Inverse, ranking and the fourth person

Legacy link recovery
Legacy section links (scroll sideways); old articles remain available.
who-acts-on-whomeight-ranksinverseinverse-agreedeep-or-shallow-inversealgonquian-inversetwo-osvone-proximateranked-obviativeshale-general-bfourth-personfourth-person-usesfourth-person-comparedclause-chainsreported-speechfourth-subject-p

On the spine (2026-11-14 summary)

2026-11-14 overhaul (approved by Grant for the live site, 2026-11-14). This summary states the topic on the spine (V→v→T and subject raising to Spec,TP assumed; each pattern is what else moves). It supersedes the framing of the dated sections below, whose data, judgments and citations are unchanged.

  • HYPOTHESIS The inverse is one extra movement: the patient moves from VP to a position above the agent, which has already raised to Spec,TP (H3). The upper copy is pronounced; the lower copy keeps the theta role, so who does what to whom is unchanged: ashkii atʼééd yi yiiłtsą́. (NX-001) vs ashkii atʼééd b-iiłtsą́. (NX-002). Only a definite or “which one” patient can move (H6′). Results §2.
  • HYPOTHESIS On the way the patient passes an outer Spec,vP, where the plural da- also counts it (H21b): *NX-278. Results §7.
  • HYPOTHESIS In relatives, the enclitic goes with the third person when the other participant is fourth person (H31): haztał =yę́ę (NX-319).
  • OPEN The trigger and landing site of the inverse movement. The animacy ranking describes when the inverse is used but is not yet stated as a trigger. Whether the yi-/bi- switch across =go clauses keeps or swaps who does what: Willie’s printed subscripts for (9a)/(9b) and her prose disagree (H45b-θ vs H45b-pos; clause spine, L4; results §11).
  • HYPOTHESIS Fourth person across clauses (reviews 12–13): a fourth-person sayer of jiní is a respected person or nobody in particular (H43); two fourth persons in one sentence can be the same person (H32b); in =go pairs a fourth-person second patient needs a fourth-person participant in the first clause, so *yiztał=go náhodííłtsʼin (NX-403) is out (H45a). Results §10–11.

Who can act upon whom? Witherspoon 1977 on yi-/bi-, read against Dowty's proto-agent properties

Working note. Kept as written (URL, data and status labels unchanged). Filed under Ranking, animacy and yi-/bi-; the consolidated argument is the essay who-acts-on-whom. Rule IDs: rules index.

Verbs on this page → dictionary: yiztał ‘kick’ · yiyiisxį́ ‘kill (one)’ · yáłtiʼ ‘speak, talk (to)’ · yoodlą́ ‘believe’ · yoodlą́ą́ʼ ‘drink (it)’ · yikʼichʼínímááz ‘roll upon (it)’ · yikʼiikę́ę́z ‘fall upon (it)’ · ʼayííłʼééł ‘carry off (by water)’ · yistin ‘freeze (it)’ · nímááz ‘rolled; with yaa/baa ‘to’: rolled to’

Restated on the unified structure (2026-11-15). The inverse is restated as one extra Move on the clause spine: the patient moves (via an outer Spec,vP) above the agent, which has already raised to Spec,TP by the working default (H3, H6′, H21b). The event LF is the same as the direct; the trigger and landing site are OPEN. 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. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §4, inverse row. 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. The §7 badge read “CONTESTED”, which is not in the site’s label set; it is now OPEN (contested). Separately, the build no longer turns the closing quote of an English translation inside a Navajo row into a glottal stop (‘…mule’, not ‘…muleʼ).

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.

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.

This page reads the source that the theta-args thesis page §3b had only secondhand: Gary Witherspoon's Language and Art in the Navajo Universe (1977), chapter 3, section I, “Classifying Interaction” (pp. 63–81). Witherspoon states the ranking behind yi-/bi- in terms of control, intelligence, will and movement. Those are close to Dowty's (1991) Proto-Agent properties. He also supplies Grant's missing test: what happens when the two participants are of equal rank. The result is mixed, and it is stated plainly in §6: equal-ranked animates allow both voices, but equal-ranked inanimates allow no inverse at all, in either order. Written 2026-10-14 (PT), 4:46am push.

Update, 12:46pm the same day. Creamer 1974 has now been read in the original (Eight ranks). Two things on this page change. (1) INV-CAP is withdrawn: Creamer's preferred Tʼiis są́ biisxį́ ‘The tree died of old age’ (1974:37) is a bi- sentence with a tree as subject. (2) The equal-rank verdict changes from SPLIT to SUPPORTS. Creamer states that among plants and objects the one the verb says is moving “takes higher status than the more stationary object” (1974:36), so rock and tree are not of equal rank in (11)–(14), and the ordinary ranking rule already gives yi- only. Sections 6.2, 8, 9 and 10 below are kept as written at 4:46am, with notes.

On this page: 1 Question · 2 Source and fidelity · 3 Data · 4 What Witherspoon says the ranking is · 5 Mapped onto Dowty · 6 The equal-rank test · 7 Does bi- mean ‘let’? · 8 Rules · 9 Derivations C1–C4 · 10 Verdict for the thesis · 11 Open

1. The question

GRANT'S CONJECTURE The animacy ranking behind yi-/bi- is Dowty's Proto-Agent weight (sentience, volition, causation, movement), extended in two ways: the properties are weighted, and person and topicality/definiteness are added (thesis §3b). If that holds, the inverse ranks roles (participants by their proto-agent potential), not words. Grant named the remaining test: participants of equal rank.

Two items were open: TA-g (“does a source tie the scale to control and agency?”) and the equal-rank test, for which the site had Willie 1991:86 (70) (“if they are equal, the choice rests on discourse factors”) and the horse/mule pair (Willie (37), p. 71) only.

2. Source and fidelity

Opened this run (2026-10-14, ~5:00am PT). Witherspoon, Gary 1977. Language and Art in the Navajo Universe. Ann Arbor: University of Michigan Press. Open-access edition (CC BY-NC 4.0, July 2025), ISBN 978-0-472-90555-3, DOI 10.3998/mpub.9705. The OAPEN record (library.oapen.org handle 20.500.12657/105220) sits behind a bot check, so the PDF was taken from the publisher's own open-access download on Fulcrum (the same edition). The PDF is a page scan with no text layer, so everything was read on page images: printed pp. 63–82 at 110 dpi (PDF pp. 84–103; PDF = printed + 21), every Navajo line at 300 dpi, and the verb words of (11), (13), (18) and (20) at 600 dpi.

Fidelity notes. (a) Witherspoon is an anthropologist, and his argument is ethnosemantic. His judgments come from his wife (a Navajo speaker, p. 67) and “several Navajos who knew English” (p. 67). He reports degrees of “absurdity”, not grammaticality marks: (6) is “grammatically regular but culturally unacceptable” (p. 65). Stars below are ours and mean only “Witherspoon reports it unacceptable”. (b) His glosses are word-by-word in parentheses, e.g. “(it-it-kicked)”, with no yi-/bi- segmentation. They are copied as printed. The voice column is ours, read off the first letter of the verb word. (c) He prints ashkii without the initial glottal stop that Willie prints (ʼashkii). (20) abííłʼééł has no initial apostrophe, though its direct partner (18) ʼayííłʼééł has one. Both are kept as printed. Apostrophes are printed straight (') and are given here as ’, as elsewhere on the site. (d) His English translations are his own re-glosses (‘the horse let the mule kick it’); the conventional translations he argues against are given too, as he gives them.

3. Data SOURCED

All examples are Witherspoon 1977, page-checked. Each is one clause: two nouns, then the verb word. The first noun of a bi- sentence is the one acted on (as in Sandoval & Jelinek (2) and Willie (37b); inverse §2.1).

3.1 ‘Kick’: horse, mule, man, boy, baby

Witherspoon 1977:63–65, 72, 79–80
Navajo, gloss as printedvoicetranslation and judgment as reported
(1)Łį́į́ʼdzaanéézyiztałyi-‘the horse kicked the mule’ (p. 63)
(horse)(mule)(it-it-kicked)
(2)Łį́į́ʼdzaanéézbiztałbi-‘the horse was kicked by the mule’ (pp. 63–64; Witherspoon: ‘the horse let the mule kick it’, p. 67)
(horse)(mule)(it-it-kicked)
(3)Hastiinłį́į́ʼyiztałyi-‘the man kicked the horse’, acceptable (p. 64)
(man)(horse)(it-it-kicked)
*(4)Łį́į́ʼhastiinbiztałbi-‘the horse was kicked by the man’, “not just poor grammar; it is an impossibility in the Navajo world” (p. 64); Witherspoon: ‘the horse let the man kick it’ (p. 68)
(horse)(man)(it-it-kicked)
(5)Hastiinłį́į́ʼbiztałbi-‘the man was kicked by the horse’, acceptable (p. 65); Witherspoon: ‘the man let the horse kick him’ (p. 68)
(man)(horse)(it-it-kicked)
*(6)Łį́į́ʼhastiinyiztałyi-‘the horse kicked the man’, “grammatically regular but culturally unacceptable” (p. 65)
(horse)(man)(it-it-kicked)
*(15)Łį́į́ʼashkiiyiztałyi-‘the horse kicked the boy’: “one does not say” (p. 72)
(horse)(boy)(it-it-kicked)
(16)Ashkiiłį́į́ʼbiztałbi-‘the boy was kicked by the horse’ (p. 72); Witherspoon: ‘the boy had himself kicked by means of the horse’ (p. 73)
(boy)(horse)(it-it-kicked)
*(b1)Awééʼhastiinyiztałyi-‘the baby kicked the man’, “unacceptable” (p. 79; unnumbered, labelled b1–b4 here)
(baby)(man)(it-it-kicked)
(b2)Hastiinawééʼbiztałbi-“one must say ‘the man let the baby kick him’” (p. 79)
(man)(baby)(it-it-kicked)
(b3)Awééʼłį́į́ʼyiztałyi-‘the baby kicked the horse’, acceptable (pp. 79–80)
(baby)(horse)(it-it-kicked)
(b4)Łį́į́ʼawééʼyiztałyi-‘the horse kicked the baby’, acceptable (p. 80; no gloss line printed)
[horse][baby][it-it-kicked]

(1)–(6) and (15)–(16) agree with Willie (37) and Sandoval & Jelinek (1), (2), (5), (6) on the inverse page. The baby pairs are new to the site. They show that a human baby ranks with animals, not with adult humans (p. 80): “the human baby who can only call and cry is considered to be approximately equal to animals who can only call and cry.”

3.2 ‘Drink’ and ‘kill’: girl and water

Witherspoon 1977:65–66, 73
Navajo, gloss as printedvoicetranslation and judgment as reported
(7)Atʼéédtóyoodlą́ą́ʼyi-‘the girl drank the water’, acceptable (p. 65)
(girl)(water)(it-it-drank)
*(8)Tóatʼéédboodlą́ą́ʼbi-‘the water was drunk by the girl’, “unacceptable and absurd” (p. 65); = (17), p. 73; Witherspoon: ‘the water let (or caused) the girl drink it’
(water)(girl)(it-it-drank)
(9)Atʼéédtóbiisxį́bi-‘the girl can be drowned (killed) by the water’ (p. 65); “the girl can let the water kill her” (p. 69)
(girl)(water)(it-it-killed)
*(10)Tóatʼéédyiyiisxį́yi-“the water cannot drown (kill) the girl” (pp. 65–66)
(water)(girl)(it-it-killed)

P. 65: “this absurdity varies in degree. Sentence (8) is much more absurd than sentence (6), but sentence (6) is only slightly more absurd than sentence (4).” (9)/(10) match Willie's lightning/horse judgment ((73), p. 90; inverse §2.4): with a natural force acting on an animate, the inverse is the preferred form.

3.3 Two inanimates: rock and tree, water and stick, snow and milk

Witherspoon 1977:66, 74–75
Navajo, gloss as printedvoicetranslation and judgment as reported
(11)Tsétʼiisyikʼichʼínímáázyi-‘the rock rolled upon the tree’, acceptable (p. 66)
(rock)(tree)(it-upon-horizontally-it-rolled)
*(12)Tʼiistsébikʼichʼínímáázbi-‘the tree was rolled upon by the rock’, unacceptable: “an event that, strictly speaking, could not have occurred” (p. 66); Witherspoon: ‘the tree let the rock roll against it’ (p. 74)
(tree)(rock)(it-upon-horizontally-it-rolled)
(13)Tʼiistséyikʼiikę́ę́zyi-‘the tree fell upon the rock’, acceptable (p. 66)
(tree)(rock)(it-upon-it-fell)
*(14)Tsétʼiisbikʼiikę́ę́zbi-‘the rock was fallen upon by the tree’, unacceptable (p. 66)
(rock)(tree)(it-upon-it-fell)
(18)Tótsinʼayííłʼééłyi-‘the water carried off the stick’, “culturally acceptable” (pp. 74–75)
(water)(stick)(away-it-it-carried)
*(20)Tsintóabííłʼééłbi-‘the stick let the water carry it off’, “absurd” (p. 75)
(stick)(water)(away-it-it-carried)
(19)Yasabeʼyistinyi-‘the snow froze the milk’, “culturally acceptable” (pp. 74–75)
(snow)(milk)(it-it-froze)
*(21)Abeʼyasbistinbi-‘the milk let (or caused) the snow freeze it’, absurd (p. 75)
(milk)(snow)(it-it-froze)

P. 66: “Sentences (11) to (14) seem to prove that to be wrong [Hale's view that higher-ranking nouns always come first], because in sentence (11) the “rock” is allowed to precede the “tree” but not in sentence (14). Likewise, the “tree” is allowed to precede the “rock” in sentence (13) but not in sentence (12).” P. 74: “when two inanimate beings or entities are interacting, the inversion rule is not an option.”

4. What Witherspoon says the ranking is SOURCED

  • Control. The pattern “deals with the phenomena of movement and action, referring particularly to who can control whom, or who can act upon whom” (p. 67).
  • Will and intent. On (9)/(10): “The water does not possess intent or will and thus cannot control or act upon the girl” (p. 69). On (8): his wife's explanation, “The sentence attributes more intelligence to the water than it does to the girl” and “the water does not think, so how could it have the girl drink it” (p. 67).
  • Movement. “Sentences (11) through (14) reveal that movement and action are part of the considerations that decide who can control or act upon whom. In sentence (11) the rock hits the tree. The rock in this case is superior to the tree because of its locomotion” (p. 69). For two moving inanimates: “If they both have been activated (put in motion), then the one whose animation is the most potent is considered to have the capacity to act upon or control the one with the least potent animation” (p. 81; no example printed).
  • A weighting, stated by the source. Who controls whom “is determined primarily by the level of one's intelligence and, to a lesser extent, by the potency of one's animation” (p. 77). He puts this as a shorthand formula, “C(control) = M(mass) × I²(intelligence)” (pp. 77, 79, 80). It is his metaphor for a graded scale, not a measurement, and it is not used on this site beyond that.
  • The classes (Fig. 4, p. 78). Animate: speakers and callers (callers “further divided into graded categories based on the degree of intelligence and the potency of animation”). Inanimate: corporeal (subdivided into activated and stationary) and incorporeal (abstractions such as hunger and thirst, “the least potent of all things”, p. 79). The speaker/caller line follows Haile 1943:76–77 (quoted p. 79): yáłtiʼi ‘one who speaks’ vs do yáłtiʼi ‘non-speaker’, and animals' ʼání ‘it gives a call’ vs yáłtiʼ ‘he speaks’. (The site's dictionary entry for yáłtiʼ ‘speak’ is here.)
  • Not deep grammar. “They are not generated by a set of operations at the deep structural level of Navajo grammar; they are generated by a set of cultural rules which are ultimately derived from more fundamental metaphysical propositions” (pp. 66–67).
  • Hale 1973, as quoted by Witherspoon (Hale's paper itself is still not opened; SOURCED (secondhand, page-checked in Witherspoon)). Hale first gave an animacy table (1973:303, quoted p. 69): animate–animate, inversion optional; animate subject–inanimate object, blocked; inanimate–inanimate, blocked; inanimate subject–animate object, obligatory. He then preferred a ranking hypothesis (1973:304, quoted p. 70): equal rank, optional; subject outranks object, blocked; object outranks subject, obligatory, because the rule “applies in such a way as to ensure that prominence is given to the higher ranking noun phrase”. He added that “the conditions on inversion cannot be formulated in absolute terms” (1973:305) and that in some areas “it is perhaps more appropriate to speak of relative acceptability” (1973:306, quoted pp. 80–81).
  • Creamer 1974, as quoted by Witherspoon. The characteristics of each being “appear to be summed up in an inherent ‘capacity to act upon’ in relation to other beings”, on the basis of “who can be expected to be able to act upon whom” (1974:30, quoted p. 71). Nouns “may freely act upon other nouns of equal or lower status but not on nouns of higher status” (Witherspoon's paraphrase of 1974:32, p. 71). The eight groups are 1974:33–37 (pp. 71). The bee/spider/mouse example: “a bee can sting a spider but not a mouse; the mouse can only let itself be stung by the bee (1974:36)” (Witherspoon's paraphrase, p. 80). Witherspoon “strongly agree[s]” with Creamer's “capacity to act upon” as the basis of the categories (p. 75).

5. Mapped onto Dowty's proto-agent properties

Baseline: Dowty 1991:572 lists the Proto-Agent properties as volition, sentience/perception, causing an event or change of state, movement relative to another participant, and independent existence. The mapping below is HYPOTHESIS (ours, in Grant's framing). Witherspoon never cites Dowty, who wrote fourteen years later. The left column is SOURCED.

Witherspoon's criterion (page)Dowty 1991 Proto-Agent propertyFit
“intelligence”, “capacity for thought”, “intent or will” (pp. 67–69, 77)volition, sentienceClose. Witherspoon makes this the main factor (“primarily”, p. 77).
“locomotion”, “movement and action”, “potency of animation”, activated vs stationary (pp. 69, 77, 81)movement; causationClose. Secondary factor (“to a lesser extent”, p. 77).
“who can control whom, or who can act upon whom” (p. 67); Creamer's “capacity to act upon”causation (of an event or change in the other participant)Close, but stated as a capacity, not an entailment of a particular verb (see §6.3).
corporeal vs incorporeal (p. 79)independent existence (loosely)Loose. Dowty's property is existence independent of the event, not having a body.
speakers vs callers (pp. 78–80)no direct counterpartWitherspoon ties speech to intelligence and control (p. 80). On the mapping it is a further grading of sentience/volition. This is an extension of Dowty's list.

Result. SOURCED A published source does tie the Navajo ranking to control, intelligence, will and movement. HYPOTHESIS Those are, nearly one for one, Dowty's volition, sentience, causation and movement. The source also states a weighting (intelligence first, animation second). That is exactly extension (a) on the thesis page, which had been a hypothesis built from one judgment of Willie's ((73), p. 90). It now has a direct source. It is still one author's ethnosemantic account, though, not a grammatical study, and the mapping onto Dowty is ours.

6. The equal-rank test

6.1 Equal-ranked animates: both voices SOURCED

Horse and mule: (1) and (2) are both acceptable (p. 63), as in Willie (37) (p. 71: “of equal animacy”, so the alternation is “discourse determined”, p. 72). Baby and horse are “approximately equal” (p. 80), and (b3) and (b4) are both acceptable direct sentences, with either one as the kicker. Witherspoon prints no bi- form for baby and horse. Hale's view, as quoted: with equal ranks “inversion [is] optional” (1973:304). Verdict: SUPPORTS the tie-breaking extension for animates. But Witherspoon adds that the two voices are not equivalent in meaning (§7), so “discourse decides” may not be the whole story even here.

6.2 Equal-ranked inanimates: no inverse at all SOURCED

Rock and tree are in the same class on every scale the site uses: Creamer's group 7 (plants and inanimate objects; p. 71), Willie's “inanimate”, and Witherspoon's corporeal inanimates. Under Willie (70) and Hale's ranking hypothesis, equal rank should make the voice free. It is not. The direct is fine in both orders ((11) rock acting on tree; (13) tree acting on rock), and the inverse is out in both ((12), (14)).

Verdict: CUTS AGAINST the rule as the site had stated it: INV-3's “r(ρA) = r(ρX) → free” (inverse §3) and the schema corollary's “When the ranks are equal, both voices are available” (schema §4). Both are too general. They hold for animates only. Hale's earlier animacy table (inanimate–inanimate: blocked, 1973:303) did predict this. Hale's later ranking hypothesis did not, and Witherspoon presses that point (p. 74).

Caveat on (18)–(21). Witherspoon calls water/stick and snow/milk “nouns of approximately the same status” (p. 75), but on Creamer's groups, as he himself lists them (p. 71), stick and milk are plants and inanimate objects (group 7), while water and snow are plausibly natural forces (group 6; the thesis table places water there, but no opened source assigns water or snow to a group). If so, INV-3 already predicts yi- only, with the higher natural force acting, so (20)–(21) are not a clean equal-rank test. Only rock/tree is.

Note, 12:46pm: Creamer's own group 7 rule makes rock and tree unequal in these sentences: the mover outranks the stationary one (1974:36). So (11)–(14) are not an equal-rank test either, and INV-3's “equal → free” is not contradicted. See Eight ranks §6.2.

6.3 The rank reads the event, not the word

SOURCED The same two nouns, tsé ‘rock’ and tʼiis ‘tree’, come out in either order, with the acting one first. Which one acts depends on the event: in (11) the rock rolls, in (13) the tree falls. Witherspoon: “The rock in this case is superior to the tree because of its locomotion … This time, however, it is the tree that is in motion and the rock which is static” (p. 69). HYPOTHESIS In Dowty's terms, the movement that decides is the movement the verb entails of one participant (‘roll upon’ entails that its subject moves; ‘fall upon’ likewise). That is an event entailment, not a fixed property of the word tsé. It answers, for inanimates, the “entailments vs capacities” worry on thesis §3b: among inanimates the deciding property is an entailment of the predicate.

With animates the deciding property is a capacity, but again not a word class. A baby is human, yet it ranks with animals (b1)–(b4), because what counts is the capacity to speak and think (pp. 79–80). Sandoval & Jelinek make the parallel point from the other side: a horse “understood as a supernatural being” outranks a man (1989:336; inverse I2). In both directions, the rank follows what the participant can do, not the lexical item that names it. Verdict: SUPPORTS “the inverse ranks participants by their proto-agent properties, not words”. Two cautions. The lexical item is still how a hearer recovers those properties when nouns are present. And Witherspoon explicitly calls these “cultural rules” rather than deep grammar (pp. 66–67).

7. Does bi- mean ‘let’? OPEN (contested)

SOURCED Witherspoon rejects the inversion analysis: “The use of the bi third person object pronoun form does not in fact cause an inversion of subject and object, and what results from its use is not, as Hale assumes, anything very similar to the passive in English” (p. 72). In (16) “the boy is the prime mover of the action and also the recipient of the action; the horse is only the agent or means by which the action is accomplished. The resulting semantic context of the sentence is more like the English reflexive than it is like the English passive” (p. 73). He reached the ‘let’ translation in discussion with Navajo speakers who knew English (p. 67), and (5) “is acceptable because the man remains the primary source and controller of the action” (p. 68).

Other sources on the site read the two voices differently. Fernald's grammar draft: direct and inverse “have the same meaning but have different uses in discourse”. Hale, “Conflicting Truths”: the same translation for both (sources §2). Sandoval & Jelinek: bi- is “a transitive, inverse construction, and therefore not a passive”, and (5) is “pragmatically unacceptable”, not ungrammatical (1989:336, 339). Willie: the inverse subject is definite and “becomes the discourse topic” (p. 100).

Two readings of Witherspoon's claim, both HYPOTHESIS:

  1. Entailment reading. bi- adds a Proto-Agent entailment (control or permission: volition, causation) to the patient. That would break INV-1 (entailments are the same under yi- and bi-; inverse §3) and the site's working view that Dowty entailments are fixed by the event. It would also conflict with the same-meaning reports above.
  2. Capacity reading (adopted here). bi- does not change the entailments. It requires that the participant it makes the subject be the kind of being that could have controlled or resisted the event. The ‘let’ in Witherspoon's translations spells out that presupposition. This keeps INV-1, fits the same-meaning reports, and still explains why (12), (14), (20) and (21) are absurd rather than merely odd: they attribute a capacity “to think and … to resist” (p. 74) to a tree, a rock, a stick or milk.

Either way, the constraint is stated over a participant's proto-agent potential. Neither reading needs the word that names it.

8. Rules

Notation as on the inverse page. r(ρ) = rank; EXT = the slot realized as subject; cap(ρ) = the potential Proto-Agent properties (sentience, volition) of ρ's referent as construed.

