: Bilasáana doo yíyą́ąda dóó doo yidlánída., translated “S/he isn’t eating an apple and isn’t drinking anything.”
Licensers other than doo … da:NX-292 (31), p. 106: Shibéeso łáʼí ndi ʼádin., translated “I don’t have even one bit of money.” · *NX-297 §5.4 (7), p. 107: Chidí bito ndóʼ ʼádin., starred (intended: “There’s not even any gasoline.”) · *NX-285 §5.2 (16), p. 104: Yishtłʼóí hanii, ʼashxaał gaʼ., starred (intended: “I’m not weaving anything; I’m drumming.”) · *NX-298 §5.4 (8), p. 107: Béeso ndóʼ hanii naashʼá., starred (intended: “I don’t have any money.”) · *NX-284 §5.2 (15), p. 104: ʼAshkii yiiyiłtsą́ní., starred, the affirmative of NX-182.
łáʼ and -í together:NX-293 fn. 12 (i), p. 106: Doo łáʼ nisinda., translated “I don’t want any.” · NX-294 fn. 12 (ii): Doo łáʼ nisinída., translated “I don’t want any one of them.”
-í from a lower clause:NX-286 §5.2 (19), p. 104: Shizhéʼé doo deeshááł nízinígóó da., translated “My father doesn’t want to go anywhere.”
-í in negative imperatives:NX-299 §6.1.3 (13), p. 110: Tʼáadoo ʼádíníní!, translated “Don’t say that!” · NX-300 (15): Tʼáadoo niyoochʼídí!, translated “Don’t be a liar/Don’t lie!”
One frame per clause; coordination inside the frame: *NX-301 §6.2.1 (1), p. 112: Doo gad ditʼin léiʼ nichxǫ́ǫ́ʼí da., starred (intended: “It is not the case that because the juniper is dense it is ugly.”) · NX-303 (3): Gad doo ditʼin da dóó nichxǫ́ǫ́ʼí., translated “The juniper is not dense and it is ugly.” · NX-304 §6.2.2 (28), p. 115: Tʼáadoo danitsaago dóó ʼádaałtʼą́ą́ʼígo ʼájiilaa da jiní., translated “She didn’t make them [tortillas] big and thin, it is said.” · NX-305 §6.2.3 (30), pp. 115–116: Nihí ʼéí náneeskaadí doo ʼakʼah dóó mandigiya bił ʼádeiilʼį́į da., translated “We (pl) don’t make tortillas with shortening and butter.” · NX-306 (31), p. 116: Nihí ʼéí náneeskaadí doo ʼakʼah bił ʼádeiilʼį́į da doodaiiʼ mandigiya da., translated “We (plural) do not make tortillas with shortening or such things as butter.”
Yes/no questions and answers:NX-307 §10 (3), p. 121: Łeetsʼaaʼísh táásinigiz?, translated “Did you wash the dishes?” · NX-308 (5): Łeetsʼaaʼísh tʼáadoo táásinigiz ?, translated “Didn’t you wash the dishes?” (the gloss ends in “neg”, but no da is printed) · NX-309 (6): Ndagaʼ tááségiz., translated “No, I washed them.” · NX-291 §5.3 (28), p. 106: Béesoósh łáʼ naniʼá?, translated “Do you have any money?” · *NX-290 (24), p. 105: Béeso łáʼí ndiísh naniʼá?, starred · cited: NX-184 §5.4 (5), p. 107, Doo ndóʼísh yiyą́ą da?, translated “Don’t you eat at all?”, and *NX-185 (6).
Our working glosses (word by word, not segmented; the source’s own glosses are on the index pages): NX-287 money MIN NEG carry.about.IPFV.1SG>3 NEG · NX-308 dishes=Q NEG wash.PFV.2SG>3 · NX-184 NEG even=Q eat.2SG NEG.
OPENNot indexed. §5.1 (9)–(13) (háʼátʼíida “whatchamacallit”), §5.2 (17)–(18), §5.3 (23), (25)–(27), (29)–(30), §5.4 (3)–(4), (9)–(15), most of §6.1, §6.2.1 (4)–(27), (29), (32)–(33), §6.3, §7, §8, §9, and §10 (1)–(2), (4), (7)–(9). Many are quoted second-hand from Young & Morgan (1987) or Reichard (1951), and nothing below depends on them. Some have printing problems we will not repair. §7 (5a) and (5b) print the same form with opposite translations. §10 (7) prints “t;ahdoo”. In §6.2.3 (33), the gloss line (“Ball-Q … 3-over-Q”) does not match the form. §8 (5), “it was never seconded”, has only a partial gloss, and an English passive translation would not fill a Navajo passive cell anyway.
Hypotheses tested, revised or added
N2 versus N2′ (reviews 4–6), the quantifier test. N2: ¬ scopes over exactly the material between doo and da. N2′: ¬ sits at the final da and covers the clause da closes. doo is free inside that clause but must precede any NPI (the “rider”).
H23 · where doo sits (new). The negation is the final head Neg⁰ da. doo (and tʼáadoo) is not its specifier. It is adjoined lower, inside the complement of Neg⁰, in the same clause, and marks that clause as negative. A phrase can therefore be pronounced left of doo and still be c-commanded by da. Rival: review 5’s placement, with doo in Spec,NegP.
H17 (review 5), NPI concord, sharpened. An NPI needs a sentential negation in its clause, reached by Agree, not just any negative meaning above it.
H18a′ (refines review 5’s H18a).łáʼ can be a determiner on the -í pronoun, so the two make up one argument between them.
H24a / H24b′ (new). Two ways a disjunct pronounced after da can be read under ¬.
H25a (new). The -í of tʼáadoo imperatives is a different, homophonous enclitic. Rival: H18c, revived (no -í is an argument).
H26 (new). Is the negation of a negative verb like ʼádin read inside the verb word, or on a clause-level head?
H12 (review 3), Q displacement. Tested on negative questions.
Working default (V→v→T, subject Spec,vP→Spec,TP, on to Spec,CP when needed): assumed in every tensed clause, not tested against a no-movement rival (audit pass 7, 2026-11-13). Each derivation says what else has to move.
Derivations, step by step
HYPOTHESIS Each derivation gives Merge with theta assignment, Agree (or “none”), Move, which copy is pronounced, what is inside the verb word versus outside it, and finally the LF. Extra assumptions are marked. Brackets put specifiers on the left. Heads are final unless a derivation says otherwise. The inflected verb is not segmented beyond the source’s hyphens.
NX-287, with *NX-289 and *NX-290 (an NPI phrase left of doo). Merge V “carry about/have” with the phrase béeso łáʼí ndi as its complement (Theme). We treat the phrase as one constituent, because it moves as a unit; its inner structure is not analyzed. Merge v. The first-person subject merges in Spec,vP (Agent). It is silent, and the verb word marks it (source gloss 3-1). Merge T. Agree: the verb’s object and subject marking matches the Theme and Agent (HYPOTHESIS). Merge the negation: doo and the final da. Move: the Theme is pronounced left of doo, but V assigns it its role inside VP, so it must have moved, on every analysis. Its lower copy is silent. LF, from the translation and the fact that the phrase is a strict NPI: ¬∃x[money(x) ∧ MIN(x) ∧ ∃e[carry_about(e) ∧ Agent(e,speaker) ∧ Theme(e,x)]], with ¬ outscoping the minimizer. N2: the moved phrase is outside the doo … da span. To get this LF, the LF must read the lower copy (reconstruction, an extra assumption). N2′ with doo in Spec,NegP: the landing site left of doo is still above NegP, so this also needs reconstruction. N2′ with H23: the phrase lands on TP, above doo but inside the complement of da. That gives [NegP [Neg′ [TP Min [TP doo [TP pro [T′ [vP pro [v′ [VP Min V] v]] T]]]] da]] (lower copies silent). da c-commands the pronounced copy, and the LF is read off the surface. *NX-289: there is no negation, so H17 leaves the NPI unlicensed. *NX-290: a yes/no question alone does not license it either. NX-291, with plain łáʼ, is fine in a question, because łáʼ is not an NPI. Result: a phrase left of doo is read inside ¬, and has to be. Strict N2, without reconstruction, is contradicted. N2 plus reconstruction survives at the cost of one assumption. N2′ derives the reading with no reconstruction, but only together with H23. N2′’s rider, that doo precedes any NPI, is contradicted by NX-287 and NX-288. It must be cut back to the h…da words and -í, the data it was built on. Call the result N2″.
NX-288 (a subject minimizer left of doo). Merge stative V “be tough” and v. The subject díí chidí bikeeʼ łáʼí ndi merges in Spec,vP (VPISH; theta: Theme of the state). Merge T. Working default: V→v→T, the subject moves to Spec,TP, and then once more, to the H23 position above doo. That second move is what else has to move (HYPOTHESIS). lá (glossed DISC) follows da, so it merges above NegP (C region, HYPOTHESIS). LF: lá(¬∃x[tire_of(x,this_car) ∧ MIN(x) ∧ ∃s[tough(s) ∧ Theme(s,x)]]). Result: the same as derivation 1, now with a subject. The move over doo would be needed on any analysis, so nothing here resists the default.
NX-296 and NX-295 (a plain bare noun left of doo).béeso is not an NPI, and it is translated “any money” under the negation. LF: ¬ EVEN ∃x[money(x) ∧ ∃s[have(s) ∧ Holder(s,speaker) ∧ Theme(s,x)]]. For NX-295, with the subject chidí bito (“gasoline”): ¬ EVEN ∃x[gasoline(x) ∧ ∃s[be_there(s) ∧ Theme(s,x)]]. The derivation is as in 1, and H23 gives the scope without reconstruction. So an ordinary indefinite left of doo also takes scope under ¬. OPEN Whether it could also take scope over ¬ (“there is some money I don’t have”) is not printed. OPENndóʼ must come right after doo (source). H23 does not derive that. One option is that ndóʼ adjoins to doo itself.
NX-292 versus *NX-297 (a negative verb as licenser).ʼádin “not exist” licenses łáʼí ndi without doo … da, but not ndóʼ. Merge V ʼádin with shibéeso łáʼí ndi. We assume “not exist” takes its one argument as its complement (Theme), with no Spec,vP subject. This is a HYPOTHESIS about this verb. LF: ¬∃x[money(x) ∧ Poss(x,speaker) ∧ MIN(x) ∧ exist(x)]. Working default: if the lone argument raises to Spec,TP (as an internal argument, by extension of the default), T, which now holds V and its negation, does not c-command it. Then the scope needs reconstruction (the lower copy is read at LF, HYPOTHESIS), or this argument does not raise (OPEN). Labelled contrast: with no movement, the Theme would stay as V’s complement under the negation inside V, and the scope would come for free. H26: ʼádin’s negation may instead Agree with a silent clause-level Neg⁰ above the subject position. Result:łáʼí ndi needs a sentential negation (from doo … da or inside ʼádin), while ndóʼ needs doo itself. That is selection, not scope. OPEN NX-292 is a sourced example that resists the default: it needs reconstruction, or no raising for this argument (H26 is a third option). It is kept as a test of the default.
*NX-285 and *NX-298 versus NX-242 (hanii does not license NPIs). H20: hanii merges right after its focus XP. NX-242, a verb word followed by hanii, is fine. *NX-285 has the same configuration and the same kind of verb focus (“not weaving …; drumming”). It differs only in having -í in the verb word, and it is out. *NX-298 puts ndóʼ under hanii, and it is out too. So the star comes from the NPI, not from where hanii sits. Note what scope alone predicts. In the intended reading, the -í pronoun would be inside the denied focus, so semantically it is under a negation: ¬[FOC = λx.∃y weave(x,y)](speaker) ∧ drum(speaker). Scope alone would let it through. Result: NPI licensing needs a sentential negation reached by Agree (H17 sharpened), and constituent negation does not count. This bears out GRANT’S CONJECTURE’s split between sentential and constituent negation in Navajo. The two differ in what they license, not only in what they mean. It does not bear on V→T. OPEN These pages print no hanii-after-verb string with a whole-clause reading, so H20 test (a) from review 6 is still open.
*NX-301 versus NX-303 (one frame, two clauses). Merge clause A, gad ditʼin “the juniper is dense”, and clause B, nichxǫ́ǫ́ʼí “(it is) ugly”, and join them with léiʼ. N2 lets ¬ cover everything between doo and da. The intended LF, ¬[dense(juniper) CAUSE ugly(juniper)], makes sense, so N2 alone does not rule out the string. N2′ with H23:da is the Neg⁰ of clause B. doo must be inside the complement of that Neg⁰, but here it is in clause A, so the string is out. NX-303 puts doo … da both in clause A: ¬∃s[dense(s) ∧ Theme(s,juniper)] ∧ ∃s′[ugly(s′) ∧ Theme(s′,juniper)]. Result: N2 as stated is contradicted (it needs a clause-mate condition added). N2′ with H23 derives the star.
NX-304 and NX-305 (∧ inside ¬) versus NX-306 (∨ after da). NX-304: merge V “make (thus)” with a Theme (“them”, silent) and a fourth-person Agent in Spec,vP. Two -go manner phrases, coordinated by dóó, merge in VP. Then T and tʼáadoo … da. jiní “it is said” follows da, so it is a higher verb that takes the negated clause (HYPOTHESIS, from the translation). LF: say(people, ¬∃e[make(e) ∧ Agent(e,she) ∧ Theme(e,them) ∧ big(e) ∧ thin(e)]). This matches the source’s truth conditions, “true as long as the tortillas were not both big and thin”. NX-305 is the same with “with shortening and butter”: ¬ over ∧. NX-306: the disjunct doodaiiʼ mandigiya da comes after the first da, yet the source says it is read under ¬ (“neither shortening nor butter”). LF: ¬∃e[make(e) ∧ Agent(e,we) ∧ Theme(e,tortillas) ∧ (with(e,shortening) ∨ with(e,butter))]. On both N2 and N2′ with H23, the pronounced disjunct is outside the material da c-commands. H24a: the disjunct moved rightward from inside the clause, and LF reads its lower copy. H24b′: it is an elided second negative clause, with doodaiiʼ meaning “and not”/“nor” and the second da as its negative marker. The plain “or not” version, ¬A ∨ ¬B, is contradicted by the source’s truth conditions. OPEN Each of H24a and H24b′ costs one assumption. This is the first sourced case in the project of material after da that is read under the negation.
NX-286 (-í from the lower clause). Lower clause: V “go” with a Goal, the -í pronoun with -góó “to”, and an Agent in Spec,vP. The source gloss is 1-go.Fut, so the complement seems to be the wanter’s own words, as in NX-238 (HYPOTHESIS). Main clause: V “want” takes that clause as Theme, and shizhéʼé merges in Spec,vP (Experiencer). Then doo … da in the main clause. -í-góó is pronounced on the main verb word (nízinígóó). By review 5’s H18a, the Goal phrase moves out of the complement to the edge of the main clause, below da, and at PF it leans on the main verb. The complement of “want” is not an island, so the move is allowed. LF: ¬∃s[want(s) ∧ Exp(s,father) ∧ Theme(s, ∃x[place(x) ∧ ∃e[go(e) ∧ Agent(e,father) ∧ Goal(e,x)]])]. Result: H18a passes out of sample with a Goal from another clause, and ¬ belongs to the main clause, closed by its da (as N2′ says). ¬ over ∃ is forced. OPEN Whether ∃ is read inside “want” (as written, which needs reconstruction) or above it is not settled by the translation.
NX-293 / NX-294 (łáʼ and -í together). “Want” has two roles, Experiencer (first person, in Spec,vP) and Theme. NX-294 has both an overt łáʼ and -í, and it is not starred. Review 6 counted -í as an argument (*NX-257/258, “too many arguments”). If łáʼ were a second Theme, the theta count would rule NX-294 out. H18a′:łáʼ is a determiner whose restriction is the -í pronoun, so together they are one Theme. This fits the translation shift from “any” to “any one of them”. LF NX-293: ¬∃x[∃s[want(s) ∧ Exp(s,speaker) ∧ Theme(s,x)]]. LF NX-294: ¬∃x[x ∈ C ∧ one(x) ∧ ∃s[want(s) ∧ Exp(s,speaker) ∧ Theme(s,x)]], where C is the contextual set “them”. *NX-258 differs: there háíida is a full pronoun phrase, not a determiner. OPEN Our prediction, not data: a full noun as object plus -í should be out, as in *NX-257.
NX-299 / NX-300 (-í in tʼáadoo imperatives). “Don’t lie!” is built on an intransitive verb (2-lie). Its one role, Agent, goes to the addressee in Spec,vP. If this -í were the argument pronoun, the theta axiom would leave it with no role, yet the sentence is fine. LF: DIR(¬∃e[lie(e) ∧ Agent(e,you)]), with the directive over ¬. No da is printed. H25a: this is a homophonous enclitic -í. The draft mentions a nominalizing -í in another context (fn. 11). H18c revived: no -í is ever an argument, and review 6’s theta count would need another explanation. OPEN H25a keeps review 6’s result, but it is a stipulation.
The yes/no question cells: NX-307, NX-308 (+ NX-309), NX-184, NX-291. NX-307: merge V “wash” with łeetsʼaaʼ “dishes” (Theme) and a second-person Agent in Spec,vP. Then T, then a silent C[Q, polar] that Agrees with the noun hosting -ísh. LF: ASK_POL(∃e[wash(e) ∧ Agent(e,you) ∧ Theme(e,dishes)]). NX-308 adds tʼáadoo, and the Theme with -ísh is pronounced to its left (Move, as in derivation 1; the noun is definite, so its scope does not matter). LF: ASK_POL(¬∃e[wash(e) ∧ Agent(e,you) ∧ Theme(e,dishes)]). The source’s answer facts force Q over ¬. “Yes” (ʼaooʼ) confirms the negative proposition. Ndagaʼ tááségiz (NX-309) rejects it and asserts the affirmative. So the question’s inner proposition is ¬p. NX-184 (tree below): -ísh is written on ndóʼ, inside the frame and under ¬, yet the question operator is over ¬. So the written -ísh is not where the question operator is interpreted. The operator is in C and reaches its host by Agree. This is review 3’s H12 (displacement), now with a negation in between. NX-291: -ísh is on béeso, ahead of łáʼ. OPEN Is béeso łáʼ one phrase (so that -ísh sits inside it) or two?
Analysis A: Working default (V→v→T, subject Spec,vP→Spec,TP); H23: doo inside the complement of Neg⁰ Hypothesis
Analysis B: Working default (V→v→T, subject Spec,vP→Spec,TP); doo in Spec,NegP (review 5 placement) Hypothesis
(1) A negative polar question with an NPI: Doo ndóʼísh yiyą́ą da? (NX-184) HypothesisNX-184, Perkins & Fernald grammar draft ch. 6 §5.4 (5), printed p. 107: Doo ndóʼísh yiyą́ą da? “Don’t you eat at all?”, source gloss “neg NDÓʼ-Q 2-eat.? neg”. Leaves doo, ndóʼ, yiyą́ą, da are printed tokens of that string; splitting ndóʼísh into ndóʼ + -ísh is our analysis. AdvP, Neg⁰ and C⁰[Q, polar] are our category assignments. The agent DP is a placeholder: no separate pronoun is printed.Bracket notation (copy)
A: [CP [C′ [NegP [Neg′ [TP [AdvP doo] [TP [AdvP ndóʼ] [TP [DP agent: you (silent; second person is marked only in the verb word)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [V⁰ yiyą́ą]]] v⁰]] T⁰]]]] [Neg⁰ da]]] C⁰]]
B: [CP [C′ [NegP [AdvP doo] [Neg′ [TP [DP agent: you (silent; second person is marked only in the verb word)] [T′ [vP [AdvP ndóʼ] [vP ⟨DP⟩ [v′ [VP [V′ [V⁰ yiyą́ą]]] v⁰]]] T⁰]] [Neg⁰ da]]] C⁰]]
A: [CP [C′ [NegP [Neg′ [TP [AdvP doo] [TP [AdvP ndóʼ] [TP [DP agent: you (silent; second person is marked only in the verb word)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [V⁰ yiyą́ą]]] v⁰]] T⁰]]]] [Neg⁰ da]]] C⁰]]
B: [CP [C′ [NegP [AdvP doo] [Neg′ [TP [DP agent: you (silent; second person is marked only in the verb word)] [T′ [vP [AdvP ndóʼ] [vP ⟨DP⟩ [v′ [VP [V′ [V⁰ yiyą́ą]]] v⁰]]] T⁰]] [Neg⁰ da]]] C⁰]]
Structure as a list
Analysis A
CP
C′
NegP
Neg′
TP
AdvP (id doo): doo
TP
AdvP (id host; goal of Agree Q-host): ndóʼ
TP
DP (id S-high; landing site of Move from subject): agent: you (silent; second person is marked only in the verb word)
T′
vP
DP (id subject; struck copy; moved to S-high): agent: you (silent; second person is marked only in the verb word)
v′
VP
V′
V⁰ (head; id V; head-moves to v): yiyą́ą
v⁰ (head; id v; head-moves to T)
T⁰ (head; id T; hosts the “verb word” (V + v + T))
Neg⁰ (head; id Neg): da
C⁰ (head; id C; [Q, polar]; probe for Agree Q-host with host)
Analysis B
CP
C′
NegP
AdvP (id doo): doo
Neg′
TP
DP (id S-high; landing site of Move from subject): agent: you (silent; second person is marked only in the verb word)
T′
vP
AdvP (id host; goal of Agree Q-host): ndóʼ
vP
DP (id subject; struck copy; moved to S-high): agent: you (silent; second person is marked only in the verb word)
v′
VP
V′
V⁰ (head; id V; head-moves to v): yiyą́ą
v⁰ (head; id v; head-moves to T)
T⁰ (head; id T; hosts the “verb word” (V + v + T))
Neg⁰ (head; id Neg): da
C⁰ (head; id C; [Q, polar]; probe for Agree Q-host with host)
Does light d earn its keep?
HYPOTHESISFair test. A light-d analysis would put a silent D⁰ on the bare nouns here. Some are read as indefinites under negation (béeso, chidí bito, bilasáana), and some as definites (łeetsʼaaʼ “the dishes”, gad “the juniper”). Result: light d is not needed, and nothing here counts against it. (a) The narrow indefinite readings come from existential closure under ¬ in each LF above (derivations 1–3). A silent D adds nothing to that. (b) The definite nouns are scopeless, so no contrast turns on them. (c) The one determiner-like element here, łáʼ, is overt (H18a′), so it is not evidence for a silent d. Light d stays idle.
Telicity test
OPENPrediction tested: whether a negated sentence’s reading, or where -í or ndóʼ can go, depends on the verb’s telicity. These pages have no pair that differs only in telicity. The ndóʼ examples have states (“have”, “be there”, “look sick”, “be cold”), and the -í examples have “see”, “weave”, “go” and “lie”, all with different verbs and frames. The source’s labels (Imp, Pf, NI, N) mark mode and viewpoint, not telicity. Neither supported nor contradicted: OPEN.
Question × polarity × voice grid (cumulative after review 8)
Question cell
Active / direct
Inverse / bi-
Passive / impersonal (source gloss)
Polar, affirmative
NX-203, 204; NX-119–134, NX-142; new:NX-307, NX-291 (*NX-290: minimizer not licensed)
NX-053; *NX-052
OPEN (NX-217 candidate only)
Polar, negative
NX-205–209 (alternative “or not”); NX-153–161, NX-184; new:NX-308 (tʼáadoo, -ísh host left of the negative word), answer NX-309
OPEN
OPEN (declaratives NX-216 and NX-236 only)
Wh, affirmative
Review 3 set; NX-212; NX-220, 221
Review 3 set
NX-213, 215
Wh, negative
NX-210, 211; NX-138; NX-247 (by reading)
OPEN
OPEN
HYPOTHESISWhat this batch adds and forces:
(a) Q over ¬ is forced in negative yes/no questions. NX-308 shows it through the source’s answer facts, and NX-184 through the NPI under ¬. In both, C c-commands NegP.
(b) Where -ísh is written relative to the negative word is optional. It can come before it (NX-308, and NX-153 on doo itself) or after doo (NX-184). The written host does not mark where Q takes scope (H12).
(c) A yes/no question does not license the NPIs łáʼí ndi or ndóʼ (*NX-290, *NX-185). Unlike English “any”, these need a sentential negation (H17).
(d) No passive or inverse question on these pages, affirmative or negative. Those cells stay OPEN. Nothing here bears on H13-int (content =sh across doo).
(e) HYPOTHESISWorking default, verdict for this batch: every derivation goes through with it. Grant’s sentential/constituent split is grammatically visible (derivation 5). What else has to move: every phrase left of doo moves above it, as on any analysis (derivations 1–2). One example, NX-292, resists the default unless the lone argument of ʼádin reconstructs or does not raise (derivation 4); this is OPEN.
Where we part from the source’s analyses (labeled contrast)
HYPOTHESISNot adopted: the draft’s “negative frame” suggests that doo and da enclose what is negated. NX-287, NX-288, NX-295 and NX-296 put negated material before doo, and NX-306 after da. In our terms, the frame marks which clause is negative, while scope comes from da’s position and c-command (H23).
HYPOTHESISAgrees with our derivations: the draft’s list of where łáʼí ndi and ndóʼ fail (questions, universal restrictions, conditionals, wishes, hanii) is what H17, sharpened, predicts once only sentential negation counts.
HYPOTHESISNot adopted as analysis: “incorporation” of -í as head movement. Review 5’s phrasal Move plus PF placement covers NX-286.
What this changes in reviews 1–7
HYPOTHESISN2 versus N2′, the quantifier version, is now tested. A phrase left of doo is read under ¬ (NX-287/288/295/296), so strict N2 is contradicted. N2′ is favored if it is paired with H23. Its NPI rider is cut back (N2″). N2 also fails the clause-mate test (*NX-301).
HYPOTHESISH17 is sharpened: licensing needs a sentential negation reached by Agree (*NX-285, *NX-298, NX-292 versus *NX-297).
HYPOTHESISH18a passes with a Goal from another clause (NX-286). H18a′ is added (NX-294). Review 6’s claim that -í is an argument holds only if the imperative -í is a different morpheme (H25a).
HYPOTHESISH12 is supported: where -ísh is written does not show where Q takes scope (NX-184).
Working default (V→v→T, subject to Spec,TP): used throughout; NX-292 resists it unless its argument reconstructs or does not raise (OPEN).
GRANT’S CONJECTURE The sentential/constituent split is confirmed by licensing.
HYPOTHESIS Light d is still idle, and telicity is still OPEN.
OPEN predictions
OPENWide scope left of doo: can a plain indefinite left of doo (not an NPI) also be read above ¬? N2′ with H23 allows it only through further movement above NegP.
OPENThe hanii test of the working default (review 6): a whole-clause reading of a hanii-after-verb string is still not printed anywhere we have read.
OPENH24a versus H24b′ for a disjunct after da (NX-306). H25a versus H18c for the imperative -í. H26 for where the negation of ʼádin is read.
OPENWhy ndóʼ must follow doo directly.
OPENStill missing from the grid: a polar passive question, and negative inverse and negative passive questions. H13-int is still untested.
OPENIndex discrepancy: NX-185’s form should be rechecked against the render (yiyą́ą́ printed).
2026-11-13 evening · First-principles review 11: Willie on “which” phrases, wh-words with “you think” and “she said”, and possessors in place phrases (ch. 8, §§8.5–8.6)
2026-11-14: the “On the spine” summary at the top of this page supersedes this section’s framing. Its data, judgments and citations are unchanged.
Source: MaryAnn Willie, Navajo Pronouns and Obviation (PhD dissertation, University of Arizona, 1991), Chapter 8 “Questions and Focus”, the remainder not covered by review 3: §8.5 “Wh- phrases” and §8.6 “Long distance wh-movement”, examples (18)–(32), printed pp. 212–218 (PDF pp. 218–224; PDF page = printed page + 6). Our queue had these as pp. 210–216; the renders show (18) at the top of p. 212 and (32) on p. 218. Every form was read from 200-dpi page renders and 400-dpi crops, with 800-dpi zooms for accents and nasal hooks; the text layer was only a finding aid. New index entries are NX-374–385. Already indexed and cited, not re-indexed: (20) = NX-113; (27a/b) = NX-087/NX-088. Not indexed: (23), (24) and (30a/b), which are English only. H-numbers continue from review 10 (H35).
2026-11-13 audit pass 7. Grant’s instruction (2026-11-13): VPISH is not a word-order account; it frames the verb’s arguments so that syntax and semantics line up. In ordinary tensed clauses, V→v→T and subject raising Spec,vP→Spec,TP (on to Spec,CP when needed) are the working default, not GRANT’S CONJECTURE to be tested against a no-movement rival. This review was posted before that instruction. Its derivations, labels and verdicts on verb movement are rewritten here to use the default: the former no-movement rival is kept only as labelled contrast, each derivation says what else has to move, and examples that seemed to resist the default are kept as tests of it (OPEN). No Navajo form, gloss, judgment or citation changed. See the 2026-11-13 note and the pass 7 report.
HYPOTHESISPlain-words verdict. When a question word sits in a clause followed by “you think” or “she said”, it stays inside that clause, and a “when” there can only ask about that clause’s event. That is easiest to explain if “you think (it)” and “she said (it)” are separate clauses set beside the question, not clauses that contain it. The rival, where the question clause is the object of “think”, gets the word order and the “Who do you think…?” paraphrase for free, but it needs one extra stipulation to block the “when did she say it” reading that Willie reports is missing. Phrases like “which boy” and “how many boys” act exactly like ordinary nouns. The yi-/bi- possessor inside “at his/her house” changes while the verb stays the same, so there it marks who the possessor is, not voice. Light d is not needed. Telicity: OPEN. No negation and no passive are on these pages, so only wh-affirmative cells gain examples.