INV-3′   Ranking, restated.  As INV-3, but the "equal → free" clause holds only when INV-CAP is met:
           r(ρ_A) = r(ρ_X) and cap(ρ_X) ≠ ∅  → free (discourse decides)
           r(ρ_A) = r(ρ_X) and cap(ρ_X) = ∅  → EXT = ρ_A  (yi- only)
INV-CAP  Capacity.  bi- (EXT = ρ_X) is defined only if the referent of ρ_X could have controlled or
         resisted the event: cap(ρ_X) ⊇ {sent} or {vol}.  (Hale 1973:303's "inanimate–inanimate: blocked".)
RANK-W   Weighted rank.  r compares cap first (intelligence: sent, vol; for animates, graded, speakers >
         callers), then movement/causation (activated > stationary; for inanimates, read off the event's
         entailments), then nothing.  (Witherspoon 1977:77 "primarily … to a lesser extent".)
RANK-EV  Event-relative rank among inanimates.  If neither referent has cap, the one the verb entails
         moves outranks the static one; voice is still yi- (INV-CAP), so the mover is EXT = ρ_A.
RuleStatusBasis
INV-3′HYPOTHESIS from SOURCED factsWillie (37), (70); Witherspoon (1)–(2), (11)–(14)
INV-CAPHYPOTHESIS (our formulation of Witherspoon pp. 72–74 and Hale 1973:303 as quoted)Every two-noun bi- sentence checked for this page has an animate subject: Willie (37b), (45), (73b); S&J (2), (6); Witherspoon (2), (5), (9), (16), (b2). The bi- sentences with an inanimate subject are all out: Willie *(10b) (p. 206) and Witherspoon *(8), *(12), *(14), *(20), *(21). (*(4) is out for a different reason: its subject is animate but ranked below the agent, so INV-3 excludes it.) No counterexample found yet (T20b). WITHDRAWN 12:46pm: counterexample Creamer 1974:37 Tʼiis są́ biisxį́ ‘The tree died of old age’ (Eight ranks §6.1). INV-3′ is withdrawn with it; RANK-EV becomes RANK-EV′, which sets rank rather than breaking a tie.
RANK-WSOURCED as Witherspoon's claim; the Dowty labels are HYPOTHESISWitherspoon pp. 77–81; Willie (73) (lightning/horse, p. 90)
RANK-EVSOURCED (Witherspoon p. 69, (11)–(14)); as an entailment of the verb, HYPOTHESIS(11) vs (13)

Note on *(8). Water is a natural force (group 6) and the girl a person (group 1), so *(8) is already out by INV-3. INV-CAP adds nothing there, but it predicts the degree: (8) is “much more absurd” than (6) (p. 65), because (8) breaks two conditions (INV-3 and INV-CAP) and (6) only one. That last step is HYPOTHESIS.

9. Derivations C1–C4, step by step

Conventions as on the derivations page. Tags: Σ schema step, A word-internal linearization (rule set A's mirror, inside the left-headed prefixal word), B clause level (rule set B's adjunct ordering). V-H = in-situ licensing (Hale 2003:29). V-P = VPISH raising of EXT to Spec,M/SP inside the word. Verb words are not segmented beyond Witherspoon's glosses. Entailment labels are ours, from the translations.

C1. (11) Tsé tʼiis yikʼichʼínímááz ‘the rock rolled upon the tree’ vs *(12) Tʼiis tsé bikʼichʼínímááz

  1. Σ Σ(√ROLL.UPON) = {ρA (mov, caus; no vol), ρX (contact; static)}. Witherspoon's gloss “it-upon-horizontally-it-rolled” has an ‘upon’ element; whether ρX is the object of a postposition inside the word is not shown by the gloss (OPEN, T20c).
  2. Merge ρX low (Comp,V or Comp,P); Merge v; Merge ρA in Spec,v. A head-final inside the word: ρX⌢V, v⌢VP.
  3. Σ π(ρA) = π(ρX) = 3: VOICE is defined (INV-0). Rank: both inanimate (Creamer group 7), r equal. INV-CAP: cap(ρX) = ∅ (a tree cannot resist), so bi- is undefined. INV-3′: EXT = ρA, VOICE = yi-. RANK-EV agrees: ρA is the one the verb entails moves.
  4. Agree M/S ⟶ EXT = ρA. Move V-P: ρA → Spec,M/SP (nothing crossed). V-H: none.
  5. Spell-out A yikʼichʼínímááz.
  6. B INV-6: tʼiis, next to the verb, links to the internal slot ρX; tsé, outer, links to EXT = ρA. ✓ (11).
  7. *(12): choosing bi- at step 3 would set EXT = ρX (the tree). INV-CAP fails at the Σ step, before any noun is placed. Witherspoon's paraphrase ‘the tree let the rock roll against it’ names the capacity INV-CAP requires. ✗

C2. (13) Tʼiis tsé yikʼiikę́ę́z ‘the tree fell upon the rock’ vs *(14) Tsé tʼiis bikʼiikę́ę́z

  1. Σ Σ(√FALL.UPON) = {ρA (mov; no vol), ρX (static)}. The tree is now ρA, because ‘fall upon’ entails its movement.
  2. Merge, Agree, Move as in C1. Σ INV-CAP: cap(rock) = ∅, so yi- only.
  3. Spell-out A yikʼiikę́ę́z. B tsé (inner) → ρX; tʼiis (outer) → ρA. ✓ (13). *(14) fails as *(12) does.
  4. What C1 and C2 show together. The words tsé and tʼiis have no fixed order relative to each other. Each comes first exactly when it fills ρA, and the schema step, not the lexical items, decides which slot that is. This is the clearest case on the site of a ranking computed over role slots and the entailments the verb assigns them.

C3. (9) Atʼééd tó biisxį́ ‘the girl can be drowned (killed) by the water’ vs *(10) Tó atʼééd yiyiisxį́

  1. Σ Σ(√KILL.SG) = {ρA (caus, mov; no vol), ρP (chg, aff)}, as on the kill (one) entry.
  2. Merge ρP in Comp,V; v; ρA in Spec,v. A ρP⌢V, v⌢VP.
  3. Σ Rank by RANK-W: the girl has sent, vol; water has caus, mov only. Intelligence outweighs animation (p. 77), so r(ρP) > r(ρA). INV-3: EXT = ρP. INV-CAP: cap(girl) ≠ ∅ ✓. VOICE = bi-.
  4. Agree M/S ⟶ EXT = ρP. Move V-P HYPOTHESIS: ρP → Spec,M/SP, crossing ⟨ρA⟩, which stays in Spec,v as Willie's “internal (Ergative)” agent. V-H: none.
  5. Spell-out A biisxį́.
  6. B INV-6: tó (inner) → internal = ρA; atʼééd (outer) → EXT = ρP. ✓ (9).
  7. *(10): yi- sets EXT = ρA (water) against INV-3 at the Σ step. Every later step goes through, as in inverse I2, so the only violation is the ranking. That matches “the water cannot drown (kill) the girl” and Willie's “objectionable, as an animacy violation” for the lightning/horse pair (p. 90).

C4. *(b1) Awééʼ hastiin yiztał vs (b2) Hastiin awééʼ biztał; (b3)/(b4) both yiztał

  1. Σ Σ(√KICK) = {ρA (vol, caus, mov), ρP (aff)}, as on the kick entry.
  2. Σ (b1)/(b2): ρA = baby, ρP = man. RANK-W: speaker > caller, so r(ρP) > r(ρA). INV-3: EXT = ρP, VOICE = bi-. *(b1) chooses yi- and fails at this step; (b2) ✓.
  3. Σ (b3)/(b4): baby and horse are both callers, “approximately equal” (p. 80). INV-CAP holds for both. INV-3′: voice is free. Both printed sentences choose yi-, with either participant as ρA. (A bi- form for this pair is predicted to be possible but is not printed: OPEN, T20a.)
  4. Agree, Move, Spell-out A as in inverse I1 (bi-) and I3 (yi-). B INV-6 links the inner noun to the internal slot.
  5. What C4 shows. The noun awééʼ denotes a human, but the rank computed at the Σ step is that of a caller. The rank reads the participant's capacity, not the noun's lexical class.

VPISH. None of Witherspoon's data tests V-P against V-H. In every derivation the ranking, INV-CAP and the voice choice are settled at the Σ step, before any noun is adjoined (rule set B). That is consistent with Willie's finding that the ranking applies “even when the sentence does not contain any nominals” (p. 86). It is the same result as inverse §5.1, now with inanimate pairs as well.

10. Verdict for the thesis

ClaimEffect of Witherspoon 1977Verdict
Extension (a): the animacy scale is weighted Proto-Agent potential, sentience/volition above causation/movementStated by the source: control “primarily” by intelligence, “to a lesser extent” by animation (p. 77); water lacks “intent or will” (p. 69)SUPPORTS. Upgraded from a one-datum hypothesis to a sourced claim (one ethnosemantic author; Dowty labels ours)
Natural forces are not a hard case (causation, movement)Water acting on a girl needs bi- ((9)/(10)); water acting on a stick takes yi- ((18)); force > inanimate, animate > forceSUPPORTS Grant's framing: forces rank by causation and movement, below anything with sentience
Equal rank → topicality/discourse decidesTrue for animates ((1)/(2); (b3)/(b4)). False for inanimates: rock/tree allow no inverse in either order ((11)–(14))SPLIT: SUPPORTS for animates, CUTS AGAINST for inanimates. Repaired by INV-CAP: the inverse subject must have potential sentience or volition. Revised 12:46pm: SUPPORTS. INV-CAP is falsified, and rock/tree are unequal by Creamer's movement rule (Eight ranks §6.2)
The inverse ranks roles, not wordsRock and tree swap places with the event (C1–C2); a baby ranks as a caller, not as a human (C4); the ranking is computed before any noun is adjoinedSUPPORTS, the strongest sourced case so far. Caution: the source calls it “cultural rules”, not grammar (pp. 66–67)
INV-1 / Dowty entailments fixed under both voicesWitherspoon reads bi- as ‘let’, or reflexive (pp. 72–73)CONTESTED. On the entailment reading it cuts against INV-1; on the capacity reading (adopted, §7) it does not
T-STRONG / T-MID / T-WEAK (who the arguments are)Witherspoon does not discuss argumenthood or adjunctionNEUTRAL. T-STRONG still fails for the reasons on the thesis page

11. Open

#QuestionWhat would settle it
T20aDo equal-ranked animates (baby/horse; Creamer's bee/spider) really allow bi- freely, or does bi- carry Witherspoon's ‘let’ even there?Speaker judgments of bi- forms for (b3)/(b4) and of (2) vs (1) in context
T20bIs there any bi- form with an inanimate subject (a counterexample to INV-CAP)?ANSWERED 12:46pm: yes. Creamer 1974:37 Tʼiis są́ biisxį́ (Eight ranks §6.1)
T20cIs the ‘upon’ in yikʼichʼínímááz / yikʼiikę́ę́z a postposition inside the word, and is the yi-/bi- on it the inverse marker or the postposition's object pronoun (as on yichʼįʼ/bichʼįʼ, inverse §2.5)?Young & Morgan; a morpheme-by-morpheme gloss
T20dEntailment reading or capacity reading of ‘let’ (§7)?A context where the patient plainly could not have resisted (asleep, unaware) but is high-ranked: is bi- still used?
T20eTwo moving inanimates: does “the most potent animation” (p. 81) decide, and how?A sourced example; none is printed
T20fGraded absurdity ((8) ≫ (6) > (4), p. 65): does it count violated conditions (INV-3, INV-CAP), as suggested in §8?Graded judgments from several speakers (bears on T6)

Eight ranks: Creamer 1974 read in the original, and an inanimate inverse subject

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.

Working note. Kept as written (URL, data and status labels unchanged). Filed under Ranking, animacy and yi-/bi-; the consolidated argument is the essay who-acts-on-whom. Rule IDs: rules index.

Verbs on this page → dictionary: yiztał ‘kick’ · yiyiisxį́ ‘kill (one)’ · yikʼichʼínímááz ‘roll upon (it)’ · yikʼiikę́ę́z ‘fall upon (it)’ · nímááz ‘rolled; with yaa/baa ‘to’: rolled to’

Restated on the unified structure (2026-11-15). The inverse is restated as one extra Move on the clause spine: the patient moves (via an outer Spec,vP) above the agent, which has already raised to Spec,TP by the working default (H3, H6′, H21b). The event LF is the same as the direct; the trigger and landing site are OPEN. 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. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §4, inverse row. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

Until this run the site knew Mary Helen Creamer's “Ranking in Navajo Nouns” (1974) only through other authors' quotations. This page reads it directly. Creamer was a native speaker who checked her judgments with other speakers. She sorts nouns into eight rank groups, explains the ranking by a “capacity to act upon” others, and prints two-noun pairs for every group. Three results matter for Grant's thesis. (1) Her data contain a bi- sentence whose subject is a tree, so the rule this site proposed this morning (INV-CAP: an inverse subject must be able to resist) is wrong and is withdrawn. (2) She states the rock/tree rule in her own words: among plants and objects, the one the verb says is moving outranks the stationary one. That explains the rock/tree data without INV-CAP, and the old rule “equal rank → either voice” survives. (3) One datum cuts against a simple count of the event's proto-agent entailments: old age, which causes a tree's death, ranks below the tree. Written 2026-10-14 (PT), 12:46pm push.

On this page: 1 Question · 2 Source and fidelity · 3 Data · 4 What Creamer says the ranking is · 5 Mapped onto Dowty · 6 Results · 7 Rules · 8 Derivations K1–K4 · 9 Verdict for the thesis · 10 Open

1. The question

GRANT'S CONJECTURE The ranking behind yi-/bi- is Dowty's (1991) Proto-Agent weight, extended (weighting; person and topicality), so the inverse ranks roles, not words (thesis §3b). The who-acts-on-whom page (4:46am today) ran Grant's equal-rank test on Witherspoon 1977 and found a split: equal-ranked animals allow both voices, but rock and tree allow no inverse in either order. It repaired this with a new hypothesis, INV-CAP (“the inverse subject must have potential sentience or volition”), and left open T20b: is there any bi- sentence with an inanimate subject? The obvious place to look was the source of the eight groups, Creamer 1974, which the site had only secondhand.

2. Source and fidelity

Opened this run (2026-10-14, ~1:20pm PT). Creamer, Mary Helen 1974. “Ranking in Navajo Nouns.” Navajo Language Review 1(1): 29–38 (the journal also carries a Navajo title, not reproduced here). Navajo Linguistic Society; published by the Center for Applied Linguistics. Read in ERIC document ED 101 549 (the whole issue, 55 pp.; PDF). PDF = printed + 3 (printed pp. 29–38 = PDF pp. 32–41). The PDF has an OCR layer, which was used to find passages and English quotations only. Every Navajo line below was read on page images (PDF pp. 36–40 at 200 dpi, each Navajo line at 500 dpi). The issue's editorial Preface (pp. 5–6) summarizes her paper and is quoted once.

Fidelity notes. (a) Creamer marks with * the structure that is “not preferred” (p. 34), not “ungrammatical”. Stars below mean exactly that. (b) She prints no word glosses, only free translations. The bracketed gloss lines are ours, taken from her translations; the voice column is ours, read off the yi-/bi- at the start of the verb word or postposition. (c) She types barred ł as a crossed letter and hooks under vowels; both are rendered in standard form, and straight apostrophes as ’. (d) Kept as printed: wóláchííʼ ‘ant’ (other sources write wólázhiní); dáʼdeestlʼin ‘dam’ with a plain l; yishjizh/bishjizh with no accent; the ‘roll upon’ verb written as two words, yikʼi chʼínímááz, where Witherspoon writes one word, yikʼichʼínímááz (see §6.6). (e) The bi- partner of the bee/mouse pair is printed with a damaged first letter (it looks like “Lishish”). We read it as bishish, because it is the passive translation and matches bishish in the next pair; that reading is ours.

3. Data SOURCED

All examples are Creamer 1974, page-checked. In a bi- sentence the first noun is the one acted on, as on the inverse page. Creamer: “in cases of unequal rank, the topic of the sentence must be the higher ranking noun phrase” (p. 32).

3.1 Group 1 (persons), the newborn, and lightning (pp. 33–34)

Creamer 1974:33–34
Navajo, gloss oursvoicetranslation as printed
(c1)Iiʼniʼdinéyiłłeeʼ dííłchʼilyi-‘Lightning struck the man.’ (p. 33)
[lightning][man][3-with][struck]
(c2)Dinéiiʼniʼbiłłeeʼ dííłchʼilbi-‘The man was struck by lightning.’ (p. 33)
[man][lightning][3-with][struck]
*(c3)Awééʼchíʼídinéyiztałyi-‘The baby kicked the man.’ (p. 34)
[baby][man][kicked]
(c4)Dinéawééʼchíʼíbiztałbi-‘The man was kicked by the baby.’ (p. 34)
[man][baby][kicked]
(c5)Dinéawééʼchíʼíyiyiisxį́yi-‘The man killed the baby.’ (p. 34)
[man][baby][killed]
*(c6)Awééʼchíʼídinébiisxį́bi-‘The baby was killed by the man.’ (p. 34)
[baby][man][killed]
(c7)Chį́į́h yeeʼidilohiiawééʼchíʼíyishjizhyi-‘The elephant crushed the baby.’ (p. 34)
[elephant][baby][crushed]
(c8)Awééʼchíʼíchį́į́h yeeʼidilohiibishjizhbi-‘The baby was crushed by the elephant.’ (p. 34)
[baby][elephant][crushed]

In (c1)/(c2) the yi-/bi- is on the postposition -ił ‘with’, written as its own word, not on the verb. Group 1 also has four man/boy, boy/girls and leader/white-man pairs, all acceptable in both voices (p. 33); they are not repeated here.

3.2 Groups 2–5: animals and insects (pp. 34–36)

Creamer 1974:34–36
Navajo, gloss oursvoicetranslation as printed
(c9)Shashdibéyiyiisxį́yi-‘The bear killed the sheep.’ (group 2 on group 3, p. 35)
[bear][sheep][killed]
*(c10)Dibéshashbiisxį́bi-‘The sheep was killed by the bear.’ (p. 35)
[sheep][bear][killed]
(c11)Mósítązhiiyinoołchééłyi-‘The cat is chasing the turkey.’ (both group 3, p. 35)
[cat][turkey][is.chasing]
(c12)Tązhiimósíbinoołchééłbi-‘The turkey is being chased by the cat.’ (p. 35)
[turkey][cat][is.chasing]
(c13)Mósínaʼazísíyinoołchééłyi-‘The cat is chasing the gopher.’ (group 3 on group 4, p. 35)
[cat][gopher][is.chasing]
*(c14)Naʼazísímósíbinoołchééłbi-‘The gopher is being chased by the cat.’ (p. 35)
[gopher][cat][is.chasing]
*(c15)Tsísʼnánaʼastsʼǫǫsíyishishyi-‘The bee stung the mouse.’ (group 5 on group 4, p. 36)
[bee][mouse][stung]
(c16)Naʼastsʼǫǫsítsísʼnábishishbi-‘The mouse was stung by the bee.’ (p. 36; first letter damaged in print, see §2)
[mouse][bee][stung]
(c17)Tsísʼnánaʼashjéʼiiyishishyi-‘The bee stung the spider.’ (both group 5, p. 36)
[bee][spider][stung]
(c18)Naʼashjéʼiitsísʼnábishishbi-‘The spider was stung by the bee.’ (p. 36)
[spider][bee][stung]

Group 2 also has bull/bear, dog/mule (both voices) and *bull/man, *dog/boy (p. 34–35); *sheep/baby (p. 35); group 4 rattlesnake/mouse (both voices, p. 35). They follow the same pattern and are not repeated.

3.3 Groups 5–8: a flood, a dam, a rock, a tree and old age (pp. 36–37)

Creamer 1974:36–37
Navajo, gloss oursvoicetranslation as printed
*(c19)Tówóláchííʼyiyiisxį́yi-‘The flood killed the ant.’ (group 6 on group 5, p. 36)
[flood][ant][killed]
(c20)Wóláchííʼtóbiisxį́bi-‘The ant was killed by the flood.’ (p. 36)
[ant][flood][killed]
(c21)Tódáʼdeestlʼinyinahjįʼninígoʼyi-‘The flood came to rest against the dam.’ (group 6 and group 7, p. 36)
[flood][dam][3-against][came.to.rest]
*(c22)Dáʼdeestlʼintóbinahjįʼninígoʼbi-‘The dam stopped up the flood (lit.: The dam had the flood come and lean on it.)’ (p. 36)
[dam][flood][3-against][came.to.rest]
(c23)Tsétʼiisyikʼichʼínímáázyi-‘The rock rolled upon the tree.’ (both group 7, p. 36)
[rock][tree][3-upon][rolled]
*(c24)Tʼiistsébikʼichʼínímáázbi-‘The tree was rolled upon by the rock.’ (p. 37)
[tree][rock][3-upon][rolled]
(c25)Tʼiistséyikʼiikę́ę́zyi-‘The tree fell upon the rock.’ (p. 37)
[tree][rock][3-upon-fell]
*(c26)Tsétʼiisbikʼiikę́ę́zbi-‘The rock was fallen upon by the tree.’ (p. 37)
[rock][tree][3-upon-fell]
*(c27)Są́tʼiisyiyiisxį́yi-‘Old age killed the tree.’ (group 8 on group 7, p. 37)
[old.age][tree][killed]
(c28)Tʼiissą́biisxį́bi-‘The tree died of old age.’ The preferred structure (p. 37)
[tree][old.age][killed]

(c23)–(c26) are the same four sentences as Witherspoon 1977 (11)–(14) (who acts on whom §3.3), with the same judgments; Witherspoon prints ‘roll upon’ as one word. P. 37, on group 8: “You may find it of some comfort to know that in Navajo old age cannot kill you. You may only die of it.”

4. What Creamer says the ranking is SOURCED

  • Five qualities. “Qualities which tend to give higher status are: 1. capacity for having intent or purpose; 2. intelligence; 3. strength, vigor, aggressiveness, or special potency; 4. usefulness to man, or relatedness to man; and 5. animation, or capacity for movement” (p. 30).
  • Capacity to act upon. These are “summed up in an inherent ‘capacity to act upon’ in relation to other beings”. Each group “is ranked above all other groups upon which its members can act freely, but which cannot freely act upon them” (p. 30).
  • From usage, not theory. “The actual rankings … have not been derived from this common sense rationale, but from the review of a large number of typical Navajo sentences … The rationale was derived by inference out of usage, rather than the other way around” (p. 30).
  • The verb is secondary. The inversion “is governed by reference to a system of classification of nouns into status groups … Within this system, the nature of the verb plays a secondary role” (p. 29).
  • Equal rank: either voice. “If there is no difference in status … either noun may be chosen to be the derived, or surface subject of the sentence. The subject chosen by the speaker is the object of major interest or emphasis” (p. 32). Editorial Preface: “The rule applies freely when the subject and object are equal in rank” (pp. 5–6).
  • The eight groups (pp. 33–37): 1 persons; 2 larger animals and medium-sized ones “of special intelligence or relationship to man (such as the dog) and predators”; 3 medium-sized animals; 4 small animals; 5 insects, spiders, worms; 6 natural forces (“windstorms, flood, sunshine (heat), and forest or range fire”); 7 plants and inanimate objects; 8 abstractions (“old age, hunger, disease … emotions, or other things of which no action is seen other than effects”).
  • Group 7 rule. “In each case, the semantic content of the verb implies movement on the part of one of the nouns in the sentence, and this noun then takes higher status than the more stationary object” (p. 36).
  • Two group-1 exceptions. Lightning, “when used with a verb denoting its characteristically potent striking force”, ranks with persons. A newborn ranks below other humans and with the larger animals, “only as long as the infant retains the helplessness characteristic of the newborn. As soon as he becomes capable of intent and movement, he is treated as fully human” (p. 33).
  • Personification. Group 7 nouns and some group 2 and 3 nouns, when personified in stories, are “treated as of equal status with whatever category of nouns it is interacting with … either yi- or bi- form verbs may be used” (p. 37).

5. Mapped onto Dowty's proto-agent properties

Baseline: Dowty 1991:572, Proto-Agent properties volition, sentience/perception, causation, movement, independent existence. Left column SOURCED; mapping HYPOTHESIS (ours).

Creamer's quality (p. 30)Dowty 1991Fit
1. intent or purposevolitionClose.
2. intelligencesentienceClose.
3. strength, vigor, aggressiveness, special potencycausation (as a capacity to cause effects)Partial. It reads as a degree of causal power. Dowty's causation is yes/no per event, so this is part of extension (a), weighting.
4. usefulness to man, relatedness to mannoneNo counterpart. It puts the dog in group 2. It is a closeness-to-the-speaker factor, which belongs with extension (b) (person, topicality), not with proto-role weight.
5. animation, capacity for movementmovementClose.

Result. Four of the five qualities are Dowty properties or degrees of them. The fifth (relatedness to man) is not, and has to sit in the extension that already holds person and topicality. This is a second source, and a native speaker, tying the ranking to intent, intelligence and movement, which supports extension (a). It also shows that the extension needs a non-Dowty factor, as the thesis page already says for person.