Starting point
Assumed, by instruction: VPISH (the subject argument first merges in Spec,vP) and theta roles; since audit pass 7 (2026-11-13), also the V→v→T working default stated below. HYPOTHESIS Everything else is tested: H1–H35 from reviews 1–10 (here mainly H3, H3-loc, H4″, H9′, H11, H12, H32a, H34), light d, the categories D⁰, C⁰ and P⁰, Agree. Since audit pass 7 (2026-11-13), V→v→T and subject raising Spec,vP→Spec,TP (on to Spec,CP when needed) are the working default in tensed clauses, assumed rather than tested: VPISH frames where the arguments start and get their roles; it is not a word-order account. An example that seems to resist the default is kept and flagged OPEN. Adverbs are pied-piped with VP. Every derivation ends in a logical form (LF) whose scope order matches c-command. ASK_WH(λx …) is the question operator, written where its C⁰ is. Tense is left out of the LFs. Willie’s analysis (nominals and wh-words as adjuncts, no complement clauses, “inner speech”, wh-movement only at LF) appears only as labeled contrast.
The source’s claims, as its own
SOURCEDWhich-phrases. “presuppositional Wh- phrases, that are used to ask about members of a presupposed set; these are expressions like “which boy?”” (p. 211). “In these phrases, we see a Wh- word combined with an ordinary nominal to build a complex adjunct; extraction from these phrases is not possible” (p. 212). No starred extraction string is printed.
SOURCEDVerbs of thinking and saying. Propositional clauses with these verbs, “when they are without complementizers, are not complement clauses, but loosely adjoined as “inner speech””; the evidence is “the patterns of pronominal agreement across the main and adjoined inner speech clauses”, which is “like direct quotes in English” (p. 213). That agreement evidence is not printed in this chapter (only English (23)).
SOURCEDNo long-distance wh. “In Navajo, the Wh- word cannot undergo this kind of movement”; “the subordinate clause must precede the main clause, and the Wh- word remains in the clause it has scope over” (p. 214). For adverbial wh, English (30) is ambiguous; “Ambiguous constructions of this kind do not occur in Navajo” (p. 217). “Wh-words, whether subcategorized for or not, must appear in the clause they have scope over at LF”; Schauber (1979) and Speas (1990) suggest “some variation across speakers”; “I report on usage known to me” (p. 218).
SOURCEDPlace phrases.baghan-di “is an adverbial adjunct … Lexical Postpositions … are exclusively locative and directional”; only Grammatical Postpositions “fall under the scope of the Inverse voice alternation” (p. 215). The possessor contrast in (28) “is not one of voice, but shows the use of this alternation to mark obviation … Speakers vary in the extent to which they use yi-/bi- in this way” (p. 216).
The checked data
SOURCED Exact forms, with example numbers and printed pages. Source glosses, printing slips and crops are at the linked index entries. A question mark is part of the form only where the Navajo line prints one.
Complex wh-phrases (p. 212):NX-374 (18): ʼashkii háí-dí-lá yiztsʼǫs, “She kissed which boy?” · NX-375 (19): ʼashiiké díkwíí=sh yiztsʼǫs, “She kissed how many boys?” · cited NX-113 (20): hastiin łį́į́ʼ háídí=ígíí=sh yizloh, “The man roped which horse?”
With nínízin “you think” (pp. 213–214):NX-376 (21): háísh łį́į́ʼ yidooloh nínízin, “Who will rope the horse, you think?” (paraphrase: “Who do you think will rope the horse?”) · NX-377 (22): John łį́į́ʼ yidooloh nínízin, “John will rope the horse, you think.” · NX-378 (25): John háísh yiztsʼǫs nínízin, “John kissed who, you think?” (paraphrase: “Who do you think John kissed?”)
Adverbial wh and place phrases (pp. 214–217):NX-379 (26a): hádáaʼshąʼ John Mary yiztsʼǫs?, “When did John kiss Mary?” · NX-380 (26b): háísh baghan=di Mary yiztsʼǫs?, “At whose home did he kiss Mary?” · NX-381 (28a): John háísh baghan=di yiztsʼǫs?, “Who did John kiss at his house?” · NX-382 (28b): John háísh yighan=di yiztsʼǫs?, “Who did John kiss at her house?” · NX-383 (29): John háísh haghan=di yiztsʼǫs, “Who did John kiss at that other person’s house?”
“When” with “say” (pp. 217–218):NX-384 (31): hádááʼshąʼ dah diiyá ní, “When did he leave, she said.” · NX-385 (32): hádááʼshąʼ ʼání, dashdiiyá=hágíí ʼáhóótʼiid, “When did she say it, his leaving took place.” (paraphrase: “When did she say that he left?”)
SOURCEDPrinting slips, kept as printed. (26a) has hádáaʼshąʼ with no acute on the last a of the stem, where (31) and (32) have hádááʼshąʼ. The prose writes baghan-di (p. 215) and háísh bighan-di (p. 216), while the examples print baghan=di. (21) is glossed “3.3.roped” but translated “will rope”. =sh is glossed FOCUS in (19) and Q in (20). (29) has no question mark. (28a) and (28b) gloss both baghan and yighan as “3POSS:home”.
HYPOTHESISOur editorial glosses (not the source’s). Only where the parts are clear; the segmentations are ours. (19): boys how.many=Q 3SG.SBJ>3.OBJ.kiss.PFV. (26b): who=Q 3.POSS-home=at Mary 3SG.SBJ>3.OBJ.kiss.PFV. (28b): John who=Q 3.POSS-home=at 3SG.SBJ>3.OBJ.kiss.PFV. (32): when=Q 3.SBJ>3.OBJ.say.PFV 4.SBJ.leave.PFV=NMLZ 3.SBJ.happen.PFV. We do not segment háí-dí-lá beyond the source’s hyphens, and we leave the ʼá- of ʼání unglossed.
Hypotheses tested, revised or added
Carried over (tested again): H3 (the inverse is the patient internally merged above the agent; with two nominals and a direct verb the first is the agent), H3-loc (the inverse can raise a grammatical-postposition object, which is an argument), H4″ (a lone nominal goes with the lower third-person participant), H9′ (yi- needs a distinct third-person co-argument above it), H11 (a wh-word is an indefinite plus Q), H12 (Q can be displaced), H32a (a fourth-person and a third-person participant are disjoint), H34 (an unlinked enclitic D⁰ names a fact).
H36 (new) · complex wh-phrases. A noun and a following wh-word (“which”, “how many”) form one nominal, with the wh-word last, like other final heads. The phrase gets one theta role. Q is on its right edge (-lá in (18) as our reading, =sh in (19), outside =ígíí in (20)).
H37 (new) · possessor yi-/bi-/ha- by reference. Inside a place phrase, the possessor prefix is a pronoun. It is yi- when the possessor is a third person distinct from the clause’s third-person subject, ha- when it is fourth person, and bi-/ba- otherwise. This is H9′ carried from the verb to a possessor, for the speakers Willie reports.
H38a / H38b (new, rivals) · a patient before a place phrase. H38a: a place phrase adjoins to VP, and a patient pronounced before it has moved leftward past it. H38b: a place phrase can also attach below the patient, inside VP.
H40a / H40b / H40c (new) · clauses with nínízin and ní. H40a: the clause before nínízin/ní is a complete clause of its own (with its own C⁰); the thinking/saying clause is adjoined to it, and its Theme is a silent pronoun resolved to that clause. H40b: the first clause is the Theme complement of “think/say”, and V is final, so it precedes the verb; a question operator can sit in the matrix C⁰. H40c: an add-on to H40b, saying nothing of the matrix clause may be pronounced before the embedded clause.
Derivations, step by step
HYPOTHESIS Each derivation gives Merge with theta assignment, Agree, Move, which copy is pronounced, what is inside or outside the verb word, and the LF. Extra assumptions beyond VPISH and theta roles are named, and all are HYPOTHESIS. Specifiers are on the left and heads are final. Every tensed clause below uses the working default (V→v→T, subject Spec,vP→Spec,TP), even where a step does not repeat it; with a final T⁰ nothing else has to move for the verb-final order.
NX-374 (18), NX-375 (19), cited NX-113 (20): “which boy”, “how many boys”. Merge V “kiss” with the phrase [ʼashkii + wh-word] (H36) as complement; it takes the Patient role. Merge v, then a silent third-person pronoun in Spec,vP, which takes the Agent role (VPISH). Agree: none needed. Move: none. The only overt nominal is the wh-phrase, so H4″ puts it with the lower participant, the patient, and the printed “She kissed which boy?” follows. In (20) hastiin and the wh-phrase are two nominals with a direct verb, so hastiin is the agent (H3). Spell-out: the phrase outside the verb word, then the verb word. LF (18): ASK_WH(λx[boy(x) ∧ x ∈ C ∧ ∃e[kiss(e) ∧ Agent(e,she) ∧ Patient(e,x)]]), with C the presupposed set (Willie, p. 211). LF (19): ASK_WH(λn ∃X[boys(X) ∧ |X| = n ∧ ∃e[kiss(e) ∧ Agent(e,she) ∧ Patient(e,X)]]). LF (20): ASK_WH(λx[horse(x) ∧ ∃e[rope(e) ∧ Agent(e,man) ∧ Patient(e,x)]]). Result: H3 and H4″ pass out of sample on complex wh-phrases. H36 also offers an answer to review 3’s OPEN question about *NX-064 (háísh shizhéʼé naalnish): with the wh-word first, the two words cannot be one H36 phrase, so they are two nominals competing for one role (theta biuniqueness). A rival (háí “who” never takes a noun) gives the same star. OPEN The test, shizhéʼé before háísh, is not printed. Willie’s “no extraction” claim has no starred string, so it is OPEN too. H12 makes no prediction here: Q already sits after the first constituent, the whole phrase.
NX-379 (26a): “when”. Merge V “kiss” with Mary (Patient), v, John in Spec,vP (Agent). Two nominals, direct verb: H3 gives John = agent, which is correct. Merge the time indefinite + Q (H11) as an adjunct. It is pronounced before John, so it is above Spec,vP: under the working default (John raised to Spec,TP), in Spec,CP or adjoined to TP (which of the two is OPEN). Agree: Q with C⁰. LF: ASK_WH(λt ∃e[kiss(e) ∧ Agent(e,John) ∧ Patient(e,Mary) ∧ at(e,t)]). Result: fits. OPEN Whether “when” must come first is open: no order with it after the subject is printed. Review 3’s háadish “where” (NX-063, NX-065) is also first.
NX-380 (26b) versus NX-381 (28a): one substring, two parses. (26b): Merge háí+Q as possessor inside [baghan]=di. The postposition relates “home” to the event (location), and the phrase gets no role from “kiss” (an adjunct, as Willie says). Merge V with Mary (Patient) and a silent agent in Spec,vP. Mary is the only nominal outside the phrase, so H4″ makes her the patient, which matches “did he kiss Mary”. The place phrase comes before Mary, so it is adjoined at VP or higher. LF: ASK_WH(λx ∃e[kiss(e) ∧ Agent(e,he) ∧ Patient(e,Mary) ∧ at(e, home-of(x))]). (28a): John and háísh are two nominals with a direct verb, so John = agent and háísh = patient (H3), as printed. Here the patient comes before the place phrase. If the place phrase is adjoined to VP as in (26b), the patient must have moved left past it (H38a); otherwise the phrase attached lower (H38b). LF: ASK_WH(λx ∃e[kiss(e) ∧ Agent(e,John) ∧ Patient(e,x) ∧ at(e, home-of(y))]). Result, and a sharp prediction: H3 + H4″ predict that each string has a second parse. (26b) should also mean “Who kissed Mary at his house?” (háísh = agent, H3). (28a) should also mean “At whose house did he kiss John?” (háísh = possessor, John the lone nominal = patient, H4″). Willie prints one reading for each. An unprinted reading is not a star, so this is OPEN. If a speaker rejects these readings, H4″ or H3 needs a condition (perhaps that a nominal directly before a place phrase is read as its possessor). H38a versus H38b is also OPEN (the order “John place-phrase háísh verb” is not printed).
NX-381–383 (28a), (28b), (29): the possessor prefix is not voice. The verb is yiztsʼǫs (direct) in all three, so H3 has not moved anything, and the clause is the same. Only the possessor changes: ba- “his”, yi- “her”, ha- “that other person’s”. Under H37: yi- needs a possessor distinct from the third-person subject John, which fits “her” (the translations treat John as “he”). ba- allows John, which fits “his”. ha- is fourth person, disjoint from third-person John by H32a, which fits “that other person”. LF (28b): ASK_WH(λx ∃e[kiss(e) ∧ Agent(e,John) ∧ Patient(e,x) ∧ at(e, home-of(y))]), y ≠ John. LF (29): the same with y = z₄, z₄ ≠ John. Result: H37 fits all three, but the evidence is weak, because English “his”/“her” do not say who is meant. The possessor is not a clause argument, so H3-loc predicts that a lexical place phrase cannot take part in the inverse, unlike the grammatical postposition in NX-078–081. That matches Willie’s statement and the unchanged verb. SOURCED Speakers vary in using yi-/bi- this way, so H37 is limited to the usage reported.
NX-376–378 (21), (22), (25): a clause before nínízin “you think”. First clause: (21) háísh and łį́į́ʼ are two nominals with a direct verb, so háísh = agent and łį́į́ʼ = patient (H3), matching “Who will rope the horse”. In (25), John = agent and háísh = patient, the same as review 3’s NX-067. (22) shows that the frame is not specific to questions. H40b: merge the whole first CP as the complement of V “think”, which gives it the Theme role; V is final, so the CP comes first. A silent second person in Spec,vP is the Experiencer. Matrix C⁰ has Q and Agrees with the wh-indefinite. LF (25): ASK_WH(λx ∃s[think(s) ∧ Exp(s,you) ∧ Theme(s, ∃e[kiss(e) ∧ Agent(e,John) ∧ Patient(e,x)])]). This is exactly the paraphrase “Who do you think John kissed?” H40a: the first CP has its own C⁰ with Q. “Think” gives its Theme role to a silent pronoun, resolved to the first clause, and that CP is adjoined to the right. LF (25): ASK_WH(λx ∃e[kiss(e) ∧ Agent(e,John) ∧ Patient(e,x)]), plus the adjoined ∃s[think(s) ∧ Exp(s,you) ∧ Theme(s,q)], with q the proposition asked about. Here the paraphrase has to come from the speech act (asking the addressee’s opinion), not from Q c-commanding “think”. Result: both derive (21), (22) and (25). H40b derives the paraphrase more directly; H40a uses a silent Theme pronoun. See the tree below.
NX-384 (31) and NX-385 (32): “when” with “say”. This is the contrast that decides. Willie reports that (31) has only the “when did he leave” reading, and that asking about the time of saying takes (32). Under H40b, the embedded CP is inside the matrix VP. A matrix adjunct “when” + Q can be merged above the matrix vP; the matrix subject is silent, so “when” is pronounced first and the string is still hádááʼshąʼ dah diiyá ní. That gives the reading ASK_WH(λt ∃e1[say(e1) ∧ Agent(e1,she) ∧ Theme(e1,p) ∧ at(e1,t)]), p = ∃e2[leave(e2) ∧ Agent(e2,he)], which Willie says (31) does not have. So H40b needs H40c, a stipulation with no other job. Under H40a, “when” in the first clause can only modify “leave”, because the “say” clause is a separate CP. LF (31): ASK_WH(λt ∃e2[leave(e2) ∧ Agent(e2,he) ∧ at(e2,t)]), plus the adjoined ∃e1[say(e1) ∧ Agent(e1,she) ∧ Theme(e1,q)]. No extra assumption is needed. (32): merge V “say” with a pronoun object (glossed “3.3.said”, so the Theme is a third person, “it”) and a silent agent; “when” + Q modifies this clause: ASK_WH(λt ∃e1[say(e1) ∧ Agent(e1,she) ∧ Theme(e1,q) ∧ at(e1,t)]). The second clause is dashdiiyá with a fourth-person subject, closed by unlinked =hágíí, which names the fact (H34, out of sample). That fact is the Theme of “happened”: ∃e3[happen(e3) ∧ Theme(e3,p)], p = ιp[p = ∃e2[leave(e2) ∧ Agent(e2,z₄)]], with q resolved to p and z₄ ≠ she (H32a). Result: (31) against (32) favors H40a over H40b, unless H40b adds H40c. (32) shows overtly what H40a uses silently: a “say” verb whose object is a pronoun pointing to a separate clause. Clauses can be arguments when an enclitic names them as facts (H34 here and in review 10), so “Navajo has no argument clauses” is too strong; what is supported is that a bare clause before nínízin/ní is not one. OPEN Willie’s claim is prose, not a starred reading. Her translation of (31) ends in a period, so it may also report a question she asked (“she said, ‘When did he leave?’”), with Q inside what was said. The gloss shows no object on ní (“3.said”) or nínízin (“2.think”), unlike ʼání (“3.3.said”), so H40a’s silent pronoun has no overt mark there (NOT UNDERSTOOD).
Analysis A: H40b: the kiss clause is the Theme complement of “think”; Q is in the matrix C⁰ Hypothesis
Analysis B: H40a: two clauses; “you think (it)” is adjoined to the question clause; Q stays in the kiss clause Hypothesis
(1) A wh-word inside a clause before nínízin “you think”: John háísh yiztsʼǫs nínízin (NX-378) HypothesisNX-378, Willie 1991 ch. 8 (25), printed p. 214: John háísh yiztsʼǫs nínízin, glossed “who 3.3.kissed 2.think”, translated “John kissed who, you think?” with the paraphrase “(Who do you think John kissed?)”. Leaves are printed tokens. C⁰, T⁰, v⁰, the silent “you” and the silent “it” are our assignments.Bracket notation (copy)
DP (id you-high; landing site of Move from you): Experiencer: silent second person “you”
T′
vP
DP (id you; struck copy; moved to you-high): Experiencer: silent second person “you”
v′
VP
V′
CP (id kissCP)
C′
TP
DP (id john-high; landing site of Move from john): John
T′
vP
DP (id john; struck copy; moved to john-high): John
v′
VP
V′
DP (id who; goal of Agree Q (H11/H12): matrix C⁰ binds the wh-indefinite in situ): háísh
V⁰ (head; id V1; head-moves to v1): yiztsʼǫs
v⁰ (head; id v1; head-moves to T1)
T⁰ (head; id T1; hosts the “verb word” (V1 + v1 + T1))
C⁰ (head; id C1)
V⁰ (head; id V2; head-moves to v2): nínízin
v⁰ (head; id v2; head-moves to T2)
T⁰ (head; id T2; hosts the “verb word” (V2 + v2 + T2))
C⁰ (head; id C2; probe for Agree Q (H11/H12): matrix C⁰ binds the wh-indefinite in situ with who)
Analysis B
CP
CP (id kissCP)
C′
TP
DP (id john-high; landing site of Move from john): John
T′
vP
DP (id john; struck copy; moved to john-high): John
v′
VP
V′
DP (id who; goal of Agree Q (H11/H12): C⁰ of the kiss clause binds the wh-indefinite): háísh
V⁰ (head; id V1; head-moves to v1): yiztsʼǫs
v⁰ (head; id v1; head-moves to T1)
T⁰ (head; id T1; hosts the “verb word” (V1 + v1 + T1))
C⁰ (head; id C1; probe for Agree Q (H11/H12): C⁰ of the kiss clause binds the wh-indefinite with who)
CP
C′
TP
DP (id you-high; landing site of Move from you): Experiencer: silent second person “you”
T′
vP
DP (id you; struck copy; moved to you-high): Experiencer: silent second person “you”
v′
VP
V′
DP (id it): Theme: silent “it”, resolved to the kiss clause
V⁰ (head; id V2; head-moves to v2): nínízin
v⁰ (head; id v2; head-moves to T2)
T⁰ (head; id T2; hosts the “verb word” (V2 + v2 + T2))
C⁰ (head; id C2)
Does light d earn its keep?
HYPOTHESISFair test. Complex wh-phrases are a good place to look. (18) “which boy” has no enclitic, while (20) “which horse” has =ígíí (glossed DET) between the wh-word and =sh. A light-d analysis would give (18) a silent D⁰ that matches the overt =ígíí of (20). Result: light d is not needed. Both phrases are translated as presuppositional “which” questions, and no reading, word order or judgment on these pages differs between them, so the silent D⁰ would do nothing. The two examples also differ in noun and verb, so this is not a minimal pair, and nothing here counts against light d either. The only determiner-like elements that do work are overt (=ígíí, =hágíí). Review 10’s finding stands: an always-definite silent d is contradicted elsewhere (NX-372).
Telicity test
OPEN No pair on these pages differs only in telicity. A narrow candidate claim would be: “hádááʼshąʼ ‘when’ needs a telic predicate.” It occurs with “kiss” (NX-379), “leave” (NX-384) and “say” (NX-385). All three are bounded events by their meaning (our classification), and there is no atelic control. So the claim is neither supported nor contradicted. yidooloh “will rope” (NX-376/377) differs from the other verbs in mode, not in a controlled telicity contrast.
Negation
SOURCED No negation is on these pages. A text-layer search of printed pp. 212–218, checked against the renders, finds doo only inside yidooloh (NX-376/377). So there is nothing new on sentential versus constituent negation, and no negative cell changes.
Question × polarity × voice grid (cumulative after review 11)
HYPOTHESIS Only the wh-affirmative × direct cell gains examples. All new verbs are direct (yi-). The yi-/bi- possessor in (28a/b) is not voice: the verb is unchanged. No passive, inverse question or negation is on these pages. New sub-type: wh with a verb of thinking or saying (NX-376, 378, 384, 385).
OPEN Still empty: polar-affirmative × passive; polar-negative × inverse and × passive; wh-negative × inverse and × passive; any negative passive question. Places to search next: Willie ch. 2 and ch. 5 (passive and indefinite agents) for questions; the Navajo Questions database for bi- verbs with doo … da; Young & Morgan 1987 (not on the box); Perkins & Fernald grammar draft, content-question chapter, for a wh-word inside a clause before nízin or ní together with negation.
Where we part from the source’s analyses (labeled contrast)
HYPOTHESISKept, restated: “the clause before nínízin is not its complement” survives as H40a, which beats H40b on (31)/(32). The reason here is the missing “when did she say” reading and the overt pronoun object of ʼání, not Willie’s claim that nominals are adjuncts.
HYPOTHESISNot adopted: “Wh-words … are adjuncts that do not undergo movement from A-positions” (p. 219) as a general claim. Inside one clause, wh-words still take theta roles exactly where nouns do (H3/H4″, derivations 1–3), and review 3’s H14 fronting is still needed. “No complement clauses” is too strong: fact clauses with an enclitic are arguments (NX-385, review 10’s NX-372).
HYPOTHESISNot adopted: “obviation” as an unanalyzed label for (28). H37 states what is checkable (distinct from the subject or not, third or fourth person).
What this changes
HYPOTHESISH36, H37, H38a/b, H40a/b/c added. H40a is favored over H40b on (31)/(32); H40b survives only with the stipulation H40c.
HYPOTHESISH3, H4″, H3-loc, H32a and H34 pass out of sample (NX-374–385). H9′ extends to possessors as H37 (weak evidence).
Working default (V→v→T, subject to Spec,TP): every derivation uses it (the figure now draws it in both clauses), and nothing here resists it. What else has to move: the patient before the place phrase in (28a) takes one extra step (H38a or H38b), and “when” before the subject in (26a) is in Spec,CP or adjoined to TP (OPEN which). With no negation here, there is no verb-movement test of the default.
HYPOTHESIS Light d is not needed. Telicity is OPEN. The grid gains wh-affirmative × direct examples only.
OPEN predictions
OPENSecond parses: can (26b) mean “Who kissed Mary at his house?” and can (28a) mean “At whose house did he kiss John?” H3 + H4″ say yes.
OPENH40a versus H40b + H40c: a string with a matrix adverb placed between the embedded clause and ní (for example “when” right before ní), with its readings; and Willie’s pronominal-agreement evidence for “inner speech”, which is not printed in this chapter.
OPENH36: a noun before háísh “who” in a one-role clause (for example shizhéʼé before háísh), and any starred extraction from a “which” phrase.
OPENH38a versus H38b: the order agent, place phrase, wh-patient, verb.
OPENGrid: still no polar passive question and no negative inverse or negative passive question.
2026-11-14 early morning · First-principles review 12: Willie on “he said …” sentences, with first, second, third and fourth person (ch. 5, §5.3)
2026-11-14: the “On the spine” summary at the top of this page supersedes this section’s framing. Its data, judgments and citations are unchanged.
Restated on the unified structure (2026-11-15). This section’s results now sit on the one structure for Navajo: unified structure, reported-speech row. Its restatement as derivations is in Results on the spine. Data, judgments and citations below are unchanged.
Source: MaryAnn Willie, Navajo Pronouns and Obviation (PhD dissertation, University of Arizona, 1991), Chapter 5, §5.3 “The fourth person in reported speech”, examples (21)–(30), printed pp. 143–147, with the section’s closing prose on p. 148 and the chapter summary on pp. 148–149 (PDF pp. 149–155; PDF page = printed page + 6). Every form was read from 200-dpi page renders and 400-dpi crops, with 800-dpi zooms for accents and the index subscripts; the text layer was only a finding aid (it garbles ł and the accents). New index entries are NX-386–394. Not indexed: (27a–d), an English summary with no Navajo. Why this span: review 11 favored H40a (the clause before ní “said” is a separate clause) over H40b (it is the complement of “say”), and Willie’s ch. 8 cites “the patterns of pronominal agreement” in these sentences as her evidence for “inner speech”. This review checks whether that evidence decides between them. The same examples were read on the earlier site page Same person or someone else? (2026-10-10) under the older framework; its results are not premises here. H-numbers continue from review 11 (H40c).
HYPOTHESISPlain-words verdict. Willie’s person facts are real and fully derivable, but they are about who a person form can point to, not about how the two clauses are put together, so they do not by themselves decide whether the “I will come” clause is a separate clause or the object of “said”. Two smaller points pull in opposite directions: these clauses never carry the enclitic that every other subordinate clause has (Willie, p. 149), which suits the separate-clause analysis (H40a), and a second person can be meant as the real hearer only when “say” itself carries a second-person object (30), which is easiest to state if the clause is inside the “say” clause (H40b). H40a stays ahead on the strength of review 11’s “when” contrast. A plain third person in the reported clause never means the sayer, because Navajo has a dedicated way to say that (a shifted “I”, or fourth person on both verbs). Light d is not needed. Telicity: OPEN (one verb only). No questions, negation or passives are on these pages, so the grid is unchanged.
Starting point
Assumed, by instruction: VPISH (each verb’s subject argument first merges in Spec,vP, where it gets its theta role; this frames the arguments and is not a word-order account), theta roles, and the working default for tensed clauses: V→v→T head movement, with the subject moving Spec,vP→Spec,TP (on to Spec,CP or C when needed). HYPOTHESIS Everything else is tested: H1–H40c from reviews 1–11 (here mainly H32a, H32c, H34, H40a, H40b, H40c), light d, the categories D⁰ and C⁰, Agree, and any context-shifting device. The live question for word order is what else moves. Every derivation ends in a logical form (LF) whose scope order matches c-command. Tense is left out of the LFs except for the future mode, written FUT where the source translations depend on it. Willie’s analysis (a fourth person “marks” co- and disjoint reference; “inner speech” clauses are loosely adjoined) appears only as labeled contrast.
The source’s claims, as its own
SOURCEDAim. “I will survey briefly the use of the fourth person to mark co- and disjoint reference in reported speech contexts” (p. 143).
SOURCEDSummary (27), p. 146. “a. First person in the complement clause and third person in the main clause can either signal coreference with the subject of the verb of speech, or reference to the speaker of the complex utterance; b. When both the main and complement verbs are in the third person, disjoint reference is obligatory; c. When both verbs are in the fourth person, coreference between the subjects of the main and complement clauses is signaled; d. Fourth person on the verb of speech can signal a quotative.” (27c) and (27d) “overlap”, which is “the source of the ambiguity” of (26).
SOURCEDSecond person. “The use of the second person in the complement clause adds a modal feature (obligation) to the reading” (p. 147).
SOURCEDAll persons. “First person in the complement verb can be used to mark coreference with the main clause subject; second person forms carry a modal implication; third person forms in both clauses mark disjoint reference; and fourth person forms can be used for quotatives, to mark coreference across clauses, and as a respect form -- where it necessarily marks disjoint reference with all other pronouns” (p. 148).
SOURCEDNo enclitic. “It is often the case that across languages that verbs of speech have a special syntax. It is important to note that in Navajo these are the only subordinate clause types that do not have an enclitic that serves as determiner or nominalizer” (p. 149).
Willie calls the first clause the “complement clause” and the “say” clause the “main clause” in this chapter; in ch. 8 (p. 213) she says such clauses “are not complement clauses, but loosely adjoined as “inner speech””. We keep her words in quotes and use “reported clause” and “say clause” for our own neutral names.
The checked data
SOURCED Exact forms, source word glosses and translations, with example numbers and printed pages. Every example is two words: a verb meaning “arrive / come / go / be there” (the source glosses it “arrive” throughout) followed by a verb of saying. Source glosses, printing notes and crops are at the linked index entries.
NX-386 (21), p. 143: deesháał ní “1.arrive 3.said”: (a) “He said "I will be there".” (b) “He said that I would be there.”
NX-387 (22), p. 143: doogááł ní “3.arrive 3.said”: “Heᵢ said heⱼ will be there.”