6. Results

6.1 An inanimate inverse subject exists: INV-CAP is withdrawn

SOURCED (c28) Tʼiis są́ biisxį́ ‘The tree died of old age’ is the preferred structure, and (c27), with yi-, is not preferred (p. 37). It is a bi- sentence whose subject (topic, first noun) is a tree. INV-CAP said bi- is defined only if its subject “could have controlled or resisted the event”. A tree cannot. Verdict: INV-CAP is falsified by a native speaker's judgment and is withdrawn (with INV-3′, which depended on it). This was the site's own hypothesis from this morning, not Grant's claim. It also counts against Hale's early table “inanimate–inanimate: blocked” (1973:303, as quoted by Witherspoon), if old age counts as inanimate. Hale's later ranking hypothesis (object outranks subject → inversion obligatory, 1973:304 as quoted) predicts (c28) correctly, because group 7 outranks group 8.

6.2 The rock/tree block without INV-CAP, and the equal-rank test revisited

SOURCED Creamer states the group 7 rule herself: the noun the verb says is moving “takes higher status than the more stationary object” (p. 36). So in (c23) ‘the rock rolled upon the tree’ the rock outranks the tree, and in (c25) ‘the tree fell upon the rock’ the tree outranks the rock. The two nouns are not of equal rank in these sentences. The mover is also the agent, so the ordinary ranking rule INV-3 (higher rank = subject) gives yi- and excludes bi-, just as it excludes *(c10) and *(c14). No capacity condition is needed.

Verdict on the equal-rank test, revised from SPLIT to SUPPORTS. Every equal-rank pair Creamer prints allows both voices: groups 1, 2, 3, 4 and 5 (pp. 33–36), and the newborn with the elephant, (c7)/(c8). The inanimate pairs that looked like exceptions are unequal once the verb's movement entailment is read, by Creamer's own rule. OPEN No source on the site gives two group 7 nouns where neither moves, or where both do (T20e). Those are the cases that would test “equal → free” for inanimates directly.

6.3 Old age: causation alone ranks lowest. This cuts against one version of the mapping

SOURCED In (c28) old age causes the tree's death, yet the tree outranks it. Group 8 is things “of which no action is seen other than effects” (p. 37). HYPOTHESIS In Dowty's terms, in this event old age has one Proto-Agent entailment (causation) and the tree has Proto-Patient ones (change of state, affected). If the rank were a count of the event's Proto-Agent entailments, old age would outrank the tree. It does not. Verdict: CUTS AGAINST a per-event count. It does not cut against the version the site has been using since Witherspoon: rank reads a participant's potential proto-agent properties (intent, intelligence, movement), weighted, with the event's entailments deciding only within group 7. A tree has independent physical existence and can move in other events; old age has neither. A possible repair inside Dowty's own list is his fifth property, existence independent of the event (1991:572): old age does not exist apart from its effects. That repair is HYPOTHESIS and untested.

6.4 Lightning's rank depends on the verb

SOURCED Lightning ranks with persons only “when used with a verb denoting its characteristically potent striking force” (p. 33), and with ‘strike’ both voices are fine, (c1)/(c2). HYPOTHESIS With other verbs it falls back to the natural forces (group 6). That would reconcile Creamer with Willie's judgment that ‘The lightning killed the horse’ with yi- is “an animacy violation” (1991:90, (73); inverse §2.4): Willie's verb is ‘kill’, not ‘strike’. The two authors may simply differ as speakers (Willie reports variation). Either way, a rank that changes with the verb is a rank read off the event, not a fixed property of the word iiʼniʼ. SUPPORTS D12.

6.5 The newborn and personified nouns: construal, not the word

SOURCED The same noun awééʼchíʼí ranks with the large animals while the baby is helpless, then as fully human once it is “capable of intent and movement” (p. 33): (c3)–(c8). Personified objects and animals take any rank their co-participant has (p. 37). This repeats Witherspoon's baby (pp. 79–80) and Sandoval & Jelinek's supernatural horse (1989:336) from a third source. HYPOTHESIS The properties that move the infant up are Dowty's volition and movement, nearly word for word. SUPPORTS T-ROLE: the rank follows what the participant is construed to be able to do, not the lexical item.

6.6 The yi-/bi- is sometimes on a postposition, written as a separate word

SOURCED Creamer writes yikʼi chʼínímááz ‘rolled upon’ as two words (pp. 36–37), and likewise yinahjįʼ ninígoʼ ‘came to rest against’ and yił łeeʼ dííłchʼil ‘struck’ (pp. 33, 36). In each, yi-/bi- is the first element of a postposition word (-kʼi ‘upon’, -nahjįʼ ‘against’, -ił ‘with’; meanings from her translations), and the ranking still applies. HYPOTHESIS This partly answers T20c. In these sentences the two ranked slots are the subject of the verb and the object of a postposition. The ranking is then computed over role slots in two words, at the clause level (rule set B), rather than inside one verb word (rule set A). Witherspoon's one-word spelling yikʼichʼínímááz is the opposite choice. Whether the postposition is inside the verb word is a spelling question that the two sources answer differently; the slots are role-bearing pronouns either way. OPEN A morpheme-by-morpheme analysis (Young & Morgan) is still needed.

7. Rules

Notation as on the inverse page and the who-acts-on-whom page. g(ρ) = Creamer group of ρ's referent as construed (1 highest, 8 lowest); r(ρ) higher when g(ρ) is smaller.

RANK-G     Groups.  r(ρ) is set by g(ρ), 1 > 2 > … > 8; same group = equal rank.  Basis: "capacity to act
           upon" (Creamer 1974:30, 32–37).
RANK-EV′   Group 7 by the event.  If g(ρ_1) = g(ρ_2) = 7, the participant the verb entails moves outranks
           the stationary one (p. 36).  Replaces RANK-EV, which only broke a tie after INV-CAP.
RANK-LT    Lightning.  g(lightning) = 1 with a verb of striking (p. 33); otherwise 6 (our default).
RANK-INF   Newborn.  g = 2 while helpless; g = 1 once capable of intent and movement (p. 33).
RANK-PERS  Personification.  A personified noun is equal in rank to its co-participant (p. 37).
INV-3      (restored as stated on the inverse page)  r(ρ_A) > r(ρ_X) → EXT = ρ_A (yi-);
           r(ρ_X) > r(ρ_A) → EXT = ρ_X (bi-);  r equal → free (topic = "object of major interest", p. 32).
INV-CAP    WITHDRAWN (counterexample (c28), p. 37).   INV-3′  WITHDRAWN (depended on INV-CAP).
RuleStatusBasis
RANK-GSOURCED (Creamer's groups); as proto-agent potential, HYPOTHESIS(c9)–(c20), (c27)–(c28); p. 37 “consistent with the examples shown” for groups further apart
RANK-EV′SOURCED (p. 36); as a verb entailment, HYPOTHESIS(c23)–(c26); Witherspoon (11)–(14)
RANK-LTSOURCED for ‘strike’; the default of 6 is HYPOTHESIS(c1)/(c2); Willie (73)
RANK-INF, RANK-PERSSOURCED(c3)–(c8); p. 37
INV-3SOURCED as a generalization (Willie (70); Creamer pp. 31–32)All pairs in §3

8. Derivations K1–K4, step by step

Conventions as on the derivations page. Σ schema step, A word-internal linearization (left-headed prefixal word), B clause level. V-H = in-situ licensing (Hale 2003:29); V-P = VPISH raising of EXT to Spec,M/SP inside the word. Verb words are not segmented. Entailment labels are ours, from the translations.

K1. (c28) Tʼiis są́ biisxį́ ‘The tree died of old age’ vs *(c27) Są́ tʼiis yiyiisxį́

  1. Σ Σ(√KILL.SG) = {ρA (caus; no vol, no sent, no mov), ρP (chg, aff)}, as on the kill (one) entry. ρA = old age, ρP = tree.
  2. Merge ρP in Comp,V; Merge v; Merge ρA in Spec,v. A head-final inside the word: ρP⌢V, v⌢VP.
  3. Σ π(ρA) = π(ρP) = 3: VOICE is defined (INV-0). RANK-G: g(tree) = 7, g(old age) = 8, so r(ρP) > r(ρA). INV-3: EXT = ρP, VOICE = bi-. (Withdrawn INV-CAP would have blocked this step, since a tree cannot resist. That wrong prediction is why it is withdrawn.)
  4. Agree M/S ⟶ EXT = ρP. Move V-P HYPOTHESIS: ρP → Spec,M/SP, crossing ⟨ρA⟩ in Spec,v. V-H: none.
  5. Spell-out A biisxį́.
  6. B INV-6: są́, next to the verb, links to the internal slot ρA; tʼiis, outer, links to EXT = ρP. ✓ (c28).
  7. *(c27): yi- sets EXT = ρA (old age) against INV-3 at the Σ step. ✗ Same failure as *(c10) and *(c19).

K2. (c23) Tsé tʼiis yikʼi chʼínímááz vs *(c24) Tʼiis tsé bikʼi chʼínímááz, without INV-CAP

  1. Σ Two role slots, in Creamer's spelling in two words: ρA = subject of chʼínímááz (mov, caus of contact; no vol), ρX = object of the postposition -kʼi ‘upon’ (static). HYPOTHESIS the two-word analysis follows the spelling (§6.6).
  2. Merge ρX as Comp,P; A inside the postposition word: pronoun ⌢ P (yi-/bi- + -kʼi). Merge PP as an adjunct of the verb word's projection; ρA in the verb's subject slot.
  3. Σ RANK-G: g = 7 for both. RANK-EV′: the verb entails that ρA (the rock) moves, so r(ρA) > r(ρX). INV-3: EXT = ρA, so the postposition takes yi-.
  4. Agree M/S ⟶ EXT = ρA. Move V-P: ρA → Spec,M/SP (nothing crossed). V-H: none.
  5. Spell-out A yikʼi and chʼínímááz; B PP ⌢ verb (head-final at clause level).
  6. B INV-6: tʼiis, inner, links to ρX; tsé, outer, to ρA. ✓ (c23).
  7. *(c24): bi- sets EXT = ρX, the lower-ranked tree, against INV-3. ✗ The same rule that excludes *(c14). (c25)/*(c26) go the same way with the tree as mover and yikʼiikę́ę́z as one word (C2 there, with RANK-EV′ in place of INV-CAP).

K3. (c20) Wóláchííʼ tó biisxį́ ‘The ant was killed by the flood’ vs *(c19)

  1. Σ Σ(√KILL.SG) as in K1, with ρA = flood (caus, mov; no vol, no sent), ρP = ant.
  2. Merge as in K1. Σ RANK-G: g(ant) = 5 > g(flood) = 6. INV-3: EXT = ρP, bi-.
  3. Agree, Move V-P, Spell-out A biisxį́; B tó (inner) → ρA, wóláchííʼ (outer) → EXT. ✓
  4. What K3 shows. An insect outranks a flood. HYPOTHESIS On the mapping, minimal sentience outweighs causation plus movement. That is extension (a), weighting, from a second source (Witherspoon's girl/water (9)/(10) was the first).

K4. (c7)/(c8) elephant and newborn, both voices; *(c3) vs (c4)

  1. Σ Σ(√CRUSH) = {ρA (caus, mov), ρP (chg, aff)}; Σ(√KICK) = {ρA (vol, caus, mov), ρP (aff)}.
  2. Σ RANK-INF: the newborn has g = 2, the elephant g = 2. Equal: INV-3 leaves voice free. (c7) chooses yi-, (c8) chooses bi-, with the baby as topic. Both ✓.
  3. Σ (c3)/(c4): baby g = 2, man g = 1. INV-3: the man must be EXT. *(c3) makes the baby EXT with yi- ✗; (c4) bi- ✓. *(c6) makes the baby EXT with bi- ✗; (c5) yi- ✓.
  4. Agree, Move, Spell-out A as in inverse I1/I3; B INV-6.
  5. What K4 shows. The noun awééʼchíʼí is the same in every line. The rank computed at the Σ step is that of a helpless participant, and Creamer says it changes as soon as the child is capable of intent and movement.

VPISH. No Creamer datum tests V-P against V-H. In K1–K4 the rank and the voice are fixed at the Σ step, before any noun is adjoined, as on the who-acts-on-whom page. K2 adds one thing: if Creamer's spelling is taken at face value, the ranked slots can sit in two different words (a postposition and a verb), so the ranking step must be able to see across a word boundary, at the level where rule set B applies.

9. Verdict for the thesis

ClaimEffect of Creamer 1974Verdict
INV-CAP (site hypothesis, 4:46am today)(c28) Tʼiis są́ biisxį́, preferred, has a tree as bi- subjectFALSIFIED; withdrawn
Equal rank → either voice (topic decides)Holds for every equal-rank pair printed (groups 1–5, newborn/elephant). The rock/tree pairs are unequal by Creamer's movement ruleSUPPORTS (revised from SPLIT). Untested for two inanimates where neither or both move (T20e)
Extension (a): weighted proto-agent potentialNative-speaker source names intent, intelligence, potency and movement (p. 30); an ant outranks a flood (c20)SUPPORTS, now from two sources
Rank as a count of the event's proto-agent entailmentsOld age causes the tree's death but ranks below it (c28)CUTS AGAINST that version. The capacity version survives. Possible repair via Dowty's independent existence (HYPOTHESIS)
The ranking is pure Dowty“Usefulness to man, or relatedness to man” (p. 30) has no Dowty counterpartNeeds extension (b), as already stated for person; not a departure from Dowty, but not covered by it
The inverse ranks roles, not words (T-ROLE, D12)Group 7 rank set by the verb's movement (p. 36); lightning's rank set by the verb (p. 33); newborn and personified nouns re-ranked by construal (pp. 33, 37)SUPPORTS, third source. Creamer treats it as a rule of grammar (a syntactic inversion governed by rank), which weakens Witherspoon's “cultural rules” caution. Caution: she also says the verb's role is “secondary” (p. 29); the main factor is the participant's inherent capacity
T-STRONG / T-MID / T-WEAKNot discussedNEUTRAL. T-STRONG still fails (thesis page)

10. Open

#QuestionWhat would settle it
T21aTwo group 7 nouns where neither moves (or both do): either voice, or no inverse?A sourced example; speaker judgments (= T20e)
T21bWhy does old age rank below a tree? Independent existence, absence of visible action, or something else?Group 8 nouns with other group 7 and group 6 nouns (hunger, disease); Young & Morgan
T21cIs lightning's high rank tied to ‘strike’ only? Does it reconcile Creamer with Willie (73)?Lightning with ‘kill’, ‘burn’ and ‘strike’, judged by the same speakers
T21dIs the yi-/bi- on -kʼi, -nahjįʼ, -ił the same alternation as on transitive verbs, and is the postposition inside the verb word (Witherspoon) or not (Creamer)?Young & Morgan 1987; a morpheme-by-morpheme gloss (extends T20c)
T21eDoes relatedness to man (the dog in group 2) behave like person and topicality (extension (b)) or like proto-agent weight?Dog vs another group 3-sized animal with the same verb; discourse tests

The inverse: when a lower argument outranks the agent

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.

Working note. Kept as written (URL, data and status labels unchanged). Filed under Ranking, animacy and yi-/bi-; the consolidated argument is the essay who-acts-on-whom. Rule IDs: rules index.

Verbs on this page → dictionary: yiztał ‘kick’ · yiztsʼǫs ‘kiss’ · yishxash ‘bite’ · yiyiisxį́ ‘kill (one)’ · dabiiłtsą́ ‘see’ · yoolóós ‘lead (one animate)’ · yáłtiʼ ‘speak, talk (to)’

Restated on the unified structure (2026-11-15). The inverse is restated as one extra Move on the clause spine: the patient moves (via an outer Spec,vP) above the agent, which has already raised to Spec,TP by the working default (H3, H6′, H21b). The event LF is the same as the direct; the trigger and landing site are OPEN. 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. 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, §4, inverse row. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

With two third persons, a Navajo transitive verb comes in two shapes: a yi- form (‘the horse kicked the man’) and a bi- form (‘the man got kicked by the horse’). Who did what to whom is the same in both. What changes is which participant is realized as the subject. This page states the choice as rules over proto-role entailments plus an animacy ranking, asks where in the derivation the choice is made, and derives one sourced pair step by step. This is open question T5. The other response to the same ranking clash — take the human off the scale and keep the direct form — is the fourth-person page (T8).

On this page: 1 Question · 2 Data · 3 Rules · 4 Which orders are impossible · 5 Where the choice is made · 6 Derivations I1–I5 · 7 Verdict · 8 Open

1. The question

GRANT'S CONJECTURE The project is argument ranking. The natural alignment is that the higher proto-agent argument is the subject. The edge case is when a lower argument outranks it. Then the passive-like expression (the Navajo bi- form) is the natural one, and some permutations are impossible. In Navajo the units being ranked and moved are theta roles, not lexical items. Dowty's proto-role entailments are fixed by the event. The inverse does not change them; it changes only which argument is realized where.

Open question T5, as posed on the open-questions page: does the inverse change the agent's position (Willie's "internal (Ergative) agent"), or only how the nominals are linked (Willie & Jelinek's 3ACC-3NOM gloss, which is the same in both voices)? The earlier pages state the yi-/bi- rule over the topic DP (schema §4, D4). This page reads Willie 1991 §3.3 in full and Sandoval & Jelinek 1989 against that rule.

2. Data SOURCED

Sources opened this run (2026-10-08). (a) Willie, MaryAnn 1991, Navajo Pronouns and Obviation, PhD dissertation, University of Arizona (UMI 9210298 scan). Page images of printed pp. 16, 27–28, 71–77, 86–88, 90–92, 94–95, 97 and 100 were read at 110 dpi. The Navajo lines of (37), (73)–(76) and (79)–(80) were re-read at 300 dpi. The typescript prints nasal hooks as cedilla-like marks; they are transcribed as hooks here. (b) Sandoval, Merton & Eloise Jelinek 1989, "The bi-construction and pronominal arguments in Apachean," in Rice & Cook (eds.), Athapaskan Linguistics, Mouton, pp. 335–377. This was read as page images (the scan has no text layer) from Carnie & Harley's 2012 compilation of Jelinek's papers. Printed pp. 335–361 and 368–371 were read at 90 dpi, and the Navajo examples on pp. 335–336 at 300 dpi. Most of their examples are Jicarilla Apache; only those they mark as Navajo are printed below. (c) Willie & Jelinek 2000, printed p. 268 at 300 dpi and p. 273 at 90 dpi (same compilation).

Glosses are copied exactly as printed. Willie glosses the verb as a whole (3.3.kicked, INV3.3.kicked). Sandoval & Jelinek separate only yi-/bi- (yi-kicked). No further segmentation is offered here.

2.1 The basic pair, and the edge case

Willie 1991:71, (37)
a.łį́į́ʼdzaanéézyiztał
horsemule3.3.kicked
‘The horse kicked the mule.’
b.dzaanéézłį́į́ʼbiztał
mulehorseINV3.3.kicked
‘The mule was kicked by the horse.’

Willie, pp. 71–72: "In (37a) the NP occurring immediately before the yi-form is a patient adjunct, and by default, the other NP is an adjunct to the agent argument. In (37b) the NP occurring nearest the bi-form is the agent adjunt [sic] of that sentence … Therefore, the immediately preceding NP, in both voice forms, is an adjunct to the "internal" argument, regardless of whether the theta role of that argument is agent or patient." Horse and mule are "of equal animacy", so here "the voice alternation is optional, that is, discourse determined" (p. 72).

Sandoval & Jelinek 1989:335–337, (1), (2), (5), (6), (7) (all marked as Navajo)
(1)łį́į́ʼdzaaneezyiztał‘The horse kicked the mule.’
horsemuleyi-kicked
(2)łį́į́ʼdzaaneezbiztał‘The horse got kicked by the mule.’
horsemulebi-kicked
(5) ??łį́į́ʼhastiinyiztał?? ‘The horse kicked the man.’
horsemanyi-kicked
(6)hastiinłį́į́ʼbiztał‘The man got kicked by the horse.’
manhorsebi-kicked
(7)łį́į́ʼshiztał‘The horse kicked me.’
horseme-kicked

On (5), p. 336: it "is culturally "odd" and not employed, since it carries the implication that the horse has more volition and intelligence than the man. If the horse is understood as a supernatural being, with greater than human powers, then the sentence is acceptable; this fact demonstrates that the sentence is not ungrammatical, but pragmatically unacceptable." On (7), p. 337: "first and second person do not participate in the animacy hierarchy." Spelling: Sandoval & Jelinek print dzaaneez with no accents, and Willie prints dzaanééz. Both are recorded as printed. The two sources order the words of the bi- sentence differently: S&J (2) has horse–mule, Willie (37b) has mule–horse. Both follow the same rule, because in each one the first noun is the one kicked.

2.2 With one nominal or none

Willie 1991:73, (38)–(39)
(38a)dzaanéézyiztał‘It kicked the mule.’
mule3.3.kicked
(38b)łį́į́ʼbiztał‘It was kicked by the horse.’
horseINV3.3.kicked
(39a)yiztał‘It kicked it.’
3.3.kicked
(39b)biztał‘It was kicked by it.’
INV3.3.kicked

p. 73: "When only one adjunct is present, it must be coindexed with the internal argument, and there is no adjunct to the external argument." And: "When no NPs are adjoined, the verb-sentences in the Direct and Inverse forms still differ in meaning." Sandoval & Jelinek print the same bare pair, (9a) yiztał yi-kicked ‘He kicked him.’ and (9b) biztał bi-kicked ‘He got kicked by him.’ (p. 338).

2.3 The ranking, and the stars it produces with no nominals at all

Willie 1991:86, (70): "If the agent is higher on the animacy scale than the patient, then the Direct voice is used: if the patient is higher on the scale, then the Inverse must be used. And if they are equal, the choice rests on discourse factors." Same page: "even when the sentence does not contain any nominals, the sentence must be interpreted with reference to the animacy scale."

Willie 1991:87, (71)–(72) (her ranked interpretations, "in order of preference")
(72a) yishxash 3.3.bither label(72b) bishxash INV3.3.bither label
1)he bit him.equal rankedhe was bitten by him.equal ranked
2)he bit it.pat. lower rankedhe was bitten by it.patient higher
3)it bit it.lower equalsit was bitten by it.lower equals
4)*it bit him.agent lower*it was bitten by him.higher agent

pp. 87–88: "what is ruled out (72a.4) is a lower-ranked agent in the yi- verb, and what is ruled out in (72b.4) is the higher agent in the bi- verb … Each paradigm permits what the other excludes." Also p. 88: "on the preferred interpretation of these sentences, for both the Direct and Inverse voices, the internal argument is taken to be a lower or indefinite being."

2.4 Speaker variation, and fronting under the inverse

Willie 1991:90–92, (73)–(76)
(73a)ʼiiʼniłį́į́ʼyiyiisxį́‘The lightning killed the horse.’
lightninghorse3.3.killed sg.
(73b)łį́į́ʼʼiiʼniʼbiisxį́‘The horse was killed by lightning.’
(74a)iiʼniʼłį́į́ʼbiisxį́‘The lightning killed the horse.’
(74b)lííʼiiʼniʼbiisxí‘The lightning killed the horse.’
(74c)*iiʼniʼłííʼ łaʼbiisxį́
(75a)łííʼiiʼniʼyiyiisxį́‘The horse killed the lightning.’
(75b)iiʼniʼłííʼyiyiisxį́‘The lightning killed the horse.’
(76a)*iiʼniʼishłííʼbiisxį́
(76b)łííʼ-íshiiʼniʼbiisxį́‘Did the lightning kill the horse?’

Only (73a) has a gloss line in the source. Recorded as printed, not repaired: ‘lightning’ appears as ʼiiʼni (73a), ʼiiʼniʼ (73b) and iiʼniʼ (74)–(76). ‘Horse’ appears as łį́į́ʼ, łííʼ (no hooks) and, in (74b), lííʼ with a plain l. The verb in (74b) is biisxí with no hook. The translation printed under (76b) is the only one given for (76).

Willie's comments. On (73a), p. 90: "Actually I find (73a) objectionable, as an animacy violation, and (73b) more preferable." On (74), p. 90: "you can change the order of the nounphrases in the bi- form but not with yi- form -- and not change the meaning", and (74c) is out because "the external argument in a bi- construction must be definite". On (75), p. 91: "when the Inverse is used, the speaker presupposes that the agent is lower on the animacy scale … Therefore, it is possible to put the highly salient nominal 'ii'ni' in sentence initial position, by focus movement. Whereas, when the speaker hears a Direct construction, he does not presuppose a lower ranked agent; and Ex. (75a), though extremely odd pragmatically, can have only the interpretation shown." On (76), p. 92: "When there is focus movement, as in (76a), this clitic does not attach to the moved element." On variation, p. 16: Young (1977; 1991) "found only about half of the speakers obeying the animacy hierarchy" in an informal survey (secondhand through Willie; Young not opened).