NX-388 (23), p. 144: deesháał jiní “1.arrive 3.said” [jiní glossed 3 here, 4 elsewhere]: (a) “That personᵢ said that heᵢ would go.” (b) *“They say I will go.” (c) *“That person says that I will go.”
NX-389 (24), p. 144: doogááł jiní “3.arrive 4.said”: (a) “That personᵢ said heⱼ will come.” (b) “It is said that he will come.”
NX-390 (25), p. 145: jidoogááł ní “4.arrive 3.said”: “He said that person will come.” (respect reading only)
NX-391 (26), p. 145: jidoogááł jiní “4.arrive 4.said”: (a) “Heᵢ said that heᵢ would come.” (b) “They say that person will come.”
NX-392 (28), p. 147: díínááł ní “2.arrive 3.said”: “He said that you should go.”
NX-393 (29), p. 147: díínááł jiní “2.arrive 4.said”: (a) “It is said that you should go.” (b) *“That person said you should go.”
NX-394 (30), p. 147: díínááł nijiní “2.arrive 2.4.said”: “That person said [of you] that you should go.”
SOURCEDWhat the stars are on. No Navajo string is starred. The stars in (23b), (23c) and (29b) exclude a reading of an otherwise acceptable string. The source glosses are object.subject for transitive verbs, so nijiní “2.4.said” has a second-person object and a fourth-person subject.
HYPOTHESISOur editorial glosses (not the source’s). Only the person and mode values the source gloss and translations already show: (21) 1SG.SBJ.arrive.FUT 3.SBJ.say; (25) 4.SBJ.arrive.FUT 3.SBJ.say; (30) 2SG.SBJ.arrive.FUT 2SG.OBJ-4.SBJ.say. FUT is our reading of the translations (“will”, “would”, “should”); we do not segment the stems or the d-/-ł parts.
Hypotheses tested, revised or added
Carried over: H40a (the reported clause is a complete clause of its own; the say clause is adjoined to it, and “say” gives its Theme role to a silent pronoun resolved to that clause), H40b (the reported clause is the Theme complement of “say”; V is final, so it comes first), H40c (an add-on to H40b: nothing of the say clause may be pronounced before the reported clause), H32a (a silent pronoun shares reference only with one of the same person, so a fourth and a third person are disjoint), H32b (two fourth-person pronouns in one sentence refer to the same person), H32c (Willie’s “parallel processing” as a stipulation: with all third persons in two transitive clauses, the silent pronouns pair up), H34 (an enclitic D⁰ on a clause makes it name a fact and lets it be an argument).
H41 (new) · context shift. A first or second person in the reported clause is evaluated either in the actual speech situation or in the reported one (speaker = the sayer). Two ways to build it: H41a (with H40a) the reported clause is a root clause, and its context is the one fixed by the say clause it is linked to, as in an English quotation followed by “he said”; H41b (with H40b) an optional context-shifting operator in the embedded C⁰, inside the scope of “say”.
H42 (new) · a dedicated form blocks the plain third person. When the reported clause’s participant is the sayer, Navajo has a dedicated way to say so: a shifted first person, or (with a fourth-person sayer) a fourth person in the reported clause. Where a dedicated form exists, a plain third person is read as someone else.
H43 (new) · two fourth persons on “say”. The fourth-person subject of jiní is either a specific respected person (written z₄) or nobody in particular (a quotative, written as an unspecified sayer, “it is said”). This is Willie’s respect/quotative split restated as two values of one silent subject in Spec,vP.
H44 (new, a stipulation) · a respected sayer shifts first and second person. With a specific fourth-person sayer, a first or second person in the reported clause must be shifted (refer to the sayer or the sayer’s addressee), unless that participant is also an argument of “say”, as the second-person object of nijiní is in (30).
H40d (new, an add-on to H40b). A bare CP with no enclitic can be an argument only of “say” and “think”.
Derivations, step by step
HYPOTHESIS Each derivation gives Merge with theta assignment, Agree, Move, which copy is pronounced, what is inside or outside the verb word, and the LF. Extra assumptions beyond VPISH, theta roles and the working default are named, and all are HYPOTHESIS. Specifiers are on the left and heads are final. The role of the subject of “arrive/go” is written Agent, as for “leave” in review 11 (the label is ours). Every clause below is tensed, so each uses the working default: V→v→T, with its silent subject raised from Spec,vP to Spec,TP. The person and future-mode prefixes are spelled out inside the verb word that ends up in T⁰; whether a person prefix is Agree on T⁰ or the pronoun itself is left OPEN, since nothing on these pages separates the two. Because every subject is silent and T⁰ is final, nothing else moves: these strings put no extra demand on the default, and none of them resists it.
The shared clause-internal steps (all of NX-386–394). Reported clause: Merge V “arrive” (its single argument is the goer); merge v; merge a silent pronoun in Spec,vP, which gets the Agent role (VPISH), with person 1, 2, 3 or 4 as the verb shows. V→v→T; the pronoun moves Spec,vP→Spec,TP (working default); the lower copy is silent. C⁰ is declarative. Say clause: Merge V “say”; merge its internal argument (a CP under H40b, a silent “it” under H40a, see 2); merge v; merge the silent sayer in Spec,vP (Agent). V→v→T; sayer Spec,vP→Spec,TP. Spell-out: two verb words, reported clause first. Under H40b the order comes from V being final: the CP complement precedes V. Under H40a the say CP is adjoined to the right of the reported CP (rightward adjunction is HYPOTHESIS; if adjunction can only be leftward, the reported CP must move to a higher left position, one extra Move). Neither order needs anything else to move.
NX-386 (21): “I” can be the sayer or the actual speaker. The reported clause has a first-person subject. Reading (a), quoted: “I” = the sayer. Reading (b), indirect: “I” = the actual speaker. Under H40b + H41b: an optional shifting operator in the embedded C⁰. LF (a): ∃e1[say(e1) ∧ Agent(e1,x₃) ∧ Theme(e1, OP_shift(λc. FUT ∃e2[arrive(e2) ∧ Agent(e2, author(c))]))], with c set to the context of e1, so author(c) = x₃. LF (b): the same without OP_shift, so the Agent of e2 = the actual speaker. Under H40a + H41a: the reported CP is a root clause; its context is either the actual one (b) or the one of the saying event named by the adjoined say clause (a). LF (a): p = FUT ∃e2[arrive(e2) ∧ Agent(e2, author(c_e1))] · adjoined: ∃e1[say(e1) ∧ Agent(e1,x₃) ∧ Theme(e1,p)], author(c_e1) = x₃. In both, the say verb’s scope over the content matches the structure: under H40b by c-command, under H40a only through the silent Theme pronoun (no c-command, so no scope interaction is predicted; review 11’s (31) bears that out). Result: both derive (21); each needs one context-shifting assumption (H41a or H41b).
NX-387 (22): two plain third persons are disjoint. Under H40b, the sayer x₃ in matrix Spec,TP c-commands the reported clause’s subject y₃. Nothing in clause-internal syntax forbids y₃ = x₃ (it is not in the same clause), so c-command alone predicts that “he said he would come (himself)” is possible, as in English. Under H40a there is no c-command at all, so coreference is again free. Both analyses need H42: “he said he himself would come” has a dedicated form, the shifted first person of (21a), so the plain third person picks someone else. LF: p = FUT ∃e2[arrive(e2) ∧ Agent(e2,y₃)], ∃e1[say(e1) ∧ Agent(e1,x₃) ∧ Theme(e1,p)], x₃ ≠ y₃ (under H40b p is inside the Theme by c-command; under H40a by the silent pronoun). Result: Willie’s (27b) follows from H42. It also suggests why Willie’s “parallel processing” cases go the other way, where two all-third-person transitive clauses share their participants (review 10, H32c; her conditional sentences, summarized on p. 148): in a clause that is not reported speech there is no shifted first person to compete, so nothing blocks coreference. This gives the reported-speech half of the contrast a reason; H32c itself is still a stipulation.
NX-389 (24) and NX-391 (26): fourth person on “say”, respect or quotative. Under H43, the Spec,vP of “say” holds either z₄ (a specific respected person) or an unspecified sayer. (24): reported subject y₃. On the respect reading y₃ ≠ z₄ (H32a), matching “That personᵢ said heⱼ will come”; on the quotative reading there is no specific sayer: ∃e1[say(e1) ∧ Agent(e1,UNSPEC) ∧ Theme(e1,p)], p = FUT ∃e2[arrive(e2) ∧ Agent(e2,y₃)]. (26): reported subject w₄. Respect reading: w₄ = z₄, matching (26a) “Heᵢ said that heᵢ would come”: two fourth persons can be the same person (as in review 10, H32b), and this is the dedicated coreference form of H42 for a fourth-person sayer. Quotative reading (26b): unspecified sayer, w₄ a respected third party. Result: H32a, H42 and H43 together derive (24) and (26) with no new syntax. The fourth person in the reported clause is not shifted (a respected person would not use the respect form for themself), so fourth person behaves like third person, not like “I” and “you”, under H41.
NX-390 (25): fourth in the reported clause, plain third on “say”. Reported subject w₄, sayer x₃: w₄ ≠ x₃ (H32a), and no quotative reading is possible, because ní has no fourth-person subject to be a quotative (H43). LF: ∃e1[say(e1) ∧ Agent(e1,x₃) ∧ Theme(e1,p)], p = FUT ∃e2[arrive(e2) ∧ Agent(e2,w₄)], w₄ ≠ x₃. Result: derived; matches “Only the respect reading appears”.
NX-388 (23): a respected sayer forces “I” to be shifted. Only (23a) is allowed: “I” = the respected sayer. (23b) quotative is excluded; Willie gives a pragmatic reason (a reportative about one’s own actions is inappropriate), which we keep as such, labeled. (23c) “I” = the actual speaker is excluded. Nothing so far derives (23c): H41 allows an unshifted first person after ní (21b), and z₄ ≠ the actual speaker would satisfy H32a. So we add H44 (a stipulation): with a specific fourth-person sayer, first and second persons in the reported clause are shifted. LF (23a): p = FUT ∃e2[arrive(e2) ∧ Agent(e2, author(c_e1))], ∃e1[say(e1) ∧ Agent(e1,z₄) ∧ Theme(e1,p)], author(c_e1) = z₄. NOT UNDERSTOOD: why a respect sayer should force the shift. Willie’s account (“reference to the speaker of the whole complex utterance is not present” with a fourth-person sayer) is a restatement, kept as labeled contrast.
NX-392–394 (28)–(30): second person. (28) ní: “you” = the actual hearer, with an obligation reading. (29) jiní: the quotative (a) is fine, the respect reading with “you” = the hearer (b) is excluded, which H44 derives (the unshifted second person is blocked with a specific fourth-person sayer). (30) nijiní: “say” has a second-person object. Merge it as a second internal argument of “say” in Spec,VP (left), with an “about / for” role (our label for Willie’s bracketed “[of you]”); its person is spelled out on the say verb word. Under H44’s exception the hearer may then be meant in the reported clause. LF (30): p = SHOULD ∃e2[go(e2) ∧ Agent(e2,you)], ∃e1[say(e1) ∧ Agent(e1,z₄) ∧ About(e1,you) ∧ Theme(e1,p)]. Under H40b the exception is ordinary binding: the object “you” c-commands the CP complement and so can license the second person inside it. Under H40a the object of “say” does not c-command the reported clause, so the licensing has to go through the same anaphoric link as the silent Theme pronoun, and “say” now carries two internal arguments (the silent “it” and “you”). Result: (30) is the one fact on these pages that is easier to state under H40b. The obligation reading of a second person (SHOULD) is not derived by either analysis (OPEN); see the tree below.
Analysis A: H40b: the go clause is the Theme complement of “say”; the object “you” c-commands it Hypothesis
Analysis B: H40a: two clauses; “that person said (it) of you” is adjoined to the go clause Hypothesis
(1) A respected sayer with a second-person object: díínááł nijiní (NX-394) HypothesisNX-394, Willie 1991 ch. 5 (30), printed p. 147: díínááł nijiní, glossed “2.arrive 2.4.said”, translated “That person said [of you] that you should go.” Leaves are printed tokens. C⁰, T⁰, v⁰, the silent pronouns, the About role and the silent “it” are our assignments.Bracket notation (copy)
A: [CP [C′ [TP [DP Agent: silent fourth person z₄ (respected sayer)] [T′ [vP ⟨DP⟩ [v′ [VP [DP About: second person “you”, spelled ni- on the verb] [V′ [CP [C′ [TP [DP Agent: silent second person “you” (the goer)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [V⁰ díínááł]]] v⁰]] T⁰]] C⁰]] [V⁰ nijiní]]] v⁰]] T⁰]] C⁰]]
B: [CP [CP [C′ [TP [DP Agent: silent second person “you” (the goer)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [V⁰ díínááł]]] v⁰]] T⁰]] C⁰]] [CP [C′ [TP [DP Agent: silent fourth person z₄ (respected sayer)] [T′ [vP ⟨DP⟩ [v′ [VP [DP About: second person “you”, spelled ni- on the verb] [V′ [DP Theme: silent “it”, resolved to the go clause] [V⁰ nijiní]]] v⁰]] T⁰]] C⁰]]]
A: [CP [C′ [TP [DP Agent: silent fourth person z₄ (respected sayer)] [T′ [vP ⟨DP⟩ [v′ [VP [DP About: second person “you”, spelled ni- on the verb] [V′ [CP [C′ [TP [DP Agent: silent second person “you” (the goer)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [V⁰ díínááł]]] v⁰]] T⁰]] C⁰]] [V⁰ nijiní]]] v⁰]] T⁰]] C⁰]]
B: [CP [CP [C′ [TP [DP Agent: silent second person “you” (the goer)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [V⁰ díínááł]]] v⁰]] T⁰]] C⁰]] [CP [C′ [TP [DP Agent: silent fourth person z₄ (respected sayer)] [T′ [vP ⟨DP⟩ [v′ [VP [DP About: second person “you”, spelled ni- on the verb] [V′ [DP Theme: silent “it”, resolved to the go clause] [V⁰ nijiní]]] v⁰]] T⁰]] C⁰]]]
Structure as a list
Analysis A
CP
C′
TP
DP (id z4-high; landing site of Move from z4): Agent: silent fourth person z₄ (respected sayer)
T′
vP
DP (id z4; struck copy; moved to z4-high): Agent: silent fourth person z₄ (respected sayer)
v′
VP
DP (id you2; probe for Agree 2nd person licensed by a c-commanding argument of “say” (H44 exception as binding) with you1): About: second person “you”, spelled ni- on the verb
V′
CP (id goCP)
C′
TP
DP (id you1-high; landing site of Move from you1): Agent: silent second person “you” (the goer)
T′
vP
DP (id you1; struck copy; moved to you1-high; goal of Agree 2nd person licensed by a c-commanding argument of “say” (H44 exception as binding)): Agent: silent second person “you” (the goer)
v′
VP
V′
V⁰ (head; id V1; head-moves to v1): díínááł
v⁰ (head; id v1; head-moves to T1)
T⁰ (head; id T1; hosts the “verb word” (V1 + v1 + T1))
C⁰ (head; id C1)
V⁰ (head; id V2; head-moves to v2): nijiní
v⁰ (head; id v2; head-moves to T2)
T⁰ (head; id T2; hosts the “verb word” (V2 + v2 + T2))
C⁰ (head; id C2)
Analysis B
CP
CP (id goCP)
C′
TP
DP (id you1-high; landing site of Move from you1): Agent: silent second person “you” (the goer)
T′
vP
DP (id you1; struck copy; moved to you1-high): Agent: silent second person “you” (the goer)
v′
VP
V′
V⁰ (head; id V1; head-moves to v1): díínááł
v⁰ (head; id v1; head-moves to T1)
T⁰ (head; id T1; hosts the “verb word” (V1 + v1 + T1))
C⁰ (head; id C1)
CP
C′
TP
DP (id z4-high; landing site of Move from z4): Agent: silent fourth person z₄ (respected sayer)
T′
vP
DP (id z4; struck copy; moved to z4-high): Agent: silent fourth person z₄ (respected sayer)
v′
VP
DP (id you2): About: second person “you”, spelled ni- on the verb
V′
DP (id it): Theme: silent “it”, resolved to the go clause
V⁰ (head; id V2; head-moves to v2): nijiní
v⁰ (head; id v2; head-moves to T2)
T⁰ (head; id T2; hosts the “verb word” (V2 + v2 + T2))
C⁰ (head; id C2)
H40a or H40b: what these pages add
HYPOTHESISNeutral: the person and reference facts of (21)–(29) (Willie’s “pronominal agreement”). Each analysis derives them with the same added hypotheses H41 (one version each), H42, H43 and H44. They concern who a pronoun can refer to, which needs no particular constituent structure; so they are not, by themselves, evidence that the clause is “not a complement”.
HYPOTHESISFor H40a: the missing enclitic (p. 149). Elsewhere a clause that is an argument carries an enclitic (=hágíí in review 11’s NX-385, =ígíí in review 10, H34). H40a expects no enclitic on the reported clause, because it is not an argument. H40b needs H40d, an exception for “say” and “think” only.
HYPOTHESISFor H40b: (30). The real hearer can be meant in the reported clause after a respected sayer only when “say” has a second-person object, which H40b states as binding under c-command and H40a states only through an anaphoric link.
HYPOTHESISUnchanged: review 11’s (31)/(32) contrast, where H40b needs the stipulation H40c. With one point and one stipulation on each side here, H40a stays favored, but these pages do not settle it.
Does light d earn its keep?
HYPOTHESISFair test. There are no overt nouns on these pages, so the only place a light d could work is on the reported clause itself: a silent D⁰ over it, matching the overt =hágíí of review 11’s fact clause, would make it an argument and so support H40b. Result: light d is not needed, and a silent d with =hágíí’s meaning would do harm: under H34 that would make the reported content a fact, but the reported content here is not treated as true. The quotative (24b), (29a) and Willie’s note on (30), that the hearer “has not necessarily been informed” (p. 147), report content without taking it as fact (our reading of her translations). A silent d with no meaning would do no work. The only determiner-like enclitics that do work remain overt.
Telicity test
OPEN Every reported clause uses the same verb, glossed “arrive” and translated “be there”, “go” or “come”, and every saying verb is “say”. With one predicate there is no telic/atelic contrast, so no claim can be tested. A candidate for later: “the obligation reading of a second person (28) needs a bounded event.” It needs the same frame with a state or activity verb.
Negation
SOURCED No negation is on these pages. A text-layer search of printed pp. 139–150, checked against the renders, finds doo only inside doogááł and jidoogááł (NX-387, 389, 390, 391), and no da, hanii or question marker. Nothing new on sentential versus constituent negation.
Question × polarity × voice grid (cumulative after review 12)
HYPOTHESIS No cell changes. These pages have no question and no negation. The quotative jiní “it is said” (NX-389, 391, 393) has an unspecified sayer, which is relevant to the impersonal column, but it is a declarative, and an English passive translation does not fill a Navajo passive cell.
OPEN Still empty: polar-affirmative × passive; polar-negative × inverse and × passive; wh-negative × inverse and × passive; any negative passive question. Places to search next: Willie ch. 2 and ch. 5 §§5.0–5.2 (passive and indefinite agents, conditionals) for questions; a question inside a reported clause before jiní (quotative + Q); the Navajo Questions database for bi- verbs with doo … da.
Where we part from the source’s analyses (labeled contrast)
HYPOTHESISKept, restated: Willie’s (27a)–(27d) are all derived: (27a) by H41, (27b) by H42, (27c) by H32b (two fourth persons are the same person), (27d) by H43.
HYPOTHESISNot adopted: “the fourth person … mark[s] co- and disjoint reference” as a primitive. On our account the fourth person is a person value (H32a/H32b); disjointness of two plain third persons comes from competition with the dedicated forms (H42), not from a marking rule.
HYPOTHESISNot adopted as evidence: the person facts as proof of “inner speech” (ch. 8, p. 213). They are equally derivable with a complement clause. The evidence that does bear on the structure is the missing enclitic (p. 149) and review 11’s “when” contrast.
HYPOTHESISNot adopted: Willie’s explanation of (23c). It restates the fact; H44 is our equally stipulative description, marked NOT UNDERSTOOD.
What this changes
HYPOTHESISH41a/b, H42, H43, H44 and H40d added. H42 derives Willie’s (27b) and suggests why her parallel-processing cases (H32c) differ (no competing form there). H44 is a stipulation.
HYPOTHESISH40a vs H40b: H40a stays favored (review 11), but the person facts are neutral; the missing enclitic counts for H40a, (30) counts for H40b.
HYPOTHESISH32a, H32b, H34 pass out of sample (NX-388–391, light-d test).
Working default (V→v→T, subject to Spec,TP): used in all eighteen clauses; nothing else has to move, and nothing here resists it. Mode (FUT) is spelled out inside the verb word in T⁰. The only order question is how the say clause comes second: a final V (H40b) or right adjunction (H40a; a leftward-only adjunction theory would need one extra Move of the reported clause).
HYPOTHESIS Light d is not needed. Telicity is OPEN. No grid change.
OPEN predictions
OPENH40a vs H40b: a sourced reported clause that carries an enclitic (=go, =ígíí) before ní, with its readings; and a matrix adverb between the reported clause and ní (review 11).
OPENH41: does a shifted first person ever share a reported clause with an unshifted second person? H41b (an operator in C⁰) shifts the whole clause at once; H41a allows the same. A mixed case would count against both as stated.
OPENH42: H42 predicts that a plain third person after ní stays disjoint even with a noun (“John” before the reported verb); and that a non-reported two-clause sentence with the same verbs allows coreference again.
OPENH44: can (29) have a shifted second person (the respected sayer’s own addressee)? H44 says yes. Is nijiní also possible with a shifted reading?
OPENSecond-person obligation (28): is the obligation reading also found in a main clause with the same second-person future verb? If so, it is a root property and fits H40a; if only under “say”, it fits H40b.
OPENGrid: a question inside a reported clause (with ní vs quotative jiní), and still no polar passive question and no negative inverse or negative passive question.
2026-11-14 afternoon · First-principles review 13: Willie on “when / as” clauses with =go, who pairs with whom across two clauses, and event vs person focus (ch. 5, §§5.0–5.2)
2026-11-14: the “On the spine” summary at the top of this page supersedes this section’s framing. Its data, judgments and citations are unchanged.
Restated on the unified structure (2026-11-15). This section’s results now sit on the one structure for Navajo: unified structure, =go row. Its restatement as derivations is in Results on the spine. Data, judgments and citations below are unchanged.
Source: MaryAnn Willie, Navajo Pronouns and Obviation (PhD dissertation, University of Arizona, 1991), Chapter 5 “The Fourth Person in Complex Sentences”, §5.0 introduction, §5.1 “Adjoined Temporal Clauses” (§§5.1.1–5.1.3) and §5.2 “Event vs. Person Focus”, examples (2)–(20), printed pp. 132–142 (PDF pp. 138–148; PDF page = printed page + 6). Every form was read from 200-dpi page renders, with 400-dpi renders and 800-dpi zooms for accents, nasal hooks and index subscripts; the text layer was only a finding aid (it garbles ł, the accents and the subscripts). New index entries are NX-395–414. Already indexed and cited, not re-indexed: the first (3), p. 133 = NX-110. Not indexed: (1) (English only), and (6) and (15) (the source’s principles, no Navajo). Why this span: review 12 proposed H42 (a dedicated coreference form makes a plain third person disjoint) and suggested it explains why Willie’s “parallel processing” sentences go the other way (two all-third-person transitive clauses share their participants; review 10, H32c). These pages hold her parallel-processing data for =go clauses, so they test that suggestion directly. H-numbers continue from review 12 (H44, H40d).
HYPOTHESISPlain-words verdict. In two “kicked him … hit him” clauses joined by =go, Navajo makes the second clause reuse the first clause’s people: the one hit must be someone already in the first clause (unless it is “a girl” with léiʼ), and the hitter is the kicker whenever the person forms allow it. One rule with those two parts (H45) predicts every good and every starred sentence on these pages, including the fourth-person ones. It replaces the old “parallel processing” stipulation, though it is still a stipulation at heart. Review 12’s idea that “nothing competes, so the two clauses can corefer” explains only half of the pattern: it says why coreference is possible, not why it is forced, and a general “the special form wins” rule is contradicted by (2). One example pair, (9a)/(9b), has printed index letters that disagree with Willie’s own prose about it, so whether the yi-/bi- switch can swap who does what stays OPEN. The “event vs person” contrast of (17)–(20) follows from person alone: an event can only be referred to by a third person, and a fourth person only by a person. Light d is not needed. Telicity: the =go form does not change with telicity; the event/person contrast holds with the verbs, and so their telicity, kept the same. No questions, negation or passives, so the grid is unchanged.
Starting point
Assumed, by instruction: VPISH (each verb’s subject argument first merges in Spec,vP, where it gets its theta role; this frames the arguments and is not a word-order account), theta roles, and the working default for tensed clauses: V→v→T head movement, with the subject moving Spec,vP→Spec,TP (on to Spec,CP or C when needed). HYPOTHESIS Everything else is tested: H1–H44 and H40d from reviews 1–12 (here mainly H3, H6, H4″, H31, H32a–c, H34, H42), light d, the categories D⁰ and C⁰, Agree, binding conditions, and any topic position. The live question for word order is what else moves. Every derivation ends in a logical form (LF) whose scope order matches c-command. Tense is left out of the LFs. The meaning of =go is written WHEN(p, q) (“when / as / if p, q”), its first argument being the =go clause. Willie’s analysis (“parallel processing” as a principle of interpretation, nominals as adjuncts, “focus” on a person or an event) appears only as labeled contrast.
The source’s claims, as its own
SOURCEDAim. The fourth person’s function is “that of marking disjoint reference across clauses in complex sentences” (p. 132); §5.1 treats “the temporal/conditional constructions where an initial subordinate clause is under the scope of the complementizer enclitic -go” (p. 132).
SOURCED(6) Parallel Processing, p. 135. “In a sequence of two transitive clauses with all third person arguments, the two arguments of the first and second clause are interpreted as coreferential.” With =go, “coreference is obligatory unless disjoint reference is overtly marked” (p. 135).
SOURCEDRoles. “The yi/bi- alternation can be used to mark a change in thematic roles across clauses” (p. 136); “each clause must have the same two referents, although their status as agent or patient may change, as we have seen in (9a,b)” (p. 136).
SOURCED(15), p. 139. “a. If arguments are all third person, coreference of the arguments across clauses is obligatory; the yi-/bi- alternation permits a switch in thematic roles. b. If the arguments in the first clause are third person, fourth person cannot appear in the second clause. c. A fourth person argument may be used in the first clause to mark disjoint reference of the agents across the two clauses -- but not the patients.” Also: “Navajo requires that in conditional constructions, transitive clause sequences have coreferent patients” (p. 139).
SOURCEDException. “It is possible to create disjoint reference of patient arguments across clauses in a temporal construction if the patient of the second clause is less definite than that of the first” (pp. 139–140).
SOURCED§5.2. The fourth person “excludes the adjunction of a simple nominal”, but in (17)–(20) there is “focus on a person rather than the event the person participates in” (p. 141); in (18) “the nominal Mary cannot be syntactically associated with the subject of the subordinate clause … But this nominal can be a topic for the whole complex sentence” (pp. 141–142); “in (20), the fourth person patient pronoun in the main verb can refer only to the person and never to an event” (p. 142).
The checked data
SOURCED Exact forms, source glosses and translations, with example numbers and printed pages. Index subscripts are as printed. Source glosses, printing notes and crops are at the linked index entries.
Intransitive pairs (pp. 133–134).NX-395 (2): yah ʼííyáa=go dah neezdá “3.came in=when 3.sat down”, “When heᵢ came in, heᵢ,ⱼ sat down.” · NX-396 (second (3)): ʼashkii yah ʼííyáa=go dah neezdá “boy 3.came in=when 3.sat down”, “When the boyᵢ came in, heᵢ,ⱼ sat down.” · NX-397 (4): yah ʼííyáa=go ʼatʼééd dah neezdá, “When he came in, the girl sat down.” (disjoint, p. 134) · NX-398 (5): yah ʼííyáa=go dashneezdá “3-came in=when 4-sat down”, “When heᵢ came in, heⱼ sat down.” · Cited: NX-110 (first (3)): díníyáa=go shídóʼ dooleeł, “If you go, I will go too.”
Transitive, all third person (pp. 135–136).NX-399 (7): yiztał=go néidííłtsʼin, “As heᵢ kicked himⱼ, heᵢ hit himⱼ.” (source gloss: 3.3.kicked=COMP 3.3.fist-hit; the hyphen in “fist-hit” is the gloss’s, not a break in the Navajo word) · NX-400 (8): biztał=go nábiidííłtsʼin, “As heᵢ was kicked by himⱼ, heᵢ was hit by himⱼ.” (source gloss: INV.3.3.kicked=COMP INV3.3.fist-hit) · NX-401 (9a): yiztał=go nábidííłtsʼin, “As heᵢ kicked himⱼ, heⱼ was hit by himᵢ.” · NX-402 (9b): biztał=go néidííłtsʼin, “As heᵢ was kicked by himⱼ, heⱼ hit himᵢ (back).”
Fourth person (pp. 137–138).NX-403 *(10): yiztał=go náhodííłtsʼin · NX-404 (11a): jiztał=go náhodííłtsʼin, “As that personᵢ kicked himⱼ, heⱼ hit that personᵢ.” · NX-405 (11b): haztał=go nízhdííłtsʼin, “As heᵢ kicked that personⱼ, that personⱼ hit himᵢ.” · NX-407 (13): jizloh=go ʼáádóó yiztał “3.4.roped=COMP CONJ then 3.3.kicked”, “That personᵢ roped himⱼ, and then heₖ kicked himⱼ.” · NX-408 *(14): hazloh=go ʼáádóó yiztał “4.3.roped=COMP CONJ 3.3.kicked”.