2.5 Goals and benefactives can invert; the theme cannot

Willie 1991:94–95, (79)–(80), (85)–(86)
(79)ʼashkiiʼatʼéédyichʼįʼyáłtiʼ
boygirl3.to3.speaking
‘The boy is speaking to the girl.’
(80)ʼashkiiʼatʼéédbichʼįʼyáłtiʼ
boygirl3.to3.speaking
‘The boy is being spoken to by the girl.’
(85)ʼashkiiʼatʼéédłį́į́ʼyáyoolóós
boygirlhorse3.for3.3.lead [sing. +animate]
‘The boy is leading the horse for the girl.’
(86)ʼashkiiʼatʼéédłį́į́ʼbáyoolóós
boygirlhorse3.for3.3.lead [sing. +animate]
‘The girl is leading the horse for the boy.’

The bi-forms bichʼįʼ (80) and bá (86) are glossed "3.to" and "3.for", with no INV, as printed. Willie, p. 94: "with an Inverse construction, the agent is always an internal argument; and the external argument is always an argument that has an "internal" theta role in the corresponding Direct construction." p. 95: "Note that the theme argument does not take part in the inversion process." p. 97: the Inverse "always makes the argument with an agent theta role the internal argument." Sandoval & Jelinek's Jicarilla rule (55b), p. 359, says the same for three-argument verbs: "Reverse coindexing of (1) and (2) [agent, goal]; coindexing in (3) [theme] remains unchanged."

2.6 Definiteness, case and topic

  • SOURCED Willie 1991:76–77: (45) ʼashkii ʼatʼééd biztsʼǫs ‘The boy was kissed by the girl.’ vs (47) * ʼashkii léiʼ ʼatʼééd biztsʼǫs [A boy was kissed by the girl.]. Her conclusion: "with the Inverse, the external argument (the patient, with Absolutive case) must be definite." In the direct voice, either argument may be indefinite ((42)–(44), p. 76).
  • SOURCED Case. On p. 72: "Case is assigned to the pronouns before they are incorporated into the verb. In the Inverse, where the agent is the "internal" argument, i.e., not the subject, I assume that the agent pronoun has Ergative case, and the patient pronoun has Accusative case (Marantz 1984)." Discrepancy in the source: pp. 77 and 100 instead give the inverse patient Absolutive case. Both are recorded; neither is chosen here.
  • SOURCED Willie 1991:100 (summary): "The Inverse makes the agent an internal argument, with Ergative case, and places an argument with an "internal" theta role in the external argument position, with Absolutive case. With simple transitives, the Inverse subject is the patient … The subject of an Inverse construction is definite, and becomes the discourse topic." Sandoval & Jelinek 1989:339 (c): "the yi-/bi- alternation involves a change in the grammatical relations of the pronominal prefixes, such that a pronominal argument with the θ-role of Patient or Goal has the grammatical relation of SUBJECT in the bi- construction"; (d): it is "a transitive, inverse construction, and therefore not a passive".
  • SOURCED Clause positions. Willie & Jelinek 2000:268: "The position immediately preceding the verb is the focus position … By default, an "outer" nominal must be coindexed with the pronoun that is not coindexed with the inner nominal; it is a "topic-adjunct"". Their (46) is ʼashkii łį́į́ʼ yiztał, boy horse 3ACC-3NOM-kicked (Direct), TOP-ADJ FOCUS Focus-Topic V, ‘The boy, he kicked [the horse]F.’ On p. 273: "The bi-pronoun marks a topic/focus inversion that is incompatible with focus movement of the Wh-word."
  • SOURCED The prefixes. Willie 1991:27: "The subject is the innermost (rightmost) argument, closest to the verb stem, while the object is to the left of (external to) the subject." p. 28: "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".

3. The rules

Notation as on the schema page. Σ = {ρA, ρX}. ρA is the slot with the most Proto-Agent entailments (C5). ρX is the competing slot: a patient, or the object of a "grammatical" postposition (goal, benefactive). r(ρ) is the rank of the slot's referent on the animacy scale (Creamer's eight groups; Willie's "animacy scale"). EXT(Σ) is the slot realized as the external argument, i.e. the subject.

NAT    Natural alignment.  EXT = ρ_A.  The slot with the most Proto-Agent entailments is the subject
       (Dowty's ASP, 1991:576 (31)), spelled out yi-.
EDGE   If r(ρ_X) > r(ρ_A), then EXT = ρ_X.  ρ_A becomes internal, and the verb is spelled out bi-.

INV-0  Domain.   VOICE is defined iff every slot in the competition is 3rd person, and Σ has ρ_A and some ρ_X.
INV-1  Fixity.   E_ρ is the same under yi- and bi- for every ρ ∈ Σ.                               [= C2]
INV-2  Exponence.  yi-  iff  EXT = ρ_A ;   bi-  iff  EXT = ρ_X  (and then pos ρ_A = internal).
INV-3  Ranking.  EXT = the slot whose referent ranks higher:
         r(ρ_A) > r(ρ_X) → EXT = ρ_A          r(ρ_X) > r(ρ_A) → EXT = ρ_X
         r(ρ_A) = r(ρ_X) → free (discourse decides)
INV-4  Definiteness.  If EXT = ρ_X, the referent of ρ_X is definite.                               [= C7]
INV-5  Candidates.  Only ρ_A and a slot with an "internal" role (patient, goal, benefactive) compete;
       a theme in a three-slot frame never does.
INV-6  Linking (rule set B).  The nominal immediately before the verb word links to the internal slot.
       A lone nominal links to the internal slot.  An outer nominal links to EXT (topic adjunct).
INV-7  Fronting (rule set B, optional).  Under bi-, the internal-slot nominal may be focus-fronted over
       the EXT nominal, with no change of roles.  The fronted nominal cannot host -ísh.
RuleStatusBasis
NATHYPOTHESIS (the ASP is SOURCED; that it is the Navajo default is ours)Dowty 1991:576. The direct voice is the one used when the agent is higher (Willie (70)).
EDGEGRANT'S CONJECTURE, matched by SOURCED (70)Grant's "a lower argument outranks it". Willie: "if the patient is higher on the scale, then the Inverse must be used" (p. 86).
INV-0SOURCEDWillie p. 86 ("only … if all arguments are third person"); S&J (7), p. 337; W&J 2000:264 (C4).
INV-1SOURCED in effectWillie p. 71: the inverse "is a transitive sentence that has both agent and patient arguments"; the agent label survives the inversion (pp. 94, 97). S&J (d). The translations of (5)/(6) and (37a)/(37b) describe the same event.
INV-2SOURCED (restated)Willie pp. 72, 94, 97, 100; S&J p. 339 (c). The formal statement is ours.
INV-3SOURCED as a generalization; its grammatical status is OPENWillie (70) "must". S&J on (5): "not ungrammatical, but pragmatically unacceptable". Willie (73): speakers vary. Young's survey: about half obey it (via Willie p. 16). Qualified 2026-10-14: the “equal → free” clause fails for two inanimates. Rock and tree (both Creamer group 7) allow no bi- in either order (Witherspoon 1977:66, (11)–(14)). Restated as INV-3′ with INV-CAP on the who acts on whom page. Restored 12:46pm: INV-CAP is falsified (Creamer 1974:37), and Creamer's movement rule makes the rock and tree unequal (1974:36), so INV-3 stands as stated here (Eight ranks §6.2).
INV-4SOURCEDWillie pp. 77, 88; (47), (74c).
INV-5SOURCEDWillie pp. 94–95, (79)–(80), (85)–(86); S&J (55b) for Jicarilla. OPEN Why the theme cannot compete (T5d).
INV-6SOURCEDWillie pp. 72–73, (37)–(38); W&J 2000:268, (46).
INV-7SOURCED (Willie); the link to W&J is HYPOTHESISWillie pp. 90–92, (74), (76). See §5 and T5e for how this squares with W&J's "incompatible with focus movement of the Wh-word".

The cleanest restatement HYPOTHESIS. INV-3 is a rule about positions, not about roles: the external argument is the higher-ranked participant, and the internal argument is the lower one. That is Willie's p. 88 remark ("for both the Direct and Inverse voices, the internal argument is taken to be a lower or indefinite being"). The roles enter only through INV-2. bi- marks that the higher-ranked, external participant is not the one with the most Proto-Agent entailments. So the voice prefix reports a comparison between two orders over the same two slots. One is the entailment order (fixed by the event). The other is the rank order (fixed by who the participants are). This is the corollary on schema §4, bi- ⟺ ρrank ≠ ρA, now restated over EXT instead of over the topic DP.

4. Which orders are impossible

For a two-slot verb with two third persons there are three rank relations and two voices. SOURCED cells give the source's judgment. The schematic cells are labelled schematic, because no source prints that Navajo string.

Rank relationyi- (EXT = agent)bi- (EXT = non-agent)
agent > patient (e.g. horse kicks rock)✓ (72a.2); W&J (61a)✗ (72b.4) "*it was bitten by him"; W&J (61b) ?*
agent = patient (horse, mule)✓ (37a), (72a.1)✓ (37b), (72b.1); the choice "rests on discourse factors"
patient > agent (horse kicks man)✗ (72a.4) "*it bit him"; S&J (5) ??; Willie (20b) *; W&J (62a) ?*✓ (6), (72b.2), (20c), (62b)

HYPOTHESIS Two of the six cells are excluded, and they are the two where EXT is the lower-ranked participant. Add the order of the two nominals and there are 3 × 2 × 2 = 12 strings. INV-6 assigns each string exactly one reading: the nominal next to the verb is the internal participant. So the order of the nominals adds no new impossible cells. It only decides which noun is which slot. The one extra freedom is INV-7: under bi- a focus-fronted internal nominal can come first, as in (74a), and the reading is unchanged. Under yi- the reverse order is read literally, as in (75a) ‘The horse killed the lightning’. Schematically:

  [NP_high]  [NP_low]  yi-V    NP_low = internal = patient   →  high one is agent      ✓
  [NP_low]   [NP_high] yi-V    NP_high = internal = patient  →  low one is agent       ✗ (72a.4) / ?? (5)
  [NP_high]  [NP_low]  bi-V    NP_low = internal = agent     →  high one is patient    ✓ (6)
  [NP_low]   [NP_high] bi-V    NP_high = internal = agent    →  low one is patient     ✗ (72b.4)
  [NP_low]F  [NP_high] bi-V    (fronted from  [NP_high] [NP_low]F bi-V)                ✓ (74a); only under bi-
  (schematic; the sourced instances are cited in the right column)

Grant's edge, stated exactly. HYPOTHESIS The natural alignment (NAT) is overridden in one configuration only: patient > agent in rank. In that configuration the bi- form is not just allowed but the expected one (Willie's "must"; S&J's "the bi-construction is employed"), and the NAT form is excluded or marked. What is never possible is a lower-ranked participant as the external argument. The entailments are the same in every cell (INV-1). Only EXT moves.

5. Where in the derivation the choice is made

5.1 Inside the word, not in the clause HYPOTHESIS (from SOURCED facts)

  1. The choice is made with no nominals at all. (39a) yiztał ‘It kicked it.’ and (39b) biztał ‘It was kicked by it.’ "still differ in meaning". (72a.4) and (72b.4) are starred with no noun in the sentence. So the voice cannot be decided by rule set B, because B (topic/focus positions, linking of nominals) has nothing to work on here. It is fixed at Σ, by the slots and their referents, and spelled out inside the word by rule set A, as yi- or bi- to the left of the subject marker (Willie p. 27).
  2. Willie's own model puts it there. "Case is assigned to the pronouns before they are incorporated into the verb" (p. 72). The inverse "has scope over the entire MVS" (p. 100), her Maximal Verb Sentence.
  3. Rule set B only reads the result. INV-6 links nominals to slots once EXT is known. INV-7 (focus fronting) is possible only because bi- has already fixed which slot is the agent (Willie p. 91). B can be blocked by the voice, as in (76a) -ísh and W&J's wh-facts, but it never chooses the voice.
  4. Correction to the earlier pages. Schema §4 and D3/D4 define VOICE by the slot of the clause-initial topic DP (a rule-set-B notion). Because of (39) and (72), that definition is too late in the derivation. The rule should be stated over EXT (INV-2, INV-3), and the topic facts should follow from INV-6. In sentences that have two nominals the two statements give the same results, so no earlier derivation changes its output.

5.2 What VPISH predicts

Update 2026-10-16 (8:46pm): restated as probe/Agree + EPP, with Hale 1997's own two-step EPP analysis and Bruening 2005's Voice-[EPP] mechanism, on the inverse as Agree and EPP (T27k). Four options are tested there (IA-0 linking, IA-1* raising in the inverse only, IA-2 Hale 1997, IA-3 in situ). da- plus the word's spell-out order (CYC) constrains raising, and the theme exclusion (INV-5) follows from closeness under raising if the PP is above the theme.

  • HYPOTHESIS Under V-P (VPISH raising), EXT is simply the slot that raises to Spec,M/SP. In the direct voice ρA raises from Spec,v and nothing crosses. In the inverse, ρX raises from Comp,V (or from inside a grammatical PP) past ρA, which stays in Spec,v. That is Willie's "internal (Ergative)" agent. So the inverse is the one construction on this site where V-P gives the first Spec-to-Spec movement to a non-agent. The "second type" of movement that Grant allows for arguments beyond the first becomes, under the inverse, the movement to subject.
  • HYPOTHESIS Under V-H (Hale 2003, licensing in situ), nothing moves. EXT is a case designation on the pronouns (Ergative agent, Absolutive or Accusative patient; Willie pp. 72, 77). The two variants give the same string, as in D1–D5. They differ in structure: V-P makes ρX c-command ρA in the inverse; V-H does not.
  • OPEN So the inverse is where VPISH makes its sharpest testable prediction: scope or binding of the patient over the agent under bi- only. Willie's quantifier pairs (58)–(60), pp. 79–80, are the place to start. Update 2026-10-09: they are now page-checked and printed on the number & scope page. They do not settle T5c, because Willie does not say which scope readings (58) allows (T10a). That page also finds that the distributive da- counts the agent in both voices. This constrains V-P: Pl must Agree with ρA before the patient raises, or else Pl must be role-indexed (§5.1 there). It also bears on step 6 of I1 below.
  • GRANT'S CONJECTURE + HYPOTHESIS If V-P is right, what moves in the inverse is a role-slot (ρX), not a word: the nominals are adjuncts and stay where rule set B puts them (INV-6). The S&J horse case points the same way. The same word łį́į́ʼ in (5) gives ?? for an ordinary horse and ✓ for "a supernatural being, with greater than human powers". So the rank belongs to the participant as construed, not to the lexical item.

5.3 How INV-7 fits W&J's wh-facts HYPOTHESIS

W&J 2000:273 say the bi-form is "incompatible with focus movement of the Wh-word": in (60b) háí-sh Jáan biztsʼǫs, reading 2) *‘John was kissed by whom/which one?’ is out (D7). Willie (74a) allows focus fronting of the agent nominal under bi-. The two are consistent if what blocks (60b-2) is the question clitic: a focus-moved element cannot host -ísh (76a), and the wh-word in (60b) carries -sh. Then the rule is "under bi-, a fronted internal nominal cannot carry the question clitic". It is not "under bi-, nothing fronts". This also refines prediction (b) on the open-questions page (multiple-wh). OPEN (T5e).

Update 2026-10-20, 8:46pm (T43f). Willie p. 92 and pp. 210–212 and W&J pp. 272–273 were re-read on page images (the two question endings compared). One statement (QC-5, HYPOTHESIS) now covers *(76a), *(60b) reading 2 and Willie *(17) Díí=sh háí biztał: under bi- a question ending may sit only on a host in its unmoved position. *(17) involves no fronting, so the rule above (“a fronted internal nominal cannot carry the question clitic”) is too narrow. (76b) łííʼ-ísh and W&J (59) háí-sh Jáan biztsʼǫs show that the inverse subject itself can carry either ending, so the ban is not on the landing site. T5e is partly answered. Whether a fronted wh-agent with no -sh is possible is still OPEN.

6. Derivations I1–I5, step by step

Conventions as on the derivations page. Tags: Σ schema step, A word-internal linearization, B clause level. V-H = in-situ licensing (Hale 2003:29). V-P = VPISH raising. Verb words are not segmented beyond the sources' glosses. Entailment assignments are ours, from the English translations. They match D3: for ‘kick’, EρA ⊇ {vol, caus, mov} and EρP ⊇ {aff} (HYPOTHESIS).

I1. S&J (6) hastiin łį́į́ʼ biztał ‘The man got kicked by the horse.’ (edge case, inverse)

  1. Σ Σ(√KICK) = {ρA, ρP}. posρA = Spec,v; posρP = Comp,V (Hale 2003 tree (8)). C5: ρA has three P-A entailments and ρP has none, so ρA is the agent slot. These entailments are fixed here and are never touched again (INV-1).
  2. Merge V (the ‘kick’ stem; not segmented) with ρP. A lexical V is complement-final: ρP⌢V.
  3. Merge v (Voice/classifier; its value is not given by the sources). A v⌢VP. Merge ρA in Spec,v. φ unvalued.
  4. Merge M/S. A M/S⌢vP.
  5. Σ π-valuation: π(ρA) = π(ρP) = 3, so INV-0 holds and VOICE is defined. Rank: the referent of ρP is a man and the referent of ρA is an ordinary horse, so r(ρP) > r(ρA). INV-3 gives EXT = ρP, and INV-2 gives VOICE = bi-. No nominal has been used yet.
  6. Agree M/S ⟶ EXT. (Update 2026-10-16, 8:46pm: superseded under the raising version IA-1* on the inverse as Agree and EPP. There the patient slot moves first and M/S then agrees with ρA left below it (Bruening 2005 n. 22), so M/S agrees with ρA in both voices, as W&J's 3NOM gloss has it.) (Caveat 2026-10-09: if da- number agreement rides on M/S, M/S must agree with ρA, not EXT, in the inverse; see number & scope §5.1, T10b.) In Willie's terms, the external argument gets Absolutive case (p. 100; Accusative on p. 72) and the internal agent gets Ergative case.
  7. Move. V-P HYPOTHESIS: ρP goes from Comp,V to Spec,M/SP, crossing ⟨ρA⟩, which stays in Spec,v as the "internal (Ergative)" agent. V-H: no movement.
  8. Spell-out A. The position-IV exponent is bi-, to the left of the (null) 3rd-person subject marker and the stem. The word is biztał, glossed "bi-kicked" by S&J. Which slot bi- itself indexes, the external patient or the internal agent, is OPEN (T5a).
  9. B Linking (INV-6). łį́į́ʼ, immediately before the verb, links to the internal slot, which under bi- is ρA. hastiin, the outer nominal, links to EXT = ρP as the topic adjunct (W&J 2000:268). INV-4: hastiin is unmarked, so its preferred reading is definite (Willie p. 76) ✓. Order: DPEXT ⌢ DPint ⌢ V.

I2. S&J (5) ?? łį́į́ʼ hastiin yiztał ?? ‘The horse kicked the man.’ (the natural alignment, marked)

  1. Σ Same Σ, same entailments, same referents as I1 (INV-1). ρA = the horse, ρP = the man.
  2. Merge as in I1, steps 2–4.
  3. Σ NAT: EXT = ρA, so VOICE = yi-. INV-3: r(ρP) > r(ρA) requires EXT = ρP. Conflict: NAT and INV-3 pick different slots, and INV-3 is violated.
  4. Agree, Move (V-P: ρA to Spec,M/SP, nothing crosses), and Spell-out A yiztał all go through. No step of the word fails.
  5. B INV-6: hastiin, next to the verb, links to the internal slot = ρP; łį́į́ʼ links to EXT = ρA. The reading is coherent (the man is kicked), so linking does not fail either.
  6. Result. The only violation is INV-3, at the Σ step, and S&J mark it ??, not *. If the horse "is understood as a supernatural being", then r(ρA) > r(ρP), INV-3 is satisfied, and the same string is acceptable (S&J p. 336). HYPOTHESIS INV-3 reads the rank of the participant as construed in context. It does not read a fixed property of the word łį́į́ʼ.

I3. Willie (37a)/(37b): equal rank, both voices

  1. Σ Σ(√KICK) as in I1. ρA = the horse, ρP = the mule. r(ρA) = r(ρP): "of equal animacy" (Willie p. 72).
  2. Σ INV-3 leaves EXT free, and discourse chooses. (37a): EXT = ρA → yi-. (37b): EXT = ρP → bi-.
  3. Move (V-P). In (37a) ρA raises. In (37b) ρP raises past ⟨ρA⟩. V-H: none.
  4. Spell-out A yiztał / biztał.
  5. B INV-6. (37a) dzaanééz (next to V) → internal = ρP; łį́į́ʼ → EXT = ρA. (37b) łį́į́ʼ (next to V) → internal = ρA; dzaanééz → EXT = ρP. In both, the horse is the kicker.
  6. One-nominal variants (38): the lone nominal links to the internal slot, so (38a) dzaanééz yiztał ‘It kicked the mule.’ and (38b) łį́į́ʼ biztał ‘It was kicked by the horse.’ ✓ INV-6.

I4. Willie (72a.4) *yishxash ‘it bit him’ and (72b.4) *bishxash ‘it was bitten by him’: stars with no nominals

  1. Σ Σ(√BITE) = {ρA, ρP}. (72a.4): the referent of ρA is an "it" and the referent of ρP is a "him", so r(ρP) > r(ρA).
  2. Σ VOICE = yi- means EXT = ρA, the lower-ranked one. INV-3 is violated: *.
  3. (72b.4): the referent of ρA is "him" and the referent of ρP is "it", so r(ρA) > r(ρP). VOICE = bi- means EXT = ρP, the lower-ranked one. INV-3 is violated: *.
  4. B No B step exists: there are no nominals to link and nothing to front. HYPOTHESIS Both stars are therefore made by Σ and rule set A alone (§5.1).

I5. Willie (73b) → (74a), and (76a): fronting under the inverse (rule set B)

  1. Σ Σ(√KILL) = {ρA, ρP}. EρA ⊇ {caus} (no vol or sent: lightning). EρP ⊇ {chg, aff} (ours). C5: ρA = the lightning, because it has the only P-A entailment.
  2. Σ The horse outranks the lightning (Willie calls (73a) "an animacy violation"). INV-3 gives EXT = ρP, so VOICE = bi-. A biisxį́.
  3. B Base linking (INV-6), (73b) łį́į́ʼ ʼiiʼniʼ biisxį́: ʼiiʼniʼ, next to V, is the internal slot = ρA; łį́į́ʼ is EXT = ρP.
  4. Move (A′, optional; INV-7) B. iiʼniʼ focus-fronts to the left edge, leaving a copy: (74a) iiʼniʼ łį́į́ʼ ⟨iiʼniʼ⟩ biisxį́ ‘The lightning killed the horse.’ The roles do not change, because bi- already fixed them in step 2. With yi- no such parse is used. (75a) łííʼ iiʼniʼ yiyiisxį́ can only mean ‘The horse killed the lightning.’ (p. 91).
  5. B (76a): adding -ísh to the fronted element gives *iiʼniʼish łííʼ biisxį́. The clitic "does not attach to the moved element", and the question is formed instead on the unmoved order (76b) łííʼ-ísh iiʼniʼ biisxį́.

What I1–I5 show HYPOTHESIS

  1. The entailments never change. I1–I3 share Σ step 1. Only EXT differs, which is exactly Grant's "the inverse changes only which argument is realized where".
  2. The voice is decided at Σ and spelled out by rule set A. I4 has stars with no rule-set-B input at all. Rule set B links nominals afterwards (INV-6) and may front one (INV-7), which only bi- licenses.
  3. The impossible cells are the ones where the external argument is the lower-ranked participant. Whether that is a grammatical * (Willie) or a pragmatic ?? (S&J) is not settled (T5b).
  4. Under VPISH raising, the inverse is the movement of a patient slot over the agent slot into subject position. No lexical item moves. That is the testable part (T5c).

7. Verdict on T5

  • Position, not only linking. SOURCED Willie 1991's answer is explicit. The inverse "makes the agent an internal argument … and places an argument with an "internal" theta role in the external argument position" (p. 100). Sandoval & Jelinek 1989:339 agree that the patient or goal becomes SUBJECT. HYPOTHESIS W&J's constant "3ACC-3NOM" gloss labels the prefix string. It is not a claim that positions stay constant. Which prefix indexes which slot is still OPEN (T5a), and Willie's own case labels disagree (Accusative on p. 72, Absolutive on pp. 77 and 100).
  • Where. HYPOTHESIS The choice is made at Σ (which slot is external, given rank) and spelled out inside the word by rule set A. This follows from (39) and (72), where the forms differ or are starred with no nominals. Rule set B only links and fronts. The earlier pages' topic-DP definition of VOICE is restated over EXT (§5.1, item 4).
  • Grant's ranking picture fits the data. HYPOTHESIS NAT (agent as subject) holds unless the non-agent outranks the agent. Then bi- is the expected form, and the two cells with a lower-ranked external argument are out ((72a.4), (72b.4)). Theme arguments never compete (INV-5).
  • Not settled. OPEN Whether VPISH raising is real here. That needs scope or binding evidence that the inverse patient is structurally above the agent (T5c). Also open: whether the ranking is grammar or usage (T5b). (2026-10-23: Oxford 2023's tests are run on deep or shallow inverse?; deep-like where runnable, binding still open, T53a.)