A noun in the second clause (pp. 137–140).NX-406 *(12): yiztał=go ʼatʼééd néidííłtsʼin · NX-409 *(16a): yiztał=go ʼáádóó ʼatʼééd néidííłtsin · NX-410 (16b): yiztał=go ʼáádóó ʼatʼééd léiʼ néidííłtsin (“indf.” under léiʼ), “Heᵢ kicked himⱼ and then heᵢ hit a girl.”
Event vs person (pp. 141–142).NX-411 (17): Mary yicha=go yiiłtsą́ “3.crying=while 3.1.saw”, “I saw Mary crying.” · NX-412 (18): Mary jicha=go hoołtsą́ “4.crying=while 4.1.saw”, “I saw Mary when she was crying.” “(Mary, I saw her when she was crying.)” · NX-413 (19): Bill łį́į́ʼ booyeéł=go yiiłtsą́ “horse 3.3.carrying=while 3.1.saw”, “I saw Bill riding the horse.” · NX-414 (20): Bill łį́į́ʼ hooyeéł=go hoołtsą́ “horse 4.3.carrying=while 4.1.saw”, “I saw Bill when he was riding the horse.”
SOURCEDWhat the stars are on. (10), (12), (14) and (16a) are starred whole strings with no translation; the prose says which reading they would need. The source glosses give object before subject for transitive verbs: jizloh “3.4.roped” is translated “That person roped him” (13), so the fourth person is the roper.
HYPOTHESISOur editorial glosses (not the source’s). Agent>patient, from the source glosses and translations only, stems not segmented: NX-399 kick.3>3=when fist.hit.3>3; NX-407 rope.4>3=when then kick.3>3; NX-408 rope.3>4=when then kick.3>3; NX-412 Mary cry.4=while see.1>4. For (10), (11a), (11b), (12) and (16) we read the person values off the matching glossed forms in (7), (11) and (13) and the translations; we do not segment náhodííłtsʼin or nízhdííłtsʼin.
Hypotheses tested, revised or added
Carried over: H3 (inverse: the patient moves above the agent), H6 (that moved patient is definite), H4″ (a single nominal goes with the lowest third-person participant), H31 (a noun is linked by Agree only to a third-person participant), H32a (a silent pronoun shares reference only with one of the same person), H32b (two fourth-person pronouns in one sentence refer to the same person), H32c (Willie’s parallel processing, as a stipulation), H34 (an enclitic D⁰ on a clause can make it name a fact), H42 (a dedicated coreference form makes a plain third person disjoint).
H45 (new; replaces H32c) · the second clause reuses the first clause’s participants. In two transitive clauses joined by =go: H45a the second clause’s patient, if definite, must be one of the first clause’s participants (of the same person, H32a); an indefinite patient (léiʼ) may be new. H45b the second clause’s agent is the first clause’s agent whenever their persons match; if they do not match, the agent is free. H45c a plain (non-reflexive) transitive verb cannot have the same individual as agent and patient. H45b is stated here by theta role (H45b-θ); a rival states it by position after inverse Move (H45b-pos: the higher argument pairs with the higher argument). They differ only on (9a)/(9b).
H46 (new) · a noun where a pronoun would do names someone else. An overt noun in the second clause, where a silent pronoun could refer to a participant already in the first clause, refers to a different individual. This restates Willie’s “nominal marks disjoint reference” (review 10, p. 167) as a competition between a noun and a silent pronoun.
H47 (new) · where the =go clause attaches. The =go clause is a CP adjoined at the left of the main clause at the TP level or higher, above the main clause’s Spec,TP. Its C⁰ is =go (Willie’s “complementizer enclitic”).
H48 (new) · events are third person. A pronoun that refers to an event or a proposition is third person; the fourth person refers only to persons. Two ways (17) can get its event reading: H48a the =go clause is itself the Theme argument of “see”; H48b it is the same adjoined time clause as everywhere else (H47), and the silent third-person object of “see” is resolved to its event (the device of H40a, used for “say” in reviews 11–12).
H49 (new) · a sentence topic. A noun at the left edge of the whole sentence, in Spec,CP, with no theta role, can be about a person who is referred to inside by fourth-person pronouns. It is not linked by Agree, so H31 does not apply to it. That a nominal can lack a theta role is itself HYPOTHESIS, beyond the theta-role axiom.
Derivations, step by step
HYPOTHESIS Each derivation gives Merge with theta assignment, Agree, Move, which copy is pronounced, what is inside or outside the verb word, and the LF. Extra assumptions beyond VPISH, theta roles and the working default are named, and all are HYPOTHESIS. Specifiers are on the left and heads are final. The role labels Agent and Patient for “kick”, “fist-hit” and “rope”, Agent for “come in”, “sit down” and “cry”, and Experiencer and Theme for “see” are ours. Silent pronouns are written with their person: x₃, y₃ (third), z₄, w₄ (fourth).
The shared clause-internal steps (all of NX-395–414). Each clause is tensed and uses the working default. Transitive clause: Merge V with a silent pronoun as its complement, which gets the Patient role; merge v; merge a silent pronoun in Spec,vP, which gets the Agent role (VPISH). V→v→T; the agent moves Spec,vP→Spec,TP; its lower copy is silent. In a bi- (inverse) clause, the patient also moves above the agent (H3); it is a silent pronoun and so definite, as H6 requires. Person and voice prefixes are spelled out inside the verb word in T⁰; whether a person prefix spells out Agree on T⁰ or the pronoun itself is left OPEN. The first clause’s C⁰ is =go, pronounced as an enclitic on the verb word; that CP is adjoined to the left of the main TP (H47). Spell-out: two verb words, the =go clause first, so with silent arguments nothing else moves. LF shape: WHEN(∃e1[…], ∃e2[…]); the =go clause c-commands the main TP, matching its place as WHEN’s first argument.
Intransitive pairs: NX-395–398 (2)–(5), and where =go attaches. (2): two one-argument clauses, subjects x₃ and x′₃. Nothing requires x′₃ = x₃; H45 is about two transitive clauses and a second-clause patient, and there is none. LF: WHEN(∃e1[come.in(e1) ∧ Agent(e1,x₃)], ∃e2[sit.down(e2) ∧ Agent(e2,x′₃)]), x′₃ = x₃ preferred, ≠ allowed (“heᵢ,ⱼ”). (Second (3)): ʼashkii precedes =go, so it is inside the first clause, linked to its only participant (H4″, H31); it raises to Spec,TP there with the default. (4): ʼatʼééd comes after =go, so it is in the main clause; it is the subject, which merges in Spec,vP and raises to the main Spec,TP (default). Since the =go clause is pronounced before it, the =go clause must sit above Spec,TP: this is the evidence for H47. A no-raising rival with the =go clause lower would need the subject to stay in Spec,vP; the default plus H47 needs nothing else to move. Disjointness (“When he came in, the girl sat down”): binding does not give it, because the first clause’s silent subject is inside an adjunct and c-commands nothing in the main clause; H46 does (the noun is used where a silent pronoun would have picked up x₃). LF: WHEN(∃e1[come.in(e1) ∧ Agent(e1,x₃)], ∃e2[sit.down(e2) ∧ Agent(e2,ιg.girl(g))]), ιg.girl(g) ≠ x₃. (5): the second subject is w₄, ≠ x₃ by H32a. Result: derived; the only new structure is H47, forced by the order in (4) under the default.
All third person: NX-399 (7) and NX-400 (8). (7): clause 1 Agent x₃, Patient y₃; clause 2 Agent x′₃, Patient y′₃. H45b: the persons match, so x′₃ = x₃. H45a: y′₃ ∈ {x₃, y₃}; H45c rules out y′₃ = x′₃ = x₃; so y′₃ = y₃. LF: WHEN(∃e1[kick(e1) ∧ Agent(e1,x₃) ∧ Patient(e1,y₃)], ∃e2[fist.hit(e2) ∧ Agent(e2,x₃) ∧ Patient(e2,y₃)]). (8): both clauses inverse; each patient moves above its agent (H3); H45 pairs by role, so the result is the same with the kicker and hitter being j, matching “heᵢ was kicked by himⱼ, heᵢ was hit by himⱼ”. Result: H45 derives (7) and (8), and H6 passes out of sample (silent definite inverse patients). These two examples do not separate H45b-θ from H45b-pos, since the voice is the same in both clauses.
Mixed voice: NX-401 (9a) and NX-402 (9b), a source conflict.OPEN By the printed subscripts, (9a) “As heᵢ kicked himⱼ, heⱼ was hit by himᵢ” keeps i as kicker and hitter, and (9b) “As heᵢ was kicked by himⱼ, heⱼ hit himᵢ” keeps j as kicker and hitter. That is exactly what H45b-θ predicts: the bi- clause moves the shared patient above the shared agent (H3), changing which participant comes first and which one the English makes the subject, but not who does what. LF (9a), printed: WHEN(∃e1[kick(e1) ∧ Agent(e1,x₃) ∧ Patient(e1,y₃)], ∃e2[fist.hit(e2) ∧ Agent(e2,x₃) ∧ Patient(e2,y₃)]), the same as (7). Willie’s prose says the opposite: the alternation marks “a change in thematic roles” and the participants’ “status as agent or patient may change, as we have seen in (9a,b)” (p. 136), and (9b) ends in “(back)”, which fits only a role swap (“he was kicked by him, he hit him back”). On that reading, the higher argument after inverse Move pairs with the higher argument (H45b-pos): LF (9b), prose reading: WHEN(∃e1[kick(e1) ∧ Agent(e1,y₃) ∧ Patient(e1,x₃)], ∃e2[fist.hit(e2) ∧ Agent(e2,x₃) ∧ Patient(e2,y₃)]). Result: the printed subscripts support H45b-θ (pairing computed on first-merge theta positions); the prose supports H45b-pos (pairing computed after inverse Move, which would make inverse Move feed interpretation). The source does not settle it, and we do not choose by guessing which part is a slip. Both versions agree on every other example on these pages.
Fourth person: NX-403 *(10), NX-404/405 (11a/b), NX-407 (13), NX-408 *(14). *(10): clause 1 has x₃, y₃; the second clause’s patient is fourth person (ho-), and there is no fourth-person participant in clause 1, so H45a fails: starred, as printed. (11a): clause 1 Agent z₄ (ji-), Patient y₃; clause 2 patient w₄ must be a clause-1 fourth person (H45a), so w₄ = z₄ (also H32b); the clause-2 agent is third person, its person does not match z₄, so it is free (H45b), and the printed reading takes y₃. LF: WHEN(∃e1[kick(e1) ∧ Agent(e1,z₄) ∧ Patient(e1,y₃)], ∃e2[fist.hit(e2) ∧ Agent(e2,y₃) ∧ Patient(e2,z₄)]). (11b): clause 1 Agent x₃, Patient z₄ (ha-); clause 2 Agent w₄ (zh-) = z₄ (H32b), Patient third person, which must be x₃ (H45a). LF: WHEN(∃e1[kick(e1) ∧ Agent(e1,x₃) ∧ Patient(e1,z₄)], ∃e2[fist.hit(e2) ∧ Agent(e2,z₄) ∧ Patient(e2,x₃)]). (13): clause 1 Agent z₄, Patient y₃; clause 2 Patient third person, so = y₃ (the only third person in clause 1, H45a); clause 2 Agent third person, its person does not match z₄, so it is free, and H45c keeps it ≠ y₃: a new person k, exactly “heₖ kicked himⱼ”. LF: WHEN(∃e1[rope(e1) ∧ Agent(e1,z₄) ∧ Patient(e1,y₃)], ∃e2[kick(e2) ∧ Agent(e2,u₃) ∧ Patient(e2,y₃)]), u₃ ∉ {z₄, y₃}. *(14): clause 1 Agent x₃, Patient z₄; clause 2 Patient third person must be x₃ (H45a; z₄ does not match); clause 2 Agent third person matches x₃, so it is x₃ (H45b); H45c forbids agent = patient: no well-formed reading, starred, as printed. Result: H45 derives all five, including the contrast (13) vs *(14) that Willie states as (15c). H32c (all-third-person clauses only) says nothing about (10)–(14), so H45 is strictly broader. OPEN H45 predicts (11a) has a second reading with a new hitter (“…, someone else hit that person”), because its agent is free as in (13). Willie prints only “heⱼ”, and her p. 137 remark that the fourth person here “cannot introduce a new referent in the second clause” conflicts with her (13)/(15c), where a new agent is fine. Not settled by the source.
Nouns and definiteness: NX-406 *(12), NX-409 *(16a), NX-410 (16b). *(12): the second clause’s patient is the bare noun ʼatʼééd, read as definite (“the girl”, as in (4)). Under the default it stays in the VP (complement of V), and the silent agent raises to Spec,TP; the order agent (silent) – noun – verb needs nothing else to move. H45a requires the patient to be x₃ or y₃; H46 says a noun where a pronoun would have picked y₃ names someone else; the two conflict: starred, as printed. *(16a) is the same with ʼáádóó “and then”. (16b): ʼatʼééd léiʼ is indefinite (“a girl”), so H45a lets it be new; H45b gives the agent = x₃. LF: WHEN(∃e1[kick(e1) ∧ Agent(e1,x₃) ∧ Patient(e1,y₃)], ∃g[girl(g) ∧ ∃e2[fist.hit(e2) ∧ Agent(e2,x₃) ∧ Patient(e2,g)]]), with ∃g inside the main clause, below WHEN, matching its position. Result: the minimal pair *(16a)/(16b) is derived, and Willie’s “less definite” exception is the indefinite clause of H45a. Willie’s “Navajo requires … coreferent patients” (p. 139) is too strong as printed: her own (11a), (11b) (patients j then i) and (16b) are counterexamples; H45a (“a participant of clause 1”, not “the patient of clause 1”) covers them.
Does H42 explain the reverse pattern? Review 12 suggested that outside reported speech no dedicated coreference form competes, so “nothing blocks coreference”. Tested here: (i) coreference is available in every =go pair: (2), (second 3), (7)–(9). This part holds. (ii) But in the transitive pairs coreference is forced (7), and the disjoint versions are starred (*(10), *(12)). H42 only removes a block; it cannot make coreference obligatory. That needs H45. (iii) A general version of H42, “wherever a dedicated form for a reading exists, the plain form loses that reading”, would make (2) coreferent only, since the noun (4) and the fourth person (5) are dedicated disjoint forms. (2) prints “heᵢ,ⱼ” and Willie allows disjoint reference “in context” (p. 133), so the general version is CONTRADICTED by NX-395 with NX-397/398. Result: H42 survives only in its specific form (dedicated coreference forms, as in reported speech); its explanation of the reverse pattern is partial. The forced pairing in transitive =go sequences is H45, which is still a stipulation. NOT UNDERSTOOD: why only two transitive clauses force pairing, and why patients (H45a, strict) but agents only when their persons match (H45b). One observation: the strict half concerns the argument merged inside VP, the soft half the argument merged in Spec,vP. That is a VPISH distinction, and it may be a lead, but nothing here derives it.