8. Open questions this page adds

#QuestionWhat would settle it
T5aIn biztał, does bi- index the external (patient) slot or the internal (agent) slot? W&J gloss the string 3ACC-3NOM in both voices. Willie calls the inverse patient Accusative on p. 72 and Absolutive on pp. 77 and 100.Young & Morgan 1987 paradigms; Willie 2000 (IJAL 66) and Aissen 2000, neither opened. Also inverse forms with a dual or plural on just one argument (Willie (63)–(69), pp. 81–85: da- scope. Page-checked 2026-10-09 on the number & scope page. da- counts the agent in both voices. That does not settle T5a, but it ties it to T7.) OPEN
T5bIs INV-3 a grammatical rule or a usage preference? Willie says "must" and stars (72a.4). S&J call (5) "pragmatically unacceptable" and acceptable for a supernatural horse. Speakers vary (Willie (73); Young's survey via Willie p. 16). HYPOTHESIS The supernatural-horse fact is Dowty's (ii)-type diagnostic (meaning-based, extendable in context) from the plain ł-causer page, so it favours a constraint that reads meaning over a fixed class of nouns.Native judgments on (5)-type pairs in plain and personifying contexts, by speaker and age (with T6). OPEN
T5cVPISH test: under bi-, does the external patient take scope over, or bind into, the internal agent, and not under yi-? V-P predicts yes; V-H predicts no difference.Willie (58)–(60), pp. 79–80. Page-checked 2026-10-09 on the number & scope page. They are not decisive, and the question is sharpened there as T10a. New elicitation with reflexives or quantifiers is still needed. OPEN
T5dWhy can a goal or benefactive become external under bi- but a theme cannot, even an animate theme (‘horse’ in (85)–(86))? Is the candidate set fixed by an entailment of the slot, or by structure (object of a "grammatical" postposition)? If it is an entailment, this is a T7 case: a generalization over Eρ that pronouns alone do not state.More three-participant verbs with animate themes; Jelinek & Willie 1996 ("Psych verbs in Navajo"), not opened. OPEN
T5eIs the clitic restriction (76a) the whole reason why W&J (60b) reading 2) is out? If so, a fronted wh-agent under bi- without -sh, or with -sh placed elsewhere, should be possible.Elicitation; W&J 2000 pp. 270–273 re-read together with Willie ch. 8. OPEN
T5fThe forms are spelled inconsistently: łį́į́ʼ / łííʼ / lííʼ and ʼiiʼni / ʼiiʼniʼ / iiʼniʼ within Willie §3.3.3, and dzaanééz (Willie) vs dzaaneez (S&J). Which spellings are typescript errors?Young & Morgan 1987 dictionary entries; a Navajo speaker. OPEN
T5gHale 1973 ("A note on subject-object inversion in Navajo"), the original statement that the higher-ranked noun comes first, is known here only through Willie (p. 12) and S&J (p. 336). Update 2026-10-14: Witherspoon 1977:69–70, 72, 80–81 quotes Hale 1973:303–306 (the animacy table, the ranking hypothesis, “cannot be formulated in absolute terms”, “relative acceptability”); page-checked there, see who acts on whom §4.A library copy of Kachru et al. (eds.) 1973, pp. 300–309. OPEN (the quotations are secondhand)

Added 2026-10-14 (8:46pm): with one overt noun, Atʼééd yiiłtsą́ is only ‘Sh/he saw the girl’ and Atʼééd b-iiłtsą́ only ‘The girl saw him/her’ (Speas 1986:250, (4.8)). So a lone noun always links to the bundle that is internal after the voice choice (FB-3a, Speas 1986 on Navajo). In a bi- relative under a yi- verb, (4.54), the horse is external in both clauses although it is the patient of one and the agent of the other.

Compared 2026-10-23: the Algonquian inverse (T52). Algonquian has two inverses: an obligatory, agreement-only one (3→1/2) and an optional voice that raises the patient (3′→3). Theta roles are unchanged in both. yi-/bi- lines up with the second only.

Follow-up (2026-10-09): number, quantifiers and da- across the two voices, on the number & scope page (T5c, T10).

Next increment (done 2026-10-08 evening): the fourth person (T8) — escape hatch from the same ranking clash, not a second inverse.

The inverse as Agree and EPP: which head probes, what moves, and what da- rules out (T27k)

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.

Working note. Kept as written (URL, data and status labels unchanged). Filed under Deep or shallow inverse; da-; the consolidated argument is the essay deep-or-shallow. Rule IDs: rules index.

Verbs on this page → dictionary: yiztał ‘kick’ · yiztsʼǫs ‘kiss’ · yishxash ‘bite’ · dabiiłtsą́ ‘see’ · nishhozh ‘tickle’ · yoolóós ‘lead (one animate)’ · yíghááh ‘join; with -tah ‘among’: go among’

Restated on the unified structure (2026-11-15). The inverse is restated as one extra Move on the clause spine: the patient moves (via an outer Spec,vP) above the agent, which has already raised to Spec,TP by the working default (H3, H6′, H21b). The event LF is the same as the direct; the trigger and landing site are OPEN. 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. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §4, inverse row. 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. IA-0/IA-1*/IA-2/IA-3 and their feature/probe choices are historical rival hypotheses. The data do not prove a unique inverse raising path, and a surface prefix is not automatically a moving DP or selecting head. Re-run these on the primary vP/V→v→T skeleton with a separately raised subject; preserve Agent/Theme at LF. Whether an inverse theme additionally raises, and to which left specifier, is OPEN. “No direct movement” belongs to the old comparator only.

The framework audit (F5, F6) flagged the site's inverse analysis because it inherits older notation: an “inversion transformation between deep structure and surface structure” (Hale 1973, as reported by Creamer p. 29 and Hale & Platero p. 15) and Case assigned to pronouns before incorporation (Willie 1991 pp. 72, 100). This page restates the yi-/bi- alternation in VPISH/minimalist terms. It says which head probes, what each probe is valued by, what (if anything) moves, and where a minimality problem arises. It tests four modern versions against every relevant form already on the site and against two analyses opened this run: Hale's own 1997 EPP analysis of the Navajo inverse (Tsaile workshop notes, not used on the site before) and Bruening 2005's Voice-[EPP] analysis of the Passamaquoddy inverse. Derivations are given step by step: Merge, Agree, Move, spell-out, with the rule set used at each step. Written 2026-10-16, 8:46pm run (PT).

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 an opened source, or from a site page that records it with page and example number. No Navajo form, judgment or scope reading is invented; where a test needs one, it is listed as OPEN or given schematically. Pre-VPISH and transitional notation is labelled and restated before it is tested (§4). Labels: SOURCED HYPOTHESIS GRANT'S CONJECTURE OPEN.

Rule-set letters and notation. The site convention is kept: A = word-internal (Merge, Agree and Move inside the verb word, and its spell-out), B = clause level (nouns, linking, clause spell-out), Σ = role-schema step. The Syntax · Semantics · Interface bot's notation conventions have still not arrived (checked: no conventions file in notes/ or the context folder), so the provisional Minimalist notation stays: copies in angle brackets, Spec,X, ⌢ for “precedes”. ρA is the role slot with the most Proto-Agent entailments for its role in the action, ρX the competing slot (patient, goal or benefactive), as on the inverse page.

Bottom line, in plain words

HYPOTHESIS Three modern versions survive the data on the site; one is in trouble.

  1. IA-1* (“the patient's slot moves up, but only in the inverse”). In the inverse, the little-v head carries an extra movement feature (an EPP feature). It lifts the patient's pronoun slot over the agent's slot, and Mode/Subject then lifts it into the subject position. In the direct voice nothing moves. This is Bruening's 2005 Passamaquoddy mechanism, minus his movement in the direct voice. That one change is needed so that Navajo subject agreement keeps tracking the agent.
  2. IA-3 (“nothing moves; the patient's slot is marked as topic”). This is Hale 2003's in-situ picture. The object prefix is yi- “when both subject and object are third person and the object is not the topic” (Hale 2003 ms. p. 27), and bi- otherwise.
  3. IA-2 (Hale 1997's own analysis). The object always moves to Spec of the Mode phrase, to get agreement. Then whichever argument goes on to the top specifier (EP) is the clause's subject or topic: the agent in the direct, the object in the inverse (Hale 1997, PDF pp. 21–25). It fits the forms, but it creates a minimality problem in the direct voice (the agent has to cross the object). It also has a freezing clause that keeps da- from reaching the agent. Both are tensions inside the theory, not contradicting data.
  4. IA-0 (“build the clause the other way round”, inverse linking) is disfavoured by da-. The distributive prefix counts the agent in both voices (Willie 1991:32, 82–85). If the patient were merged in the agent's position, a structural number probe would count the patient. This is the Navajo counterpart of Bruening's argument from Algonquian TA/TI finals (2005 pp. 4–6). It fails only if da- picks out its slot by role rather than by position.

HYPOTHESIS The new constraint: da- and the word's own order. In dabiiłtsą́ (Willie (64a)), da- comes before bi- and counts the agent. Suppose the patient's pronoun really raises inside the word, and the word is ordered by the site's spell-out rule (CYC: non-heads first, higher before lower). Then da- must sit above the raised pronoun to come first. But from there, a probe that takes the nearest pronoun would count the patient. So raising works only if one extra assumption holds: (e2) da- looks for the agent's role feature, not the nearest pronoun; (e3) bi- is an agreement head, as in Hale 1997, not the moved pronoun; or (e4) da- copies the number that Mode/Subject has already taken from the agent. The in-situ version (IA-3) needs none of these. Option (e2) is Grant's role-based reading stated as a probe.

HYPOTHESIS A point for raising. In three-participant verbs a goal or benefactive can invert with the agent, but the theme never does (Willie p. 95; INV-5). Raising explains this if the postpositional object sits above the theme: the movement feature on v takes the closest internal slot, so the theme is always blocked (§6, K6). The in-situ version has to stipulate it. Where the postpositional phrase sits is OPEN (T28d, T30a).

HYPOTHESIS A correction to the inverse page. Step 6 of derivation I1 (“Agree M/S ⟶ EXT”) is replaced. In IA-1* the patient moves first and Mode/Subject then agrees with the agent left below it (Bruening's order, 2005 n. 22). This matches Willie & Jelinek's constant 3NOM gloss. It also removes the conflict the number & scope page found between I1 and da-.

GRANT'S CONJECTURE + HYPOTHESIS For the thesis. In every surviving version, the thing the inverse ranks, and moves or marks, is a role-bearing pronoun slot, and the choice is made before any noun enters. The inverse is decided, or starred, with no noun in the sentence ((39), (72)). The same noun changes rank by construal (Sandoval & Jelinek's supernatural horse). The nouns link afterwards (INV-6). Nothing opened contradicts “the inverse ranks roles, not words”.

OPEN What decides between IA-1* and IA-3: one scope judgment per sentence, (58) with bi- (ambiguous under raising, rigid under in-situ; 4th person compared §6.2, T10a). The second decider is the height of the postpositional phrase in three-participant verbs (T30a). Neither is in any opened source.

On this page: 1 Sources opened · 2 Data used · 3 Hale 1997 and Bruening 2005: what they say · 4 Older analyses, labelled and restated · 5 Formal statement: heads, probes, options · 6 Derivations K1–K7 · 7 Results table · 8 What each gets right and misses · 9 Dowty, roles not words, feature bundles, STRATA · 10 Gaps and next (T30a–h)

1. Sources opened for this page

  • SOURCED Hale, Ken 1997, “Remarks on the syntax of the Navajo verb”, Parts I–III (Navajo Language Academy workshop, Tsaile, July 1997; MIT papers hale023.pdf, 61 pp.; box copy [local working copy, not in the repository]). It was classified on the framework audit, but its inverse analysis had not been used on the site before. Read: PDF pp. 1–2 (the template and E), 9–12 (agreement with M and E; transparency (18)), and 14–25 (Part II §§1–3: examples (2)–(16), the EPP (6), MAXIMUM LOCALITY (8), trees (4), (5), (7), (9)–(15)). Page images of PDF pp. 14, 15, 16 and 22 were checked against the text layer. Printed folios were not checked, so PDF page numbers are cited throughout. Class (b), transitional: Hale himself calls his d-/s-structure derivations “ultimately fictions” and the Merge-and-Move theory “conceptually superior” (PDF p. 25).
  • SOURCED Bruening, Benjamin 2005, “The Algonquian Inverse is Syntactic: Binding in Passamaquoddy” (ms., Univ. of Delaware; author's PDF udel.edu/~bruening/Downloads/SyntacticInversion2.pdf, re-downloaded this run, 27 pp., born-digital). Read: pp. 1–6, 13–24, examples (1)–(12), (36)–(39), (44)–(50), §§4.1–4.3, nn. 22–23. On the site since 2026-10-09 for the scope and binding diagnostics (4th person compared §6); its agreement mechanism (§4) is used here for the first time.
  • SOURCED Hale 2003 ms. (“On the significance of Eloise Jelinek's Pronominal Argument Hypothesis”; box text [local working copy, not in the repository]), re-read at pp. 10–17 and 26–27.
  • Used as recorded on the site with page and example numbers (not re-opened): Willie 1991 (37), (39), (64)–(65), (72), (79)–(80), (85)–(86) and pp. 16, 27–28, 32, 72, 77, 82–86, 94–95, 100 (inverse §2, number & scope §2); Sandoval & Jelinek 1989 (5), (6), pp. 336–339; Willie & Jelinek 2000 pp. 268, 273 and the gloss PL-3ACC-3NOM-saw; Hale 1973 as quoted by Witherspoon 1977:69–81 (who acts on whom).
  • Searched for but not usable: Perkins 1978 (the box file under that name is a 2020 Utah honors thesis on accent perception, as the audit already noted). Hale 1973 itself is still unopened (T27a).

2. Data used

SOURCED New on the site: Hale 1997's pair. Hale cites (2) from Young & Morgan 1987 (Grammar, p. 65).

Hale 1997 (2), PDF p. 14 (cited from Y&M 1987G:65); tree (4), PDF p. 16
Mósítsídiiy-ní-Ø-Ø-¬-ʼ²í
catbird3OBV.OBJ-E-IPFV-3SBJ-[V]†

OPEN † Segmentation not verifiable (2026-11-11 audit pass 4): Hale 1997 (2) prints yiní¬ʼ²í with no segmentation and no gloss; the segmentation and gloss here are ours (HYPOTHESIS). The ¬ piece has no gloss because the glyph is damaged in the PDF, and we do not reconstruct it.

‘The cat is looking at the bird.’

Hale 1997 (3) = (16), PDF pp. 15, 25; trees (5), (12)–(15), PDF pp. 17, 23–25
Tsídiimósíb-ní-Ø-Ø-¬-ʼ²í
birdcat3.OBJ-E-IPFV-3SBJ-[V]†

OPEN † Segmentation not verifiable (2026-11-11 audit pass 4): Hale 1997 (3) prints biní¬ʼ²í with no segmentation and no gloss; the segmentation and gloss here are ours (HYPOTHESIS). The ¬ piece has no gloss because the glyph is damaged in the PDF, and we do not reconstruct it.

‘The cat is looking at the bird.’ ‘The bird is being looked at by the cat.’ (both translations Hale's)

Fidelity. The surface verbs print as yiní¬ʼ²í and biní¬ʼ²í. The characters “¬” and “²” are font-substitution glyphs: they appear on the page images themselves, not only in the text layer (checked at 400 dpi). They are reproduced as printed and not repaired. The segmentation is Hale's, from trees (4) and (5): AGRo y-/b- on E, E ní-, M Ø (“zero-imperfective”), AGRs Ø (“the subject is third person”), all on PDF p. 16, and V printed “-¬-ʼýí”. The V is left unglossed (Hale gives only the sentence translation). 3OBV.OBJ follows Hale's report that y- “is glossed 3o by Young and Morgan … reflecting its obviative function”. 3.OBJ follows his statement that b- “is in fact the general third person object agreement marker” (both PDF p. 16). E is Hale's label for the position left of Mode, a mnemonic for “edge” (PDF p. 1). In Hale 2003's template it is position VI, Qualifier (ms. p. 14, (10)), so E is read as Qual below. That identification is ours (HYPOTHESIS) and rests on the position.

SOURCED Already on the site, used here (glosses as printed in the source, already in capitals; S&J segment only yi-/bi-):

FormSourceWhat it shows
łį́į́ʼ dzaanééz yiztał horse mule 3.3.kicked ‘The horse kicked the mule.’ / dzaanééz łį́į́ʼ biztał mule horse INV3.3.kicked ‘The mule was kicked by the horse.’Willie 1991:71, (37a, b)Equal rank: both voices, “discourse determined” (p. 72)
yiztał 3.3.kicked ‘It kicked it.’ / biztał INV3.3.kicked ‘It was kicked by it.’Willie 1991:73, (39a, b)Voice differs with no nominal at all
*yishxash ‘it bit him’ (72a.4); *bishxash ‘it was bitten by him’ (72b.4)Willie 1991:87Stars with no nominal: a lower-ranked external argument is out
?? łį́į́ʼ hastiin yi-ztał horse man yi-kicked ‘The horse kicked the man.’ / hastiin łį́į́ʼ bi-ztał man horse bi-kicked ‘The man got kicked by the horse.’Sandoval & Jelinek 1989:336, (5), (6)(5) “pragmatically unacceptable”, acceptable if the horse is “a supernatural being” (p. 336)
dabiiłtsą́ PL.INV3.3.saw ‘They/he (was)were seen by them/*him.’; (65b) *ʼashiiké ʼatʼééd dabiiłtsą́ (The boys were seen by the girl.)Willie 1991:82–83, (64a), (65b); W&J gloss PL-3ACC-3NOM-sawda- requires a plural agent in the inverse too (p. 84); da- precedes bi-
ʼashkii ʼatʼééd łį́į́ʼ yá yoolóós boy girl horse 3.for 3.3.lead[SG.+ANIM] ‘The boy is leading the horse for the girl.’ / … bá yoolóós ‘The girl is leading the horse for the boy.’Willie 1991:95, (85), (86)Benefactive and agent invert; “the theme argument does not take part in the inversion process” (p. 95)
ni-sh-hozh 2SG-1SG-tickle ‘I tickle you.’Hale 2003 (1), p. 5 (A/B as spell-out N1)Mode/Subject agrees with the agent (1SG) in a direct form

Willie's bracket “[sing. +animate]” is given in Leipzig style as [SG.+ANIM]. Her “3.to” and “3.for” for yichʼįʼ/bichʼįʼ and yá/bá carry no INV, as printed.

3. Hale 1997 and Bruening 2005: what they say

3.1 Hale 1997 on the Navajo inverse SOURCED

  • Spine and agreement heads. EP over MP over VP. The subject is the “distinguished adjunct” of VP. Two agreement-bearing heads: M carries AGRs, and E carries AGRo (PDF pp. 9–10, 16). An argument agrees with the closest agreement-bearing head. “Once that agreement relation is established, the subject and M are no longer accessible to further agreement relations” (PDF p. 10).
  • y- versus b-. “The obviative y- is much more restricted, being limited strictly to situations in which a third person object is in the c-command domain of a clausemate third person subject. In all other situations, b- is the agreement marker which appears” (PDF p. 16). Hale considers whether b- causes the object's movement or results from it: “the appearance of b- would be a result, not the cause, of the object moving out of the c-command domain of the subject. This seems more reasonable” (PDF p. 18).
  • Why the object moves. Speakers at the 1974 workshop “remarked that the Inverse placed the object in the position of topic”. Platero observed that a relative clause formed on the inverse “unambiguously” takes the fronted object as head (PDF p. 18, citing Hale, Jeanne & Platero 1977, not opened). Hale defines a “clausal topic” as the DP “in the highest Spec or adjunct position in the inflectional portion of the extended projection of V” (n. 11, PDF p. 18).
  • EPP. “(6) A verb and its inflectional projection is a predicate and must be predicated of a subject” (PDF p. 19). So in the direct, too, the subject raises to Spec of EP, although “there is no surface evidence” for it (PDF p. 20, tree (7)).
  • MAXIMUM LOCALITY (8) (PDF p. 21): DP may agree with H only if H c-governs DP and is the closest such head. So the object cannot agree with E from its base position. It moves to Spec of MP, in both voices (trees (10), (13)). In the direct, the subject then raises to Spec of EP, (11). In the inverse, the object “must move to Spec of MP … But it cannot do that in one step, because if it did, it would fail to be c-governed by an agreement-bearing head”, so it moves on to Spec of EP in a second step, (13)–(14) (PDF pp. 23–24).
  • Hedges. “We do not know, at this point, precisely what the mechanisms involved in the Inverse are” (PDF p. 17). Base-generating the object in Spec of EP with b- as a “resumptive pronoun” is listed as an open option (PDF p. 18).

3.2 Bruening 2005 on the Passamaquoddy inverse SOURCED

  • Arguments are merged by role: the object is the sister of V, and the subject is in Spec,VoiceP (Voice “projects the subject”, p. 19, citing Kratzer). Above them are Σ, Infl and C, (45), p. 19. He gives three A-positions (5), p. 3.
  • The inverse = an optional [EPP] on Voice. It moves the object “to a second specifier of VoiceP, above the subject”, (50), p. 20. “Now when a similar feature on Infl drives movement of an NP to Spec-IP, it will be the object that moves rather than the subject, since the object is now the closest NP to Infl” (p. 20). The direct/inverse morpheme is Voice spelled out as [−EPP] or [+EPP] (p. 20).
  • Order of operations. “It is crucial here that movement takes place prior to Agree, such that a head that has a feature driving movement may not actually Agree with the NP that it causes to move” (n. 22, p. 20). So Infl agrees with what is left in its c-command domain: the object in the direct (p. 19), the subject in the inverse (p. 20).
  • The plural head. Plural agreement sits on Σ, which agrees with “the highest NP in Spec-VoiceP” (n. 23, p. 23). In the inverse that is the raised object.
  • The hierarchy. There is no participant hierarchy in the grammar (pp. 1–3, 24). Discourse “may play a role in deciding whether to use the [EPP] feature” (p. 22; the optional feature is introduced on p. 21). Animacy is encoded as a restriction on attracting heads: they “only affect NPs with the feature [animate]” (p. 22).
  • Evidence. Variable binding and weak crossover reverse in the inverse, (36)–(37), pp. 13–14. Scope is ambiguous in the inverse, by reconstruction, (39), pp. 14–15. Inverse linking is excluded because the TA/TI final keeps tracking the logical object, (6)–(10), pp. 4–6.

4. Older analyses, labelled and restated

Analysis (class)As statedRestated in a VPISH/TP/CP spine (HYPOTHESIS)
Hale 1973, via Creamer p. 29, Hale & Platero p. 15, Witherspoon pp. 69–81: pre-VPISH (transformational), secondhand“Subject-object inversion” rule between deep and surface structure, conditioned by rank; “cannot be formulated in absolute terms” (1973:305, via Witherspoon)No levels. Either A-movement of ρX's pronoun to the highest A-specifier inside the word (IA-1*, IA-2), or a topic feature on ρX read at spell-out (IA-3). The rank condition becomes a condition on choosing the feature (§5.3).
Sandoval & Jelinek 1989: pre-VPISH (flat S)Nominals are adjuncts to S (p. 351); bi- changes the grammatical relations of the prefixes (p. 339 (c))Flat-S trees not used. Their (c) is kept as a fact: ρX bears the subject relation under bi-. Which node “subject” is (Spec,M/SP? Spec,QualP?) is what IA-1* and IA-2 disagree on (T30g).
Willie 1991: transitional (GB)Case assigned to pronouns before incorporation; inverse agent “internal” with Ergative case; inverse patient Accusative (p. 72) or Absolutive (pp. 77, 100)No Case Filter (framework audit F4). “Internal” = ρA not in the highest A-position. Her “Ergative” is restated as: M/S agrees with ρA after ρX has moved past it (IA-1*), which is also W&J's 3NOM. The two case labels for the patient are not decided here.
Hale 1997: transitional (knowingly)EP/MP/VP; Move Alpha; d-/s-structure “fictions” (PDF p. 25); AGRo and AGRs as headsEP → QualP, MP → M/SP (Hale 2003 says M “correspond[s] to I(nfl) and T(ns)”, ms. p. 10), VP-adjoined subject → Spec,v. So the Spec,MP stop is Spec,TP-like and Spec,EP is a QualP specifier above it. Stated as IA-2.
Hale 2003: minimalist, with a GB residueObject raising to Spec,M “for Case” considered and rejected as against the spirit of PA theory (ms. pp. 11–12); y- “as required when … the object is not the topic” (p. 27)IA-3: no A-movement; a [top] feature on ρX decides the exponent.

5. Formal statement

5.1 Heads, probes, and what values them

SPINE (Hale 2003 ms. pp. 10–11, 16–17; Hale 1997 E read as Qual: HYPOTHESIS)
   [CP C [QualP (Pl, Pv adjuncts) Qual(=E) [M/SP M/S [vP ρ_A [v′ v [VP (PP) [V′ ρ_X V ]]]]]]]

PROBE / FEATURE        HEAD     VALUED BY                                   STATUS
 [uφ] subject agr.      M/S      the closest φ-bundle when M/S probes        SOURCED as "closest" (Hale 1997 p. 10);
                                                                             timing = §5.2
 [uφ] object agr.       Qual/E   ρ_X (AGRo; Hale 1997 pp. 9–10, 16)          SOURCED (Hale 1997); Hale 2003 instead
                                                                             treats Obj as the pronoun itself
 [EPP]                  M/S      — (movement only; attracts the closest DP) HYPOTHESIS (Bruening's Infl, p. 20)
 [±EPP]                 v        — (optional; lifts the closest internal     HYPOTHESIS (Bruening's Voice, p. 20)
                                 slot to an outer Spec,v)
 [u#] / association     Pl (da-) ρ_A in both voices (Willie 1991:32, 82–85)  the fact SOURCED; mechanism = §5.4
 [top]                  on ρ_X   discourse / rank                           HYPOTHESIS (Hale 2003 p. 27; IA-3)

OBV (spell-out of 3rd-person object agreement, rule set A; SOURCED as a generalization, Hale 1997 p. 16)
   3.OBJ → y-  if, at spell-out of the word, ρ_X is c-commanded by a 3rd-person ρ in the same word domain
         → b-  elsewhere ("the general third person object agreement marker")
   OBV-top (Hale 2003 p. 27 version): 3.OBJ → y- if both slots are 3rd person and ρ_X is not [top]; b- elsewhere

5.2 Four options

IA-0  INVERSE LINKING   ρ_X first-merged in Spec,v and ρ_A in Comp,V under bi-.  (Bruening's rejected option, pp. 3–6)
IA-1* RAISING, INVERSE ONLY  (ours, after Bruening 2005)
      direct:  v[−EPP], M/S without EPP; nothing moves; M/S Agrees with ρ_A.
      inverse: v[+EPP]: ρ_X → outer Spec,v (crosses ρ_A inside vP: Bruening (50));
               M/S[EPP]: closest DP = ρ_X → Spec,M/SP;  THEN M/S Agrees with the closest remaining goal = ρ_A
               (Move before Agree: Bruening n. 22).
IA-2  HALE 1997         both voices: M Agrees with ρ_A first (then frozen); ρ_X → Spec,M/SP (MAXIMUM LOCALITY,
                        to Agree with Qual/E). Qual/E[EPP]: direct ρ_A → Spec,QualP; inverse ρ_X → Spec,QualP.
IA-3  IN SITU (Hale 2003)  no A-movement in either voice; ρ_X carries [top] under bi-; OBV-top decides y-/b-.

HYPOTHESIS Why IA-1* drops Bruening's movement in the direct. Bruening has Infl's EPP raise the subject in the direct and then agree with the object left behind (p. 19). In Passamaquoddy that agreement is the final suffix. Navajo's Mode/Subject agreement tracks the agent in direct forms: ni-sh-hozh, where sh is the 1SG agent. If M/S raised ρA and then agreed with what remained, it would agree with the 2SG patient. Keeping Bruening's order (Move before Agree) but giving M/S an EPP only in the inverse removes the clash. This is a choice, not a finding. Hale 1997 would object that the direct then has no clausal subject (EPP (6), PDF pp. 19–20). That is a theory-internal tension, not a datum.

5.3 Where the ranking enters

RANK→EPP (HYPOTHESIS; restates INV-3 of the inverse page as a condition on the feature choice)
   r(ρ_X) > r(ρ_A)  → v must be [+EPP]   (IA-1*)  /  ρ_X must be [top]   (IA-3)      Willie (70) "must"
   r(ρ_A) > r(ρ_X)  → v must be [−EPP]          /  ρ_X must not be [top]           (72a.4), (72b.4)
   r(ρ_A) = r(ρ_X)  → free; discourse decides (Bruening p. 22: discourse "may play a role in deciding
                      whether to use the [EPP] feature")                                Willie (37), p. 72

HYPOTHESIS Bruening's own way of getting animacy effects, a restriction on attracting heads to [animate] NPs (p. 22), cannot express Navajo's ranking. Man and horse are both animate, yet the inverse is the expected form for ‘the horse kicked the man’ (S&J (5)/(6)). Hale (1973:305, via Witherspoon) says the conditions are relative, not absolute. A binary feature on a head cannot compare two slots. So in Navajo the comparison has to happen where the feature is chosen, at Σ. Bruening's further claim that the hierarchy plays no role in the computation itself is not contradicted by this. The comparison chooses the feature; the movement it triggers is ordinary.

5.4 Four ways for da- to find the agent

P-probe  Pl is a φ-probe adjoined where Hale 2003 puts it (QualP, ms. pp. 17, 39) and Agrees with the closest bundle.
P-low    Pl is adjoined inside vP and Agrees with ρ_A before anything moves (number & scope §5.1, row 3;
         Hale 2003 p. 17 allows low adjunction for some adjuncts).
e2 P-role  Pl looks for the bundle that bears ρ_A's role feature (a probe relativized to a role, as Bruening's
         attractors are relativized to [animate]).            GRANT'S CONJECTURE (role reading) → HYPOTHESIS
e3 AGR-head  bi-/yi- are agreement on Qual/E (Hale 1997), not the moved pronoun; the moved slot is silent.
e4 P-via-M/S  da- carries the number that M/S took from its goal (concord through M/S; number & scope T10b).

6. Derivations K1–K7, step by step

Format as on A/B as spell-out: each step is Merge, Agree, Move or Spell-out, tagged Σ, A (inside the word, domain D1) or B (clause, domain D2). Spell-out uses CYC (C1 non-heads before heads, C2 non-heads higher first, C3 heads higher first). Role-in-the-action entailments are our assignments (HYPOTHESIS), with Dowty's proto-roles as the baseline. No verb is segmented beyond its source.

K1. Hale 1997 (3) Tsídii mósí biní¬ʼ²í (inverse), under IA-1*, then IA-3 and IA-2

  1. Σ Σ = {ρA (the cat: perceiver; sentience), ρP (the bird: no Proto-Agent entailment for its role in the action)}. ρA is the agent slot (C5) and stays so (INV-1). Rank: Hale gives (2) and (3) as alternatives with the same meaning (PDF p. 15), so the choice is free here. Creamer's groups for these two animals are not checked. Discourse chooses the inverse: v is [+EPP] (RANK→EPP, third line).
  2. A Merge V with ρP (Comp,V). Merge v[+EPP]. Merge ρA in Spec,v.
  3. A Move (v's EPP): ρP → outer Spec,v, crossing ρA inside vP (Bruening (50)). No minimality violation: this is movement to a second specifier of the head whose specifier ρA fills.
  4. A Merge M/S[EPP]. Move: the closest DP is ρP (outer Spec,v) → Spec,M/SP. ρA is not crossed at this step.
  5. A Agree M/S ⟶ the closest remaining goal = ρA [3] → AGRs Ø (Hale: “the subject is third person, hence Ø”, PDF p. 16). M is Ø (zero-imperfective).
  6. A Merge Qual/E ní-. Agree Qual/E ⟶ ρP [3] (object agreement).
  7. A Spell-out D1. OBV: ρP is no longer c-commanded by ρA, so the 3rd-person object exponent is b-. CYC: N = {ρP=b}, H = {ní, Ø, Ø, V}. C1 puts b first, and C3 orders ní ⌢ M Ø ⌢ AGRs Ø ⌢ V. Result: b-ní-Ø-Ø-[V], printed biní¬ʼ²í ✓.
  8. B Merge the nouns at D2 and link them (INV-6). mósí, next to the verb, links to the internal slot, which under bi- is ρA. Tsídii, the outer noun, links to the external slot ρP (topic adjunct; W&J 2000:268). Spell-out D2: Tsídii ⌢ mósí ⌢ [W] ✓.

Moves: 2 (both inside the word, both of the silent slot ρP; no noun moves).

IA-3 variant. Step 1 marks ρP [top] instead of choosing v[+EPP]. Steps 3–4 do not exist. At step 5 M/S Agrees with ρA, the closest goal at merge. At step 7 OBV-top gives b-, because ρP is [top]. Same string, 0 moves. Under Hale 1997's c-command version of OBV, IA-3 would wrongly give y-, because ρP stays c-commanded. So IA-3 needs Hale 2003's topic version of the rule.

IA-2 variant (Hale 1997 (12)–(15)). M Agrees with ρA first and freezes (PDF p. 10). ρP → Spec,M/SP to be close enough to agree with E (8). E's EPP then raises ρP on to Spec,QualP. Two moves, as Hale says (“it cannot do that in one step”, PDF p. 23). Hale moves the DP tsídii itself. In the restatement the moved element is the slot's pronoun, and the noun is an adjunct, as Hale 2003 has it (p. 27). That is a difference of notation, not of data.

K2. Hale 1997 (2) Mósí tsídii yiní¬ʼ²í (direct): where Hale 1997's minimality problem sits

  1. Σ Same Σ as K1. Discourse chooses the direct: v[−EPP].
  2. A IA-1*: Merge V + ρP, v, ρA, M/S (no EPP). Agree M/S ⟶ ρA (closest). Merge Qual/E; Agree ⟶ ρP. No movement.
  3. A Spell-out D1. OBV: ρP is c-commanded by 3rd-person ρA, so y-. CYC gives y-ní-Ø-Ø-[V], printed yiní¬ʼ²í ✓.
  4. B tsídii, next to the verb, links to the internal slot ρP, and Mósí to ρA. ✓
  5. A IA-2 instead (Hale's (10)–(11), PDF p. 22). ρP → Spec,M/SP (to agree with E). E's EPP must then raise ρA from its VP-adjoined position to Spec,EP, across ρP in Spec,M/SP, which c-commands it. If movement obeys the same closest-goal logic that Hale's (8) imposes on agreement, E would attract ρP, giving the inverse. The direct would then be underivable.

HYPOTHESIS In Hale 1997 the minimality problem is not in the inverse (where the object is already closest) but in the direct. Hale states locality only for agreement (8) and says nothing about locality of movement. This is a derived tension, not counted against him. IA-1* has no crossing at a distance in either voice. IA-3 has no movement.

K3. S&J (6) hastiin łį́į́ʼ biztał and (5) ?? łį́į́ʼ hastiin yiztał: the rank decides the feature

  1. Σ Σ(√KICK) = {ρA = the horse (volition, causation, movement in its role in the action), ρP = the man (affected)}, as on inverse I1.
  2. Σ Rank: r(man) > r(ordinary horse). GRANT'S CONJECTURE: the rank is the referent's capacity for Proto-Agent properties (sentience, volition, causation, movement), an extension of Dowty. RANK→EPP, first line: v must be [+EPP] (IA-1*) / ρP must be [top] (IA-3).
  3. A (6): as K1 steps 2–7 → biztał ✓. B łį́į́ʼ, next to the verb, links to the internal slot ρA, and hastiin to ρP ✓.
  4. A (5): v[−EPP] (or no [top]) is chosen against step 2. Every word-internal step goes through and the word yiztał is built. B linking is coherent. The only violation is RANK→EPP at Σ, which matches S&J's ?? (“not ungrammatical, but pragmatically unacceptable”, p. 336). If the horse is construed as “a supernatural being, with greater than human powers”, then r(ρA) > r(ρP) and [−EPP] is required. (5) is then fine.

HYPOTHESIS The comparison reads the participant as construed, not the word łį́į́ʼ: same noun, opposite outcome. So the rank is a feature of the bundle's referent, available at Σ, just as number is “a property of … the pronouns” (Willie p. 85).

K4. Willie (72a.4) *yishxash, (72b.4) *bishxash: the stars come before any noun

  1. Σ (72a.4) ‘it bit him’: r(ρP) > r(ρA), so RANK→EPP requires v[+EPP] / [top]. The form has yi-, i.e. [−EPP] / no [top]: *. (72b.4) ‘it was bitten by him’: r(ρA) > r(ρP), so v must be [−EPP]. The form has bi-: *.
  2. A The words themselves are buildable: there is no Agree or Move failure under any option.
  3. B No B step exists, because there is no noun to link.

HYPOTHESIS All four options locate the decision at Σ and the first word-internal feature choice, before any noun. That is the result of inverse §5.1, now stated as a feature choice.

K5. Willie (64a) dabiiłtsą́ and (65b) *: the da- trilemma

Willie 1991:82–83, (64a), (65b) (as on number & scope §2.4)
dabiiłtsą́
PL.INV3.3.saw

‘They/he (was)were seen by them/*him.’

W&J 2000 gloss the same word PL-3ACC-3NOM-saw. Neither source segments it further; it is not segmented here.

  1. Σ Σ(√SEE) = {ρA (perceiver; sentience), ρP}. #(ρA) = pl (Willie: “With dabi-, the internal argument must be plural”, p. 84). Inverse chosen.
  2. A IA-1*: K1 steps 2–6. ρP is now in Spec,M/SP, and M/S has Agreed with ρA [3, pl].
  3. A Merge Pl, adjoined to QualP (Hale 2003 pp. 17, 39). P-probe: Agree Pl ⟶ the closest bundle in its c-command domain = ρP (Spec,M/SP). That counts the patient ✗: it predicts ‘he was seen by him’ with a plural ‘he’, which Willie rules out (p. 84; (65b)).
  4. A P-low instead: Pl adjoined inside vP Agrees with ρA before ρP moves ✓. But then Spell-out D1 by CYC C2 has ρP (Spec,M/SP) higher than Pl, so b ⌢ da: *bi-da-…, against the printed da-bi-… ✗. The only way out is to pronounce the lower copy of ρP.
  5. A Ways out. (e2) Pl, adjoined high, looks for the bundle with ρA's role feature: it finds ρA ✓, and C2 gives da ⌢ b ✓. (e3) bi- is AGRo on Qual/E (Hale 1997): it is a head, so C1 puts da before it whatever their heights ✓, and a low Pl Agrees with ρA ✓. (e4) da- carries the [pl] that M/S took from ρA in step 2 ✓. Under IA-1*, M/S agrees with ρA in both voices (Move before Agree), so (e4) works only because of that ordering.
  6. A Spell-out D1 (with e2, e3 or e4): da ⌢ b ⌢ heads → dabiiłtsą́ ✓.
  7. B (65b) *ʼashiiké ʼatʼééd dabiiłtsą́: ʼatʼééd, next to the verb, links to the internal slot, which under bi- is ρA. ρA is [pl] by step 1 and the noun is singular: mismatch, * ✓ (predicted, whichever way out is taken).

The other options on the same word. IA-3: nothing moves. Pl, adjoined high, finds ρA (Spec,v) before ρP (Comp,V) ✓, and C2 gives da ⌢ b ✓. No extra assumption is needed. IA-2: after Agreeing with M, ρA is frozen for “the features person and number” (Hale 1997 PDF p. 21). A φ-probe Pl cannot reach it in either voice, so only (e4) works. IA-0: ρP is first-merged in Spec,v, so any structural Pl finds ρP ✗. Only (e2) rescues it.

HYPOTHESIS Contrast with Passamaquoddy. Bruening's plural head Σ agrees with the highest NP in Spec,VoiceP, so in the inverse it counts the raised object (n. 23). Navajo da- does the opposite: it counts the agent in both voices. So even if the two inverses use the same movement, their plural markers are not the same kind of probe. This is the strongest structural difference between them found so far.

K6. Willie (85)/(86): why the theme never inverts (INV-5) under IA-1*

  1. Σ Σ(√LEAD, animate theme) = {ρA (the leader: volition, causation, movement), ρBEN (the object of the postposition -á ‘for’), ρTH (the horse: moved)}. (85): the boy is ρA and the girl ρBEN. (86): the girl is ρA and the boy ρBEN. The roles in the action stay the same; only who fills them differs.
  2. A Merge V + ρTH (Comp,V). Merge the PP [ρBEN -á]. OPEN Where? Assume it is inside VP, above ρTH. Merge v. In (86), v is [+EPP]. Merge ρA in Spec,v.
  3. A Move (v's EPP, closest internal slot): ρBEN c-commands ρTH, so ρBEN moves to outer Spec,v, then to Spec,M/SP (as K1). ρTH can never be the closest while ρBEN is there: INV-5 follows ✓.
  4. A Spell-out of the PP: OBV applied to the P-object gives y-á → yá when ρBEN stays c-commanded by ρA (85), and b-á → bá when it has moved out (86) ✓. This extends Hale's statement about object agreement to postpositional objects, which fits Willie's remark that inverse bi- “is clearly cognate with the third person bi- pronouns that appear as possessors and postpositional objects” (p. 28). The verb yoolóós is the same in both, as printed.
  5. B Linking: the nouns are matched to ρA, ρBEN and ρTH as Willie's translations show. How three nouns are linked by position is not derived here (INV-6 is stated for two).

HYPOTHESIS IA-1* derives INV-5 from closeness, but only if the PP sits above the theme inside the verb phrase. If the PP is a higher adjunct, ρTH is the closest internal slot and IA-1* predicts the wrong inversion. IA-3 and IA-2 have to stipulate INV-5. This turns the inverse page's T5d (“entailment or structure?”) into a question about PP height (T30a).

K7. IA-0 on Willie (37b) dzaanééz łį́į́ʼ biztał

  1. Σ ρA = the horse, ρP = the mule (equal rank; discourse chooses bi-).
  2. A IA-0: Merge V with ρA in Comp,V, and ρP in Spec,v (the reversed linking). Agree M/S ⟶ ρP. OBV: ρP is not c-commanded, so b- → biztał ✓.
  3. B Linking as in I3 ✓.

HYPOTHESIS The string comes out. The problem is da- (K5), and the general preference for merging by role (Bruening p. 15, citing UTAH). The site's own schema also merges by role: pos ρA = Spec,v (C5). IA-0 would make the merge site depend on rank, against INV-1's spirit, though not against its letter.

7. Results table

Test (data)IA-0 linkingIA-1* raising (inverse only)IA-2 Hale 1997IA-3 in situ
Voice decided and starred with no noun ((39), (72))✓✓✓✓
OBV: y- only under c-command by a 3rd-person subject (Hale 1997 p. 16)✓✓✓only with OBV-top (Hale 2003 p. 27)
da- counts the agent in both voices (Willie 32, 82–85)✗ unless e2only with e2, e3 or e4only with e4 (freezing)✓ free
Order da-bi- under CYC (dabiiłtsą́)✓✓ if Pl is high, which conflicts with the row above unless e2/e4; or e3✓✓
Minimalityn/a✓ (outer Spec,v, then closest)crossing in the direct (K2)n/a
M/S agrees with the agent in direct forms (ni-sh-hozh)✓✓ (no EPP in the direct)✓✓
Theme never inverts (INV-5; Willie 95)stipulated✓ if the PP is above the theme (OPEN)stipulatedstipulated
Ranking is relative, not absolute (Hale 1973 via Witherspoon; S&J (5))all need a comparison at Σ (RANK→EPP); Bruening's [animate]-restricted attractors (p. 22) alone cannot express it
Scope in (58) tʼáá ʼałtso ʼatʼééké ʼashkii łaʼ biztsʼǫs (predicted)rigid, girls > boyambiguous (reconstruction)ambiguousrigid, boy > girls
Hale 2003's “no movement in PA languages” (p. 12; theory, not data)✓✗✗✓

HYPOTHESIS Count of contradicting data. IA-1* has none, given one of e2/e3/e4. IA-3 has none, given OBV-top. IA-2 has none; it has two internal tensions (the K2 crossing and the freezing clause). IA-0 is contradicted by da- only if da- agrees by structure. Silence counts against none of them. The scope cell is the decider (T10a, T30d).

8. What each analysis gets right and misses

Source / optionGets rightMisses or must add
Hale 1997 (IA-2)The first explicit probe/EPP statement of the Navajo inverse. b- as the elsewhere form and a result of movement, not its cause (PDF p. 18), which fits Willie's cognate remark (p. 28). Topic motivation from speakers and Platero's relative-clause fact (p. 18). Two-step movement to keep locality.The direct needs the agent to cross the raised object (K2). Freezing keeps da- from reaching the agent. It moves the noun DP itself, though the voice is decided with no noun ((39), (72)); a silent pronoun has to stand in. No account of the ranking.
Bruening 2005 (source of IA-1*)Optional [EPP] on v gives the inverse with no long-distance minimality violation. Move-before-Agree explains why Mode/Subject still agrees with the agent in the inverse (W&J's 3NOM). Evidence types that would test raising (binding, crossover, reconstruction).Made for Passamaquoddy. Its movement in the direct would make Navajo M/S agree with the patient. Its plural head counts the raised object, while Navajo da- counts the agent. Its [animate] attractor cannot express Navajo's relative ranking. Its binding tests may not be constructible in Navajo (4th person compared §6.2, W&J (71)).
Hale 2003 (IA-3)No extra movement. da- counts the agent for free. Topic statement of y- (p. 27). Consistent with the subject-over-object scope reported from Rice (secondhand).INV-5 must be stipulated. Hale 1997's c-command statement of OBV fails in situ, so the topic feature has to do the work. Predicts rigid scope in (58), which is untested.
Willie 1991The decisive facts: position change (p. 100), voice with no nominals (pp. 73, 87), da- on the agent (pp. 82–85), the theme excluded (p. 95), definiteness of the inverse subject (p. 77).The Case labels (Ergative/Accusative/Absolutive) are GB devices and disagree with each other (pp. 72 vs 77, 100). In VPISH terms they reduce to which head agrees with which slot, and when.
IA-0 (inverse linking)Gets every string and OBV without movement.da- (unless role-indexed); merge sites by rank rather than by role.

9. Dowty, roles not words, feature bundles, STRATA

Update 2026-10-17 (clause spine L5, §7). HYPOTHESIS Willie & Jelinek 2000:270 (50b) dibé łį́į́ʼ bitah yíghááh ‘The horse went among the sheep’ has the topical inverse bi- on the postposition, outside the verb word. IA-3 spells it in place with no extra assumption. IA-1* needs the raised P object's lower copy to be pronounced, or the PP to be spelled out only after v's EPP (the same escape as K5's low-Pl branch). IA-0 needs bi- to be agreement (e3). This is a derived tension, not a falsification (T31e). W&J (49) (topic/focus matched across a =go sequence) can be stated configurationally under IA-0/1*/2 and featurally under IA-3, so it does not choose.

Dowty baseline. SOURCED Which slot is ρA is fixed by Proto-Agent entailments of the role in the action (Dowty 1991; C5) and is never changed by the voice (INV-1; Willie p. 94; S&J p. 339 (d)). GRANT'S CONJECTURE The animacy ranking is Proto-Agent weight: the referent's capacity for sentience, volition, causation and movement. That makes it an extension of Dowty, not a departure. HYPOTHESIS In the restatement, the inverse arises exactly when the two disagree. The role in the action picks ρA; the referents' capacities pick the higher-ranked slot. RANK→EPP turns the disagreement into a feature choice. Nothing on this page conflicts with that reading. Bruening's denial of a grammatical hierarchy is compatible with it, because in RANK→EPP the comparison chooses the feature and does not constrain the movement (§5.3).

Roles, not words. GRANT'S CONJECTURE + HYPOTHESIS In IA-1*, IA-2 (restated) and IA-3, what is ranked and then moved or marked is a pronoun slot. Under (e2), the distributive probe targets a role feature directly. The decision precedes nouns (K4). The same noun can rank differently (K3). Nouns link afterwards (K1 step 8). Only Hale 1997's notation moves a noun phrase, and the bare-verb facts show that this is not required. The data support the weak thesis (verbs take role-bearing slots; nouns are optional add-ons linked to them). They say nothing new for or against the strong version, which already failed on postpositions.

Feature bundles (FB). HYPOTHESIS The bundle needs person, number (da-), a role index (ρA/ρX), a rank-relevant referent feature (K3), and, in IA-3, [top]. That is the FB-1 to FB-4 bundle of the feature-bundles page, plus rank and topic. No opened datum contradicts treating rank as a bundle feature.

STRATA (theta roles first, event-structure roles next, noun selection later). GRANT'S CONJECTURE → HYPOTHESIS The derivations run in that order. Layer 1, Σ: slots and ρA by entailments. Then the rank comparison over the bundles' referent features. Layer 2: v is merged as an event-structure head and receives [±EPP] (IA-1*), and Qual/E as the “edge” head (Hale 1997 p. 1). The inverse movement happens inside the word. Layer 3, D2: nouns are selected and linked (INV-6), and definiteness and “maximally referential” checks apply (INV-4, DIST-3). One refinement: the rank comparison needs referent information (who the participants are, as construed) before Layer 3, so that information has to sit on the bundles, not on the nouns. The event-structure head (v) and the role layer cooperate here: v hosts the feature, and the roles decide it. The two are compatible descriptions of the same participants, not rivals.