Event vs person: NX-411–414 (17)–(20). (17): inside the =go clause, Mary is the crier, linked to its only participant (H4″, H31), raised to that clause’s Spec,TP (default). “See”: Experiencer = silent first person in Spec,vP, raised to Spec,TP; its Theme is third person (yiiłtsą́ “3.1.saw”). Under H48a the Theme is the =go CP itself, merged as the complement of V and preceding it because V is final; its third-person object marking is Agree with the clause (H48). LF: ∃e1[see(e1) ∧ Exp(e1,I) ∧ Theme(e1, ιe2[cry(e2) ∧ Agent(e2,Mary)])]. Under H48b the =go CP is the adjoined time clause of H47, and the silent object it₃ is resolved to its event: WHEN(∃e2[cry(e2) ∧ Agent(e2,Mary)], ∃e1[see(e1) ∧ Exp(e1,I) ∧ Theme(e1,it₃)]), it₃ = e2. (18): the crier is w₄ (ji-), and the Theme of “see” is z₄ (ho-); w₄ = z₄ (H32b). By H48 a fourth-person Theme cannot be the crying event, so the =go clause cannot be the Theme and is the time adjunct of H47: “when she was crying”. Mary cannot be linked to a fourth-person participant (H31), so she is a sentence topic in Spec,CP (H49), matching Willie’s “(Mary, I saw her when she was crying.)”. LF: TOP(Mary) : WHEN(∃e2[cry(e2) ∧ Agent(e2,w₄)], ∃e1[see(e1) ∧ Exp(e1,I) ∧ Theme(e1,z₄)]), w₄ = z₄ = Mary. (19)/(20): the same pair with “carry”. In (20), hooyeéł “4.3.carrying” has a fourth-person object (Bill, the one carried, a topic by H49) and a third-person subject, the only third-person participant, to which łį́į́ʼ is linked (H4″): the horse carries Bill. In (19) the gloss “3.3” does not say who carries whom; Bill łį́į́ʼ booyeéł fits an inverse with the carried one first (H3) if b- here is the inverse prefix, but the source does not gloss it INV, so this is not counted (OPEN). Result: the event/person contrast follows from person alone (H48 with H31, H32b and H49); no “focus” primitive is needed. H48b is favored over H48a on economy: it keeps one =go (a time clause everywhere, as in review 10’s ??(68), where =go with “hear” gives only a time reading), but the pages do not decide it (OPEN prediction below). OPENNOT UNDERSTOOD: H49 also lets review 10’s *(53) ʼatʼééd joodloh=yę́ę yóóʼajoolwod (NX-359, a noun at the left edge with fourth persons in both clauses) have a topic parse, yet it is starred. Candidate differences: =yę́ę vs =go, a common noun vs a name, or a fourth-person subject vs object in the main verb. None is tested.
Analysis A: (17) NX-411 under H48a: the =go clause is the Theme of “see”; third-person object agreement with the clause Hypothesis
Analysis B: (18) NX-412: Mary is a topic in Spec,CP; the =go clause is a time adjunct on the main TP; the Theme of “see” is fourth person Hypothesis
(1) Event vs person: Mary yicha=go yiiłtsą́ (NX-411) and Mary jicha=go hoołtsą́ (NX-412) HypothesisNX-411/412, Willie 1991 ch. 5 (17)/(18), printed p. 141: Mary yicha=go yiiłtsą́, glossed “3.crying=while 3.1.saw”, “I saw Mary crying”; Mary jicha=go hoołtsą́, glossed “4.crying=while 4.1.saw”, “I saw Mary when she was crying. (Mary, I saw her when she was crying.)” Leaves are printed tokens; splitting =go off as C⁰ is our analysis (the source calls it a complementizer enclitic). C⁰, T⁰, v⁰, the silent pronouns and the role labels are our assignments.Bracket notation (copy)
A: [CP [C′ [TP [DP Experiencer: silent first person “I” (the seer)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [CP [C′ [TP [DP Mary] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [V⁰ yicha]]] v⁰]] T⁰]] [C⁰ =go]]] [V⁰ yiiłtsą́]]] v⁰]] T⁰]] C⁰]]
B: [CP [DP Mary] [C′ [TP [CP [C′ [TP [DP Agent: silent fourth person w₄ (ji-)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [V⁰ jicha]]] v⁰]] T⁰]] [C⁰ =go]]] [TP [DP Experiencer: silent first person “I” (the seer)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [DP Theme: silent fourth person z₄ (ho-), a person, never the event (H48)] [V⁰ hoołtsą́]]] v⁰]] T⁰]]] C⁰]]
A: [CP [C′ [TP [DP Experiencer: silent first person “I” (the seer)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [CP [C′ [TP [DP Mary] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [V⁰ yicha]]] v⁰]] T⁰]] [C⁰ =go]]] [V⁰ yiiłtsą́]]] v⁰]] T⁰]] C⁰]]
B: [CP [DP Mary] [C′ [TP [CP [C′ [TP [DP Agent: silent fourth person w₄ (ji-)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [V⁰ jicha]]] v⁰]] T⁰]] [C⁰ =go]]] [TP [DP Experiencer: silent first person “I” (the seer)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [DP Theme: silent fourth person z₄ (ho-), a person, never the event (H48)] [V⁰ hoołtsą́]]] v⁰]] T⁰]]] C⁰]]
Structure as a list
Analysis A
CP
C′
TP
DP (id I-high; landing site of Move from I): Experiencer: silent first person “I” (the seer)
T′
vP
DP (id I; struck copy; moved to I-high): Experiencer: silent first person “I” (the seer)
v′
VP
V′
CP (id goCPA; goal of Agree third-person object = the clause (H48))
C′
TP
DP (id S1-high; landing site of Move from S1): Mary
T′
vP
DP (id S1; struck copy; moved to S1-high): Mary
v′
VP
V′
V⁰ (head; id V1; head-moves to v1): yicha
v⁰ (head; id v1; head-moves to T1)
T⁰ (head; id T1; hosts the “verb word” (V1 + v1 + T1))
C⁰ (head; id C1): =go
V⁰ (head; id V2; head-moves to v2; probe for Agree third-person object = the clause (H48) with goCPA): yiiłtsą́
v⁰ (head; id v2; head-moves to T2)
T⁰ (head; id T2; hosts the “verb word” (V2 + v2 + T2))
C⁰ (head; id C2)
Analysis B
CP
DP (id topic): Mary
C′
TP
CP (id goCPB)
C′
TP
DP (id S1-high; landing site of Move from S1): Agent: silent fourth person w₄ (ji-)
T′
vP
DP (id S1; struck copy; moved to S1-high): Agent: silent fourth person w₄ (ji-)
v′
VP
V′
V⁰ (head; id V1; head-moves to v1): jicha
v⁰ (head; id v1; head-moves to T1)
T⁰ (head; id T1; hosts the “verb word” (V1 + v1 + T1))
C⁰ (head; id C1): =go
TP
DP (id I-high; landing site of Move from I): Experiencer: silent first person “I” (the seer)
T′
vP
DP (id I; struck copy; moved to I-high): Experiencer: silent first person “I” (the seer)
v′
VP
V′
DP (id z4; goal of Agree fourth-person object (ho-)): Theme: silent fourth person z₄ (ho-), a person, never the event (H48)
V⁰ (head; id V2; head-moves to v2; probe for Agree fourth-person object (ho-) with z4): hoołtsą́
v⁰ (head; id v2; head-moves to T2)
T⁰ (head; id T2; hosts the “verb word” (V2 + v2 + T2))
C⁰ (head; id C2)
H45: what the pages show, judgment by judgment
Example
Clause 1 (agent, patient)
Clause 2 (agent, patient)
H45 result
Printed
NX-399 (7)
x₃, y₃
= x₃ (H45b), = y₃ (H45a + c)
good, one reading
good, “heᵢ … heᵢ … himⱼ”
NX-400 (8)
x₃, y₃ (inverse)
same (inverse)
good
good
NX-401/402 (9a/b)
x₃, y₃
same by role (H45b-θ); swapped by position (H45b-pos)
good either way
good; subscripts fit θ, prose fits pos (OPEN)
NX-403 (10)
x₃, y₃
x₃, fourth-person patient
bad (H45a)
*
NX-404 (11a)
z₄, y₃
free (y₃ printed), z₄
good
good
NX-405 (11b)
x₃, z₄
z₄, x₃
good
good
NX-406 (12), NX-409 (16a)
x₃, y₃
x₃, “the girl”
bad (H45a + H46)
*
NX-407 (13)
z₄, y₃
new u₃, y₃
good
good, “heₖ … himⱼ”
NX-408 (14)
x₃, z₄
x₃, x₃
bad (H45c)
*
NX-410 (16b)
x₃, y₃
x₃, “a girl”
good (indefinite)
good
Does light d earn its keep?
HYPOTHESISFair test. Three places a silent d could work. (a) The definite/indefinite contrast that decides *(16a) vs (16b) is carried by overt léiʼ; a silent definite d on bare ʼatʼééd adds nothing, and an always-definite silent d is already contradicted by review 10’s bare łį́į́ʼ “a horse” (NX-372). (b) The =go clause as an argument of “see” (H48a) would use the overt =go, not a silent head; H48b needs no D at all. (c) The topic Mary needs no D to be a topic. Result: light d is not needed. The DP/TP symmetry point (a light n in the noun phrase parallel to light v) stays OPEN; nothing on these pages bears on it.
Telicity test
SOURCEDContradicted (narrow claim): “the form of =go depends on telicity.” The same =go attaches to “come in” and “sit down” (NX-395–398), to “kick” and “rope” (NX-399–410) and to “cry” and “carry” (NX-411–414). Calling the first two groups bounded and the last two activities is our classification, not the source’s. Not a minimal pair, but a fixed form is enough to contradict a claim about the form.
OPEN“=go is read ‘while’ with activities and ‘when / as’ with bounded events.” The source’s glosses and translations line up that way (“while” only with “crying” and “carrying”), but the verbs differ, so this is not controlled. It needs one verb in both aspects.
SOURCEDThe event/person contrast is not driven by telicity. (17) vs (18) and (19) vs (20) keep the same verbs, and so the same telicity, and change only the person prefixes, and the readings change. (Within the pair the telicity is held fixed; the conclusion concerns only this contrast.)
OPEN H45 is tested only on “kick”, “fist-hit” and “rope”. Whether forced pairing also holds with activity verbs is untested.
Negation
SOURCED No negation is on these pages. A text-layer search of printed pp. 132–142, checked against the renders, finds doo only inside dooleeł (NX-110, “will”), no da except the particle dah of dah neezdá “sat down”, and no hanii, question enclitic or question mark. Nothing new on sentential versus constituent negation.
Question × polarity × voice grid (cumulative after review 13)
HYPOTHESIS No cell changes. These pages have no question and no negation. The inverse declaratives (8), (9a), (9b) are translated with English passives, but an English passive translation does not fill a Navajo passive cell, and the source glosses them INV, not passive.
OPEN Still empty: polar-affirmative × passive; polar-negative × inverse and × passive; wh-negative × inverse and × passive; any negative passive question. Places to search next: Willie ch. 2 (passive and indefinite agents) and ch. 4 for questions with bi- verbs; the Navajo Questions database for bi- verbs with doo … da; a question inside an =go clause or a reported clause.
Where we part from the source’s analyses (labeled contrast)
HYPOTHESISRestated: Willie’s (6) and (15a–c) all follow from H45 with H32a/b: (15a) from H45a + H45b + H45c, (15b) from H45a (a fourth-person second-clause patient needs a fourth-person participant in clause 1), (15c) from H45b (agents pair only when their persons match). Her exception (16b) is H45a’s indefinite clause.
HYPOTHESISNot adopted: “Navajo requires … coreferent patients” (p. 139) as stated; contradicted by her (11a/b) and (16b). Also “a new referent cannot be introduced in the second clause” (pp. 137–138), contradicted by her (13) (heₖ) and (16b) (a girl).
HYPOTHESISNot adopted: “focus” on a person vs an event as a primitive. On our account the readings come from person: an event is third person (H48), and the topic of (18)/(20) is H49.
HYPOTHESISNot settled: whether the yi-/bi- alternation swaps roles (her prose) or keeps them (her subscripts) in (9a/b).
What this changes
HYPOTHESISH45 (a, b, c) replaces H32c and derives every judgment on pp. 133–140. Its core (why two transitive clauses, why the patient half is strict) is NOT UNDERSTOOD. H45b-θ vs H45b-pos is OPEN on a source conflict.
HYPOTHESISH42: its explanation of the reverse pattern is partial (it explains availability, not obligatoriness); the general “dedicated form wins” version is contradicted by NX-395.
HYPOTHESISH46, H47, H48 (a/b), H49 added. H47 is forced by (4) under the working default. H48 derives the event/person contrast; H48b favored over H48a, not settled. H49’s scope is limited by *(53) (NOT UNDERSTOOD).
HYPOTHESISPass out of sample: H3, H6 (silent inverse patients), H4″ and H31 (NX-396, 397, 411, 414), H32a, H32b (NX-398, 404, 405, 412).
Working default (V→v→T, subject to Spec,TP): used in all forty clauses; nothing here resists it. What else is needed for the order: H47 (the =go clause above the main Spec,TP, from (4)); objects stay in VP (*(12), (16b)); the topic of (18)/(20) is merged in Spec,CP, not moved. If (19) is an inverse, the carried one moves above the carrier (H3), the same extra Move as in earlier reviews.
HYPOTHESIS Light d is not needed. Telicity: a telicity-dependent =go form is contradicted, the event/person contrast holds with telicity fixed, and the rest is OPEN. No grid change.
OPEN predictions
OPENH45b-θ vs H45b-pos: a clean, sourced judgment on a mixed-voice pair like (9a)/(9b): does yiztał=go nábidííłtsʼin mean that the kicker was also the hitter (θ) or that the kicked one hit back (pos)?
OPENH45b: (11a) should also allow a new hitter (agent free, as in (13)).
OPENH45 and transitivity: a transitive =go clause followed by an intransitive one, and the reverse. Is pairing forced there? H45a predicts no requirement when clause 2 has no patient.
OPENH46: in a transitive pair, a noun in clause 2 that repeats clause 1’s patient (the same noun in both clauses) should still be starred or read as someone else.
OPENH48a vs H48b: does (17) also allow a person reading (“I saw her (Mary) while she was crying”)? H48b allows it (it₃ can be Mary); H48a does not.
OPENH49: what blocks the topic parse in *(53)? A minimal pair that varies only one of =yę́ę/=go, a name or a common noun, or a subject or object fourth person.
OPENTelicity: the same verb in an =go clause in a bounded and an activity form, to see whether the “when / while” reading tracks it.
OPENGrid: still no polar passive question and no negative inverse or negative passive question.