10. Gaps and next (T30a–h)

#TaskNeeds
T30aOPEN In three-participant inverses, is the goal or benefactive PP above the theme inside the verb phrase? IA-1* derives INV-5 only if it is (K6). Joins T28d and T5d.More three-participant verbs with animate themes; Jelinek & Willie 1996 (not opened); Speas 1986 on PPs
T30bOPEN Does Mode/Subject agree with the agent in the inverse? (IA-1*, IA-2 and IA-3 say yes; IA-0 says no.) Inverses are 3-on-3 only, where subject agreement is Ø, so only number can show it.Inverse forms where the agent's number changes the Mode/Subject realization, if any exist; Young & Morgan 1987 paradigms
T30cOPEN Which way does da- find the agent: e2 (role probe), e3 (bi- as agreement head) or e4 (concord through M/S)? (Joins T10b.)da- with fourth-person agents (T10f) and with intransitives; Hale 2003's “argument quantifier” (p. 17) re-read with Rice 2000
T30dOPEN = T10a/T11c: scope in (58)-type sentences, the decider between IA-1*/IA-2 (ambiguous) and IA-3 (rigid)Speaker judgments; Perkins 1978 (no copy found; the box file is a wrong file)
T30eOPEN Does the direct have an EPP raising (Hale 1997 (7)) or none (IA-1*)? A diagnostic would be an adjunct that the raised subject, but not the in-situ one, scopes over (compare Bruening (42)–(43), schematic only).Navajo adjunct-scope data; Hale 2003 pp. 16–19 on preverb scope
T30fOPEN Does Hale 1973 itself state the inversion rule with a topic condition? (= T27a)Kachru et al. (eds.) 1973, pp. 300–309
T30gOPEN Is the inverse subject's landing site Spec,M/SP (IA-1*) or Spec,QualP (Hale 1997's EP)? CYC orders it below Pl either way (update 2026-10-23 ~8:46pm PT: true only if Pl is attached high, at QualP, as in Hale 2003's trees; with Hale's low vP option, a Spec,M/SP or Spec,QualP landing site would be above Pl; see how high is da-? §8, HYPOTHESIS); Qual-level preverb scope over the raised slot would decide.Preverb or qualifier scope judgments with bi-
T30hOPEN Is the object prefix a pronoun (Hale 2003: a maximal projection, ms. pp. 10–11) or agreement on a head (Hale 1997)? This decides e3, and how CYC treats position IV.Hale & Platero 1996, Speas 1990 ch. 4 (neither opened); clitic-doubling diagnostics

Changes made elsewhere for this page. Inverse I1 step 6 now carries an update note (M/S agrees with ρA in both voices under IA-1*). Number & scope §5.1 carries a note pointing to the K5 trilemma. The framework audit T27k row is marked done. See also open questions (T30a–h) and sources.

Sources opened or re-checked for this page (2026-10-16, 8:46pm PT). Hale 1997 workshop notes, PDF pp. 1–2, 9–12, 14–25 (box [local working copy, not in the repository]; page images of pp. 14–16 and 22 checked, (2)–(3) at 400 dpi). Bruening 2005, re-downloaded from the author's page, pp. 1–6, 13–24. Hale 2003 ms. text pp. 10–17, 26–27. Willie 1991, Sandoval & Jelinek 1989 and Willie & Jelinek 2000 forms as recorded on the site with page and example numbers.

Is Navajo bi- a deep inverse? Oxford 2023's tests run on Navajo

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.

Working note. Kept as written (URL, data and status labels unchanged). Filed under Deep or shallow inverse; da-; the consolidated argument is the essay deep-or-shallow. Rule IDs: rules index.

Verbs on this page → dictionary: yiztał ‘kick’ · yiztsʼǫs ‘kiss’ · yishxash ‘bite’ · dabiiłtsą́ ‘see’ · yóóʼeelwod ‘run away’

Restated on the unified structure (2026-11-15). The inverse is restated as one extra Move on the clause spine: the patient moves (via an outer Spec,vP) above the agent, which has already raised to Spec,TP by the working default (H3, H6′, H21b). The event LF is the same as the direct; the trigger and landing site are OPEN. 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. Status on the spine: see unified structure, §4, inverse row. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

The Algonquian inverse page (T52) found that Algonquian has two inverses. A shallow one only changes the agreement marking. A deep one really moves the patient up into the subject slot (Oxford 2023). That page left T52d open: which kind is Navajo bi-? This page runs Oxford's four tests on the Navajo data in Willie 1991, one at a time. In plain words: on every test that can be run, bi- behaves like the deep kind. That holds even in the sentences where animacy forces bi-, because Oxford's Algonquian data show that a forced inverse can still be deep. The one thing that does not follow the patient is the plural prefix da-, which keeps counting the doer. So Navajo bi- is "deep" in where the participants sit, but its plural marking stays tied to the roles.

Update 2026-10-25 (~8:46pm PT): Saxon 2003 (two-osv) is a sourced "deep" account. The bi- object Moves to Spec,DiscP above the 3rd-person subject in Spec,NumP, and the prefix is read from the landing site, unlike A-bar Focus Movement, which keeps yi-. It is a structure, not a test, so D2 stays OPEN.

On this page: 1 Question · 2 Sources · 3 Oxford's tests · 4 Navajo data · 5 The tests run · 6 Forced but deep: rule RCC · 7 What does not flip: da- · 8 Derivations · 9 Grant's ideas · 10 Bottom line · 11 Open

Notation (the site's own; no Syntax · Semantics · Interface bot conventions file exists, so the provisional notation is kept). Σ = role-schema step; A = word level (rule set A: the prefixal, left-headed spine, spelled out as one word W); B = clause level (rule set B: right-headed clause syntax, noun linking, enclitics). W is the switch from A to B. ρA (agent slot), ρP (patient slot): role labels ours. r(ρ) = the animacy rank of the slot's referent as construed (eight ranks). INV-1…INV-7: inverse page. IA-1*: inverse as Agree and EPP. DIST-1: number & scope. Deep / shallow: Oxford 2023's terms (§3 below).

1. The question

  • T52d (from Algonquian inverse §10): is Navajo bi- a deep inverse by Oxford's tests?
  • T52, second half: does Navajo yi-/bi- have any obligatory-Agree, person-hierarchy side (the shallow kind), or only the deep, movement side?

Reminder (Grant mixes these up sometimes): Merge builds structure; moving something already built is internal Merge. Agree copies person and number features from an argument onto a head; nothing moves. A shallow inverse is Agree only. A deep inverse is Agree plus one internal Merge of the patient (Algonquian inverse §6).

2. Sources opened this run (2026-10-23, ~4:46am PT run)

  • Oxford, Will 2023. "A tale of two inverses." Syntax 26: 311–354. Author's preprint (lingbuzz/006567), text layer. Re-read: (4) p. 4; §4 pp. 8–17 (§4.1 contrast, §4.2 binding, §4.3 word order, §4.4 long-distance agreement); (39)–(41) pp. 25–26; §7.3, (44)–(45) and Table 4, p. 29; §8, Table 5, p. 31. Page numbers are the preprint's.
  • Willie, MaryAnn 1991. Navajo Pronouns and Obviation. PhD diss., University of Arizona. Box PDF. Navajo forms below are read from the page images (the text layer garbles Navajo): pp. 71, 77, 82–83, 86–88, 160–162. Text layer only: pp. 14, 76, 84–85, 135–139, 224, and the "no complement clauses" remark on PDF p. 230.

Nothing else was opened. Sandoval & Jelinek, Bruening 2005 and Hale 1997/2003 are cited only as already recorded on earlier pages of this site, and no Navajo form is taken from them here.

3. Oxford's tests SOURCED

Oxford 2023 (4), p. 4: a deep inverse is inverse voice (the patient is the subject); a shallow inverse is inverse alignment (the patient is only indexed like a subject). His §4 gives four ways to tell them apart:

TestDeep (3′→3 in Algonquian)Shallow (3→1/2 in Algonquian)
D1 Contrast (§4.1)A choice between direct and inverse, by topicalityNo choice: fixed by person
D2 Binding (§4.2)Reversed: the patient binds into the agentThe agent still binds into the patient
D3 Word order (§4.3)Follows the patient as subjectFollows the agent
D4 Long-distance agreement (§4.4)Targets the patientTargets the agent

Two refinements matter for Navajo.

  • Forced does not mean shallow. In §7.3 (p. 29), when an obviative acts on a proximate (3obv→3px), the inverse is obligatory, yet it is still the deep kind. Oxford's (45), the Proximate C-command Constraint: "A proximate argument cannot be c-commanded by an obviative co-argument." In Table 5 (p. 31), 3obv→3px is a forced deep inverse, 3obv→3obv an optional deep inverse, and 3→SAP a shallow inverse. He compares the forced case to English "Every mani is loved by hisi mother", where the passive is the only way to get the reading.
  • Agreement in the deep inverse follows the patient. (39)–(41), pp. 25–26: the inverse Voice head "assign[s] inherent Case to its specifier, the agent, and its EPP property attracts the patient to the edge of VoiceP. Infl will then agree only with the patient, as there are no other accessible goals: the agent is inaccessible due to its inherent Case."

4. Navajo data used SOURCED

All forms are from Willie 1991 as printed (page images). The glosses are Willie's own where she gives them. Where she gives only a translation, the word glosses (in square brackets) are ours, following her glossing of the same words elsewhere on the page.

(37a) Willie p. 71
łį́į́ʼdzaanéézyiztał
horsemule3.3.kicked
‘The horse kicked the mule.’
(37b) Willie p. 71
dzaanéézłį́į́ʼbiztał
mulehorseINV3.3.kicked
‘The mule was kicked by the horse.’

Willie p. 71: "The Inverse construction 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."

(46) Willie p. 77
’ashkiiʼatʼéédléiʼbiztsʼǫs
[boygirlINDEFINV3.3.kissed]
‘The boy was kissed by a girl.’
(47) Willie p. 77
*’ashkiiléiʼʼatʼéédbiztsʼǫs
[boyINDEFgirlINV3.3.kissed]
[‘A boy was kissed by the girl.’]

Willie p. 77: "with the Inverse, the external argument (the patient, with Absolutive case) must be definite." Examples (48)–(51) on the same page show the same pattern with łaʼ ‘some’. In the direct voice either argument can be indefinite (pp. 76, 88).

Willie (70), p. 86: "If the agent is higher on the animacy scale than the patient, then the Direct voice is used: if the patient is higher on the scale, then the Inverse must be used. And if they are equal, the choice rests on discourse factors." Same page: "The voice alternation can only be used if all arguments are third person." And: "even when the sentence does not contain any nominals, the sentence must be interpreted with reference to the animacy scale."

(71) Willie p. 87
a.yishxashb.bishxash
3.3.bitINV3.3.bit

Willie (72), p. 87, the readings: (72a) yishxash: 1) he bit him (equal ranked); 2) he bit it (pat. lower ranked); 3) it bit it (lower equals); 4) *it bit him (agent lower). (72b) bishxash: 1) he was bitten by him (equal ranked); 2) he was bitten by it (patient higher); 3) it was bitten by it (lower equals); 4) *it was bitten by him (higher agent). P. 88: "Each paradigm permits what the other excludes."

(17) Willie p. 160
a.’ashkiiyiztsʼǫs1. ‘She kissed the boy.’ 2. *‘The boy kissed her.’
b.’ashkiibiztsʼǫs1. ‘She was kissed by the boy.’ 2. *‘She kissed the boy.’
(18) Willie p. 161
yiztsʼǫs-yę́ęyóóʼeelwod
3.3.kissed=REL3.ran.away
‘The one he kissed ran away.’
(19) Willie p. 161
biztsʼǫs=yę́ęyóóʼeelwod
INV3.3.kissed=REL3.ran.away
‘The one who kissed him ran away.’

Willie (16), p. 160: "The heads of relative clauses are absolutives … the head of the relative is either the intransitive subject or an internal argument of the transitive verb -- and recall that with the Inverse voice, the agent is the internal argument". (19) is agent-headed "because the agent is the internal argument in the Inverse construction" (p. 161).

(20) Willie pp. 161–162
a.ʼatʼééd’ashkiiyiztsʼǫs=yę́ęyóóʼeelwod
[girlboy3.3.kissed=REL3.ran.away]
‘The girl who kissed the/a boy ran away.’
b.ʼatʼééd’ashkiibiztsʼǫs=yę́ęyóóʼeelwod
[girlboyINV3.3.kissed=REL3.ran.away]
‘The girl who was kissed by the boy ran away.’

Willie (21), p. 162: "When both arguments of the relative have nominal adjuncts, the relativizing enclitic binds the external argument, which must be definite." She adds that Platero 1982 reports some speakers allowing two readings, but "this is not true of my speech."

(63c) Willie p. 82
ʼashiikéʼatʼééddayiiłtsą́
boysgirlPL.3.3.saw
‘The boys saw the girl.’
(63d) Willie p. 82
*’ashkiiʼatʼéékédayiiłtsą́
[boygirlsPL.3.3.saw]
(‘The boy saw the girls.’)
(64) Willie p. 83
a.dabiiłtsą́‘They/he (was) were seen by them/*him.’
PL.INV3.3.saw
b.shizhéʼéʼashiikédabiiłtsą́
1s.fatherboysPL.INV3.3.saw
‘My father was seen by the boys.’

Willie p. 82: "the agent falls under the scope of the distributive marker". In (64b) the agent (the boys) is plural and the patient (my father) is singular, and the sentence is good. P. 84 adds that with dabi- the external argument "may optionally be interpreted as plural", and a plural external argument can escape the definiteness requirement.

5. The four tests, run on Navajo

TestNavajo factsResult
D1 ContrastEqual rank: free choice by discourse ((70), (37), (72a.1/b.1, a.3/b.3)). Unequal rank: forced ((70), (72a.4/b.4)).SOURCED facts. HYPOTHESIS This matches Oxford's 3obv→3obv (optional, deep) plus 3obv→3px (forced, deep), not his 3→SAP (forced, shallow). See §6.
D2 BindingNo Navajo binding data in any source opened. T11b (fourth person compared) found the test may not be constructible. Update 2026-10-24 ~12:46pm PT: possessor coreference data found (Young & Morgan 1980 via Aissen 2000): ‘Mr. Longi's wife looks at himi’ needs bi-, ‘the boyi saw hisi horse’ needs yi- (possessor-binding).OPEN Silence, so it counts neither way. Update: HYPOTHESIS the four cells fit a deep bi- (Oxford (45)), but obviation and word order predict them too, and the antecedents are names, not quantifiers. Still OPEN.
D3 Word order and noun linkingThe noun next to the verb is the internal argument in both voices: the patient with yi-, the agent with bi- ((37), (17)). The outer noun is the external argument: the agent with yi-, the patient with bi-.SOURCED facts (Willie pp. 71, 160). HYPOTHESIS Order follows position, and the positions swap with voice: deep-like. Caveat below.
D4 Long-distance agreementNavajo has "no complement clauses … no long-distance Wh- movement" (Willie, PDF p. 230, text layer).SOURCED Not applicable.
D4′ Substitute: relativizationThe relative head is the internal argument when it has no noun ((18) patient, (19) agent) and the external argument when both have nouns ((20a) agent, (20b) patient).SOURCED facts. HYPOTHESIS Which participant can head the relative changes with voice, so it tracks positions, not roles: deep-like.
Extra: inverse pragmaticsThe inverse subject must be definite ((46)/(47), p. 77; p. 88), unless da- licenses a plural (p. 84).SOURCED This fits Oxford's deep inverse, where the patient is the more topical argument.
Shallow side?"The voice alternation can only be used if all arguments are third person" (p. 86).SOURCED No 3→1/2 inverse, so there is nothing for a shallow inverse to be.

HYPOTHESIS Caveat on D3 and D4′. On Willie's own analysis the nouns are adjuncts linked to the verb's pronouns. A purely morphological inverse could in principle tell the hearer "the noun next to the verb is the agent" without anything moving. So D3 and D4′ show that the positions swap, but they can't by themselves rule out a morphology-only story. Oxford's decisive test is binding (D2), and that one is open. The definiteness fact helps: a morphology-only inverse gives no reason for the patient to have to be definite, whereas a raised subject that is the topic does.

6. Forced but deep: rule RCC HYPOTHESIS

In INV-3 the site already states the ranking as "EXT = the slot whose referent ranks higher". Oxford's (45) suggests restating it as a constraint on structure, which makes the parallel to the Algonquian forced deep inverse exact:

RCC (Rank C-command Constraint; ours, after Oxford 2023 (45)). In a word whose verb has two third-person slots that the voice compares, a slot whose referent ranks higher may not be c-commanded by a co-argument slot whose referent ranks lower, at the point where W is sealed.

Checked against all eight readings of (72):

(72)RanksStructure at SealRCCWillie
a.1 he bit himA = PρA over ρPokok
a.2 he bit itA > PρA over ρPokok
a.3 it bit itA = PρA over ρPokok
a.4 it bit himP > Alower ρA over higher ρPviolated*
b.1 he was bitten by himA = PρP raised over ρAokok
b.2 he was bitten by itP > AρP raised over ρAokok
b.3 it was bitten by itA = PρP raised over ρAokok
b.4 it was bitten by himA > Plower ρP raised over higher ρAviolated*

HYPOTHESIS 8 of 8 match. RCC predicts the same facts as INV-3 for two-slot verbs, so this is a restatement, not a new prediction there. What it adds: (i) the obligatory bi- comes out as a deep inverse forced by a ban on bad c-command, exactly Oxford's account of 3obv→3px. Being obligatory is therefore no sign of a shallow, Agree-only inverse. (ii) The answer to T52's second half: Navajo's only obligatory side is this rank-forced deep inverse, not a person-hierarchy Agree inverse.

OPEN How hard is RCC? Oxford's PCC is a grammatical rule in Algonquian. For Navajo, the site has already recorded that Sandoval & Jelinek call a rank-violating sentence "pragmatically unacceptable" rather than ungrammatical, and that only about half of speakers in Young's survey obey the hierarchy (via Willie p. 16; inverse page, INV-3). So RCC may be a soft constraint, or it may vary between speakers (T5b, carried into T53b).

OPEN Three slots. RCC as stated would also apply to a theme that outranks a slot above it. The site's INV-5 says a theme never competes (Willie pp. 94–95). Oxford's ditransitive data (46)–(47), p. 29, show a parallel restriction in Algonquian (only the recipient takes part in the rank comparison). Whether RCC needs a clause excluding themes, or whether themes are low enough that it never matters, is untested (T53c).

7. What does not flip: da- counts the doer

SOURCED In Oxford's deep inverse the agreement head Infl agrees with the raised patient, because the agent is made inaccessible (pp. 25–26). In Navajo the distributive plural da- counts the agent in both voices: (63c)/(63d) in the direct, (64b) in the inverse, where the plural agent and singular patient are fine with dabiiłtsą́ (Willie pp. 82–83). The site has already recorded this as DIST-1 (number & scope).

HYPOTHESIS So Navajo bi- is not a copy of Oxford's deep inverse in every respect. Its positions swap (§5), but the plural agreement stays with the agent. This is what the site's IA-1* already says, following Bruening 2005's footnote 22 as recorded on the inverse as Agree and EPP page: the patient slot moves first, and the person/number head then agrees with the agent left below it. In Oxford's terms, the Navajo agent does not get the inherent Case that would hide it from agreement. Navajo is closer to Bruening's Passamaquoddy analysis than to Oxford's Voiceerg.

HYPOTHESIS Classification. Navajo bi- = a deep inverse in position (D1, D3, D4′, pragmatics), with agreement tied to the roles. It has no shallow counterpart at all.

OPEN P. 84 says the external argument of dabi- "may optionally be interpreted as plural". Whether da- can ever count the raised patient alone (plural patient, singular agent) is not shown by the examples opened (T53d). Update 2026-10-23 ~12:46pm PT: answered, no: see can da- count the patient? (Willie (64a) ‘by *him’, (65b), (67b)). There the optional patient plural is treated as a second, closest-goal probe that reaches the patient only after it has moved (PL-2H; HYPOTHESIS).

8. Derivations

Steps are labelled by rule set: Σ role schema, A word (W), B clause. Each step names its operation: Merge (builds structure), Agree (values person/number), Move (internal Merge), Link (a noun linked to a slot), Seal (spell-out of a unit). VPISH: both participant slots are first-merged inside the verb's phrase, ρA in Spec,v and ρP as complement of V. The spine is the site's Qual > M/S > v > V, with Voice (yi-/bi-) above QualP. No flat-S tree is used. Verb words are not segmented beyond Willie's gloss line.

D-1. (37a) łį́į́ʼ dzaanééz yiztał (direct, equal rank)

  1. Σ Σ(√KICK) = {ρA (the kicker), ρP (the one kicked)}. Theta roles are fixed here and at external Merge, and nothing later changes them.
  2. A Merge V with ρP; Merge v, ρA in Spec,v (VPISH). v has no EPP (IA-1*, direct).
  3. A Merge M/S; Agree M/S ⟶ ρA [3] (the closest goal). Merge Qual; Merge Voice.
  4. A RCC check: r(ρA) = r(ρP) ("of equal animacy", Willie p. 72). ρA c-commands ρP, and that is allowed. Voice spells out as yi- because ρA is the highest slot. Seal D1 → W = yiztał, as printed.
  5. B Merge the nouns and Link them (INV-6): dzaanééz, next to W, links to the internal slot ρP; łį́į́ʼ links to the external ρA. Seal D2: łį́į́ʼ ⌢ dzaanééz ⌢ W ✓ ‘The horse kicked the mule.’

D-2. (37b) dzaanééz łį́į́ʼ biztał (inverse, equal rank: optional deep inverse)

  1. Σ As in D-1: the same two roles.
  2. A Merge V with ρP; Merge v[EPP], ρA in Spec,v. Discourse chose the EPP version (equal rank).
  3. A Move: v's EPP raises ρP to outer Spec,v, over ρA (the "crossing" of the Algonquian page). This is internal Merge, not Agree.
  4. A Merge M/S; Agree M/S ⟶ ρA [3], the agent left below (IA-1*; this is where Navajo differs from Oxford's Voiceerg, §7). Merge Qual, Voice.
  5. A RCC check: equal ranks, so ρP c-commanding ρA is allowed. Voice spells out as bi-. Seal D1 → W = biztał.
  6. B Link (INV-6): łį́į́ʼ, next to W, links to the internal slot, now ρA; dzaanééz links to the external slot, now ρP. Seal D2 ✓ ‘The mule was kicked by the horse.’ The horse is still the kicker.

D-3. (72a.4) and (72b.4): the two stars, schematic (no nouns)

  1. Σ Σ(√BITE) = {ρA, ρP}. Referents as construed: (72a.4) ρA = "it", ρP = "him", so r(ρP) > r(ρA). (72b.4) the reverse.
  2. A (72a.4) Merge with v without EPP: ρA c-commands ρP. Agree M/S ⟶ ρA. RCC: the higher ρP is c-commanded by the lower ρA → violated. Seal would give yishxash, but not on this reading: * ✓.
  3. A (72b.4) Merge with v[EPP]; Move ρP over ρA. RCC: the higher ρA is c-commanded by the lower, raised ρP → violated. Seal would give bishxash, but not on this reading: * ✓.
  4. B No B step: there are no nouns. Both stars are made inside rule set A, before W is sealed, which fits Willie's remark that the scale applies "even when the sentence does not contain any nominals" (p. 86).

D-4. (19) biztsʼǫs=yę́ę yóóʼeelwod: relativization tracks the raised structure

  1. Σ Σ(√KISS) = {ρA, ρP}.
  2. A As D-2: v[EPP] and Move ρP over ρA; Agree M/S ⟶ ρA; RCC satisfied (no rank conflict is given). Seal D1 → W = biztsʼǫs.
  3. B Merge =yę́ę (REL). Willie (16): it binds the internal argument (absolutive) when there are no nouns. After the Move in step 2, the internal slot is ρA. So the head is the kisser: ‘The one who kissed him ran away’ ✓. With yi- (no Move) the internal slot is ρP, giving (18) ‘The one he kissed’ ✓.
  4. B Merge the relative as the subject of yóóʼeelwod and Seal D2 ✓. HYPOTHESIS The enclitic reads the structure built in rule set A, so the deep-inverse Move is visible to a rule-set-B operation. That is a point where the word boundary does not hide A's work from B.

9. Grant's ideas

  • GRANT'S CONJECTURE Theta roles are a proto-role schema, fixed first; the inverse is internal Merge of the patient over the agent, and the ranking is over proto-agent capacity.
  • SOURCED Willie p. 71: the inverse "is a transitive sentence that has both agent and patient arguments", and the translations of (37a)/(37b) describe the same event. Nothing opened contradicts "roles fixed first".
  • HYPOTHESIS The deep-inverse result supports the "internal Merge over the agent" half of Grant's picture, but only for position. The da- fact (§7) shows that one agreement process ignores the Move and keeps tracking the role. That doesn't contradict Grant's idea. If anything, it suggests the agent's role bundle stays visible after the patient crosses it.
  • HYPOTHESIS Feature bundles: RCC must read ranks with no nouns present (D-3; p. 86), so the rank has to sit on the slots, the bundles, before any noun is linked in rule set B. This is consistent with Grant's bundle idea and with the remark on the inverse as Agree and EPP page that rank information has to be on the bundles. No opened data contradict it.
  • OPEN Proto-agent capacity vs animacy: Willie states the ranking as animacy (70). Whether that is a proto-agent measure is T52a/T5 territory and is not settled here.

10. Bottom line

HYPOTHESIS T52d answered as far as the data go: by every one of Oxford's tests that can be run on Navajo (contrast, word order/linking, relativization as a substitute for long-distance agreement, plus definiteness), bi- is a deep inverse. The patient really takes the subject position.

HYPOTHESIS Forced but deep. The obligatory bi- (patient outranks agent) is the Navajo counterpart of Oxford's forced deep inverse, not of the shallow SAP inverse. New rule RCC states it as a ban on a higher-ranked slot being c-commanded by a lower one, and fits all eight readings in (72).

SOURCED No shallow side. Navajo has no inverse for third person acting on first or second person (Willie p. 86).

SOURCED + HYPOTHESIS Not identical to Algonquian. da- counts the agent in both voices (Willie (63)–(64)), whereas Oxford's deep inverse makes agreement follow the patient. Navajo looks more like Bruening's Passamaquoddy analysis (IA-1*).

OPEN Binding, the decisive test, has no Navajo data in any source opened, and RCC's grammatical status (hard vs soft vs varying between speakers) is open. Update 2026-10-24: possessor coreference data now found; they fit "deep" but do not decide it (possessor-binding §5–6).

OPEN Chemistry. No source opened mentions chemistry, and this page claims no chemistry analog.

11. Open questions this page adds

#QuestionWhat would settle it
T53aBinding/scope under bi-: can the raised patient bind into or take scope over the agent (Oxford's D2)?Speaker judgments on quantifier/possessor sentences; T10a, T11b. OPEN Update 2026-10-24 ~12:46pm PT: possessor data found and they fit "deep", not decisive; see possessor-binding (T57a–d).
T53bIs RCC a hard grammatical constraint (as Oxford's PCC is), a soft one, or one that varies by speaker?Young's survey itself; Sandoval & Jelinek re-read; Willie (73). OPEN
T53cDoes RCC need a clause exempting themes (INV-5), parallel to Oxford's ditransitive restriction?A three-slot verb with a theme that outranks the agent or goal, from a source. OPEN
T53dCan da- ever count the raised patient alone (plural patient, singular agent) under bi-?Closed 2026-10-23 ~12:46pm PT: no (can da- count the patient? §4). SOURCED
T53eClause chains with =go (Willie pp. 135–139, (9)): does the "switch in thematic roles" track position (deep) or role? The subscripts in the text layer are damaged and conflict with T31b.A clean scan of Willie pp. 136–137. OPEN
T53fPlatero 1982's speakers who allow two readings of a two-noun relative: do they treat bi- as less deep?Platero 1982 (not opened). OPEN

Builds on the Algonquian inverse (T52), the inverse (T5), the inverse as Agree and EPP (IA-1*), number & scope (DIST-1), relative clauses and fourth person compared (T11).

The Algonquian inverse: same roles, a different subject slot

Working note. Kept as written (URL, data and status labels unchanged). Filed under Deep or shallow inverse; da-; the consolidated argument is the essay deep-or-shallow. Rule IDs: rules index.

Restated on the unified structure (2026-11-15). The inverse is restated as one extra Move on the clause spine: the patient moves (via an outer Spec,vP) above the agent, which has already raised to Spec,TP by the working default (H3, H6′, H21b). The event LF is the same as the direct; the trigger and landing site are OPEN. 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, §4, inverse row. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

The Morphic Tessellations bot reported "some kind of theta phenomenon ... crossing flip in inverse voice" in Algonquian languages, and Grant asked us to look into it. Algonquian verbs (Cree, Ojibwe, Passamaquoddy, Blackfoot and their relatives) have a direct form and an inverse form for the same event. In plain words: the inverse never changes who did what to whom. It changes which participant controls the verb's "subject" agreement. When a third person acts on "me" or "you", the inverse is obligatory and purely a matter of marking. When one third person acts on another, it is an optional voice, and the patient is actually raised above the agent, but unlike a passive the agent stays a full argument (Oxford 2023; Bruening 2005). "Crossing flip" is the Morphic bot's own name for this, not a linguistics term. This page compares it with Navajo yi-/bi- (inverse page, T5) and opens T52.

On this page: 1 Question · 2 Sources · 3 The pieces · 4 Examples · 5 Two inverses · 6 Merge or Agree? · 7 Grant's ranking thesis · 8 Navajo compared · 9 Bottom line · 10 Open

1. The question

GRANT'S CONJECTURE (relayed 2026-10-22) "some algonquin languages have some kind of theta phenomenon ... crossing flip in inverse voice or something... please investigate and keep track of theta phenomenon for if they pertain to chem stuff".

  • Q1. Is the Algonquian inverse a change in argument structure (different theta roles, different valency, different structure), or the same roles ranked and marked differently?
  • Q2. What is the "crossing flip"?
  • Q3. How does it bear on Grant's role-ranking thesis and on Navajo yi-/bi-?

Chemistry is tracked in Grant's research notes (lit/theta-phenomena-tracker.md; public summary on the Chemical-Theta-Roles tracker page). This page is linguistics only.

2. Sources opened this run (2026-10-23, early morning)

  • Oxford, Will 2023. "A tale of two inverses." Syntax 26: 311–354. Author's preprint (lingbuzz/006567), text layer; pp. 1–33 and the conclusion read.
  • Oxford, Will 2019. "Inverse marking and Multiple Agree in Algonquin: Complementarity and variability." NLLT 37: 955–996. Preprint (lingbuzz/004529); abstract, introduction and Table 1 read.
  • Oxford, Will. "Algonquian." Routledge handbook chapter, author's PDF; §§4.2–4.4 read (alignment, obviation, verb template).
  • Bruening, Benjamin 2005. "The Algonquian Inverse is Syntactic: Binding in Passamaquoddy." ms., University of Delaware. The box copy already used on fourth person compared §6; pp. 1–3 re-read.
  • Bliss, Heather 2005. "Topic, Focus, and Point of View in Blackfoot." WCCFL 24: 61–69 (Cascadilla Proceedings Project). Text layer; pp. 61–66 read.
  • Willie, MaryAnn 1991. Printed pp. 96–97 (page images) and p. 216, re-read for §8.
  • Aissen, Judith 1997. "On the Syntax of Obviation." Language 73: 705–750. Downloaded, but it is an image scan with no text layer. Only the abstract was read; it is not used for any claim beyond naming her approach.

Not opened (quoted only through the papers above, marked "via"): Klaiman 1991; Givón 1994; Gildea 1994; Wolfart 1973; Wolvengrey 2011; Dahlstrom 1986, 1991; Rhodes 1976, 1994; Bruening 2001; Lochbihler 2012; Bliss 2013; Zúñiga 2006, 2008; Nichols 1980; Bloomfield 1946, 1962; Valentine 2001; Hamilton 2015; McLean 2001; Oxford 2024 (LI 55, abstract only); Oxford's TWPL paper on grammar "myths" (the download hung).

3. The pieces SOURCED

3.1 The person hierarchy

Bruening 2005 (2), from Goddard & Bragdon 1988: "Participant Hierarchy: 1,2 > 3 > Obv > Inan". Oxford's Routledge chapter gives 1/2 > X > 3 > 3′ > 0 for the Ojibwe independent order (X = impersonal), but only 3 > 3′ > 0 in the conjunct (Rhodes 1994: 432, via Oxford). Most Algonquianists reject a "local inverse" for 2 acting on 1 (Oxford 2019, citing Bloomfield, Goddard, Nichols, Valentine and others). The 2 > 1 effect is real only in the person prefix.

3.2 Obviation

"When a clause contains multiple third-person referents, one, normally the most topical, is designated as proximate (notated 3 or PROX); any others must be obviative" (Routledge §4.3). Obviation is only for animate third persons (Bruening 2005: 2).

3.3 Theme signs and the verb template

The theme sign follows the stem. "The theme sign normally marks person/animacy agreement with the logical object … However, a special INVERSE theme sign -ikw appears instead when the logical object outranks the logical subject" (Routledge §4.4). Oxford, following Nichols 1980, gives the Ojibwe template as prefix · stem · theme sign · negation · central suffix · mode · peripheral suffix. For example, ni-waapam-ikw-ssii-naa-token-ak 'they might not see us' (Routledge Figure 5).

ConfigurationConjunctIndependent
2→1-i-i
1→2-in-in
1/2→3-a:-a:
3→1-i-igw
3→2-in-igw
3→3′-a:-a:
3′→3-igw-igw

Algonquin (Ojibwe) theme signs, Oxford 2019 Table 1 (data from Jones 1977). The inverse -igw is in bold. In the conjunct, 3→1/2 uses the ordinary object signs, not the inverse.

4. Examples SOURCED

All forms are copied from the opened papers. The original fieldwork or grammar is named after "via".

(A1) Menominee, Bloomfield 1962: 39, via Oxford, Routledge (10)
a.nɛ·w-ɛ·-wenoh enɛ·niwanenoh metɛ·mohs-an.
see-DIR-3.PXthat.PX man.PXthis.OBV woman-OBV
'The man (PROX) sees the woman (OBV).'
b.nɛ·w-ek-wenoh enɛ·niwanenoh metɛ·mohs-an.
see-INV-3.PXthat.PX man.PXthis.OBV woman-OBV
'The woman (OBV) sees the man (PROX).'

"Despite the reversal of thematic relations, the proximate enjoys the same morphosyntactic prominence in both sentences: it consistently precedes the obviative and is agreed with by the verb." Oxford adds that one can read this as split ergativity (Déchaine 1999, via Oxford): accusative alignment in the direct form, ergative in the inverse.

(A2) Plains Cree, SAP inverse (obligatory), Wolfart 1973 via Oxford 2023 (15)
a.ni-waːpam-aː-inaːn-ak
1-see-3OBJ-1PL-3PL
'we see them' (1→3, direct)
b.ni-waːpam-ikw-inaːn-ak
1-see-INV-1PL-3PL
'they see us' (3→1, inverse)

Oxford 2023: 10: "Each of the forms in (15) is the only way to express that particular configuration of arguments." The prefix ni- '1' is the same in both. The theme sign alone tells you who acts on whom.

(A3) Plains Cree, third-person inverse (optional), Wolvengrey 2011: 175 via Oxford 2023 (14)
a.Caːniykiː-wiːcih-eː-wMeːriy-wa.
Johnny.PX.SGPST-help-3OBJ-3SGMary-OBV
'Johnny.PX helped Mary.OBV.' (direct)
b.Caːniy-wakiː-wiːcih-ikw-wMeːriy.
Johnny-OBVPST-help-INV-3SGMary.PX.SG
'Johnny.OBV helped Mary.PX.' / 'Mary.PX was helped by Johnny.OBV.' (inverse)

The same event either way. The choice depends on "situation-specific, speaker-determined assignment of relative topicality" (Wolvengrey 2011: 176, via Oxford 2023). Oxford's (13) shows the same choice with two obviatives: 'the rabbit saw the wolf' (direct) vs 'the wolf was seen by the rabbit' (inverse).

(A4) Oji-Cree binding, Oxford's fieldwork, Oxford 2023 (18a), (19a)
a.Kahkina ihkweːwak ot-oːciːm-aː-waː-an naːpeːwan kaː꞊saːkih-ikw-waːc.
all woman.PX.PL 3-kiss-3OBJ-3PL-OBV man.OBV REL꞊love-INV-3PL
'[All the women.PX]i kiss the man.OBV who loves themi.' (direct: the agent binds into the patient)
b.Kahkina ihkweːwak ot-oːciːm-ikw-waː-an naːpeːwan kaː꞊saːkih-ikw-waːc.
all woman.PX.PL 3-kiss-INV-3PL-OBV man.OBV REL꞊love-INV-3PL
'The man.OBV who loves themi kisses [all the women.PX]i.' (inverse: the patient binds into the agent)

Only the theme sign differs (-aː vs -ikw). In (18b)/(19b) the reverse binding is impossible. Oxford 2023: 10 reports the same reversal "in all Algonquian languages in which relevant data is available": Plains Cree, Western Naskapi, Passamaquoddy, Ojibwe, Blackfoot, Meskwaki (those sources were not opened here). It is the pattern of an English active vs passive.

(A5) Passamaquoddy, Bruening 2005 (1), (3)
a.K-ucem-a-k.
2-kiss-Dir-3P
'You kissed them.'
b.K-ucem-ku-k.
2-kiss-Inv-3P
'They kissed you.'
c.Mahtoqehsʼ-toli-nuhsuphoqal-ku-lmuwinuw-ol.
rabbit3-Prog-chase-Inv-Obvbear-Obv
'A rabbit (Prox) was being chased by a bear (Obv).'

Bruening 2005: 2: "it is important to note that this is not a passive. There is no loss of an external argument here, and no change in valence."

(A6) Blackfoot, Bliss 2005 (3c, d, g, h)
c.nit-aawayaki-a-(w)ad.nit-aawayaki-ok-(w)a
1-hit-DIR-PROX1-hit-INV-PROX
'I hit him''He hit me'
g.aawayaki-(y)ii-(w)ah.ot-aawayaki-ok-(w)a
hit-DIR-PROXOBV-hit-INV-PROX
'He hit her''She hit him'
(A7) Blackfoot, a sentient obviative acting on a non-sentient proximate, Bliss 2005 (11)
Ana pokónaaniotaííspaapiksistiimáóhksinaa
an-(w)a pokon-(w)a
DEM-PROX ball-PROX
an-(y)i
DEM-OBV
ot-a-(y)iispaapiksist-(y)ii
OBV-DUR-throw-DIR
maohksinaa
be.red
'The ball that he is bouncing is red'

The obviative person acts on the proximate ball, yet the theme is direct (-yii). Bliss argues that Blackfoot ranks third persons by sentience (SENT > NON-SENT), not by PROX > OBV.

(A8) Blackfoot, a knife cannot be an agent, Bliss 2005 (9)
Amo isttoána iksikisáíikin iihtáíistiniʼpa ani íʼksisakoyi
DEM knife-PROX COP-sharp MEANS-DUR-cut-UNSPEC-PROX DEM-OBV meat-OBV
'The knife is sharp; someone cut the meat with it'

"Non-sentient nouns, like isttoán 'knife' cannot be agentive in Blackfoot." The knife is "treated as an adjunct, as indicated by the iiht- 'means' prefix on the matrix verb", and the agent is unspecified (-oʼp). Compare the Navajo instrument bee, which does not take part in yi-/bi- (instrument and voice).

5. Two inverses, not one SOURCED

Oxford 2023 builds on a distinction from Gildea 1994 and Givón 1994 (via Oxford): "inverse alignment is a morphological pattern in which the patient is indexed like a canonical subject, while inverse voice is a syntactic construction in which the patient is the subject." A deep inverse is both. A shallow inverse is only the morphology. "The two kinds of inverses can coexist within a language (Givón 1994: 29)."

SAP inverse (3 → 1/2): shallowThird-person inverse (3′ → 3): deep
Is there a choice?No: fixed by person (A2)Yes: by topicality (A3)
BindingThe agent still binds into the patient (Passamaquoddy, Bruening 2001: 131, via Oxford 2023 (22))Reversed: the patient binds into the agent (A4)
Word order, long-distance agreementLong-distance agreement targets the agent (Mi'kmaq, Hamilton 2015, via Oxford)They follow the patient (East Cree, Passamaquoddy, Mi'kmaq, Odawa, via Oxford)
MechanismAgree on Infl plus impoverishment: "Delete phi-agreement from Voice when phi-agreement on Infl is an exact duplicate" (Rule 35). The inverse marker is Voice's elsewhere form.A special Voice head gives the agent inherent case and has an EPP feature, so the patient raises over the agent.
Is the agent demoted?NoNo: not a passive
Theta rolesUnchangedUnchanged

Oxford 2019 had already derived the inverse marker from an Agree outcome: "When Infl agrees only with the object, it duplicates the result of an earlier object agreement operation on Voice. The presence of identical features on Infl and Voice triggers an impoverishment operation that deletes the features of Voice, resulting in its spellout as an underspecified elsewhere form—which is the exponent that we know descriptively as the inverse marker." When Infl agrees with both arguments (Multiple Agree), a portmanteau appears instead, so the two never co-occur.

Bruening 2005 sorts the theories into three families. (i) Inverse linking (Rhodes): the arguments are projected in reverse order, which is the only family on which the inverse changes argument structure. (ii) Morphological inverse (Aissen 1997): the syntax stays constant and the morphology indexes the hierarchy. (iii) Syntactic movement: "it involves a step of movement in the syntax, such that the object rather than the subject ends up in the highest position". Oxford 2023 in effect gives (ii) to the SAP inverse and (iii) to the third-person inverse.

So what is the "crossing flip"? HYPOTHESIS

The phrase is not in any source opened. The Morphic bot uses it for its own braid picture: two strands, one crossing, flipped (σ vs σ⁻¹). The nearest real content is the deep inverse. There, the agent and patient swap their structural order (the patient crosses over the agent) while their theta roles stay put. In the shallow inverse nothing crosses: only the agreement spell-out changes.

6. Merge or Agree? HYPOTHESIS (standard definitions; see Merge and Agree: the originals)

  • Merge builds structure. External Merge puts the agent and patient in their theta positions; this is where theta roles are fixed. Internal Merge (movement) puts an argument that is already there in a higher position.
  • Agree copies features. A probe (here Infl or Voice) finds the closest matching argument and takes its person/number. Nothing moves.
  • Shallow inverse = Agree (plus impoverishment in the morphology). Deep inverse = Agree plus one internal Merge of the patient. External Merge, and so the theta roles, is the same in direct and inverse. Only the inverse-linking family would put the difference in external Merge, and binding (A4) and Bruening's Passamaquoddy data argue against that.

7. Grant's ranking thesis

  • GRANT'S CONJECTURE Verbs rank their participants. When the natural ranking is crossed, a marked voice appears, and some combinations are ruled out.
  • SOURCED Algonquian has the shape of that rule: fixed roles, a ranking, and a marked form when the patient outranks the agent.
  • SOURCED The ranking itself is over person and obviation (1/2 > 3 > 3′ > 0). In 3/3 clauses obviation follows topicality. It is not a ranking of proto-role weight.
  • HYPOTHESIS Blackfoot comes closest to Grant. Bliss ranks third persons by sentience (A7), and a non-sentient noun cannot be an agent at all (A8). Sentience is one of Dowty's proto-agent entailments (proto-role schema). So Blackfoot ranks on one proto-agent property, but only one, not a weight.
  • OPEN Is any Algonquian inverse sensitive to a patient property (change of state, incremental theme)? Nothing opened says so (T52a).

8. Navajo yi-/bi- compared

  • SOURCED Willie 1991: 96–97 on Navajo: "the Inverse marking functions like obviative marking. However, note that the Inverse has a further property: it always makes the argument with an agent theta role the internal argument." P. 216: on a postposition the yi-/bi- alternation "is not one of voice, but shows the use of this alternation to mark obviation."
  • SOURCED Navajo yi-/bi- is limited to clauses with two third persons (inverse page). Navajo has no inverse for a third person acting on "me" or "you".
  • HYPOTHESIS So Navajo has a counterpart only to the Algonquian deep inverse. It is optional, ranked by animacy and topicality, and on Willie's wording it re-positions the agent. There is nothing like the shallow SAP inverse. This agrees with the inverse as Agree and EPP page (T27k), which already uses Bruening's Voice-[EPP].
  • OPEN Whether bi- shows Oxford's deep-inverse tests (binding reversal as in A4). T11b/T11c (fourth person compared) found that binding tests may not be constructible in Navajo; the scope test (T10a) remains.

9. Bottom line

SOURCED Theta roles never change. In every Algonquian direct/inverse pair opened (A1–A8), the agent is the agent and the patient is the patient. The inverse changes which one controls the subject agreement, and Bruening stresses it is "not a passive".

SOURCED Two inverses. 3→1/2: obligatory, Agree only, no movement (shallow). 3′→3: optional voice, the patient raised over the agent, binding reversed (deep) (Oxford 2023).

HYPOTHESIS "Crossing flip" is the Morphic bot's own braid term. It fits only the deep inverse: a swap of structural order with the roles fixed.

HYPOTHESIS Merge vs Agree. Shallow = Agree; deep = Agree plus internal Merge. Theta roles are set at external Merge, which the inverse leaves alone.

HYPOTHESIS Grant's thesis. Algonquian ranks persons, not proto-roles. Blackfoot's sentience ranking (Bliss 2005) is the one place it touches a Dowty entailment.

HYPOTHESIS Navajo. yi-/bi- lines up with the deep inverse only (Willie 1991: 96–97).

OPEN Chemistry. None of these sources mentions chemistry, and this page claims no chemistry analog.

10. Open questions this page adds

#QuestionWhat would settle it
T52aDoes any Algonquian inverse respond to a patient proto-property (affectedness, incremental theme), not only to person/obviation/sentience?Oxford 2024 (LI 55, twelve inverse distributions; abstract only so far); Bliss 2013. OPEN
T52bBlackfoot binding with a sentient obviative acting on a non-sentient proximate (A7 type): is the direct form deep or shallow?Bliss 2013 §7.4.1 (via Oxford 2023; not opened). OPEN
T52cDoes the Klaiman 1991 / Givón 1994 typology of "inverse voice" support the voice label for the third-person inverse in its own terms?Klaiman 1991; Givón (ed.) 1994. Neither opened. OPEN
T52dIs Navajo bi- a deep inverse by Oxford's tests? (= T10a/T11c, now with Oxford's list of diagnostics)Speaker judgments on scope; see number & scope. OPEN Update 2026-10-23 ~4:46am PT: run on deep or shallow inverse?: deep-like on every runnable test HYPOTHESIS; binding still open.

Builds on the inverse (T5), the inverse as Agree and EPP (T27k), fourth person compared (T11) and instrument and voice (T50a).

Two ways to put the object first: Saxon 2001 on bi- inversion vs Focus Movement, and whether the fronted patient sits in an A-position (T60c)

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.

Working note. Kept as written (URL, data and status labels unchanged). Filed under Possessors and object-first orders; the consolidated argument is the essay possessors. Rule IDs: rules index.

Verbs on this page → dictionary: dabiiłtsą́ ‘see’

Restated on the unified structure (2026-11-15). The inverse is restated as one extra Move on the clause spine: the patient moves (via an outer Spec,vP) above the agent, which has already raised to Spec,TP by the working default (H3, H6′, H21b). The event LF is the same as the direct; the trigger and landing site are OPEN. 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. Flat-S and other pre-VPISH notation below is labelled contrast. Status on the spine: see unified structure, §4, inverse row. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

The last page left a fork open (T60c). When bi- puts the unspoken person acted on first, is that person in an A-position (an argument-type position, as the "deep" inverse says) or in a topic position (A′, outside the argument system)? This run opened Leslie Saxon's "On Two OSV Constructions in Navajo: Beyond Subject-Object Inversion" (WECOL 2001, published 2003), a paper not used on this site before. In plain words, Saxon shows that Navajo has two different ways to put the object before the subject. In the familiar one the verb switches to bi- and the object must be definite. In the other, which she calls Focus Movement, the verb keeps yi-, the object can be indefinite, and question words front the same way. She treats the second as the A′ kind of movement, and keeping yi- is the sign that the object is "read back" in its old place. The first she puts in a separate, lower landing spot that also hosts 'I'/'you' subjects, and nothing there is read back. That is the A-type behavior the deep inverse needs (our reading, HYPOTHESIS). She also gives Young & Morgan sentences with two bi-s in one clause when the subject is a thing (a boat, a wagon). That is contradicting data for the "one proximate per clause" and "one topic per clause" accounts this site has been leaning on.

On this page: 1 Question · 2 Source · 3 The data · 4 Her analysis · 5 What this does to T60c and D2 · 6 Two bi-s: contradicting data for one-proximate · 7 Rulesets A and B · 8 Rules · 9 Derivations · 10 Grant's ideas · 11 Bottom line · 12 Fidelity notes · 13 Open questions

Notation as on possessor-command. Σ = the role schema. A = the word level (prefixal, left-headed; W = the verb word). B = the clause level (right-headed). ρA/ρP = agent and patient slots. Merge vs Agree: Merge builds structure: it puts two pieces together, and Move is Merge again of something already built (here: the object re-Merged in Spec,DiscP or Spec,CP). Agree values features: it checks or copies person and number between a head and a bundle (here: whether the object prefix comes out yi- or bi-). This page is about how one Move (which landing site) changes one Agree outcome (which prefix). A vs A′: an A-position is one where arguments get case/agreement and can bind (subject and object positions). An A′-position is one used for questions, focus and topics. A′-movement typically "reconstructs": the moved phrase behaves for some purposes as if it were still in its old spot.

1. Question

T60c (from possessor-command §6): W&J (70b) 'all John'si children saw himi' (bi-) is bad, and that needs the null patient above and before the doer's phrase. Option (ii) puts it there by A-movement (deep bi-), and option (iii) puts it there as a null A′ topic. Which is it? T59e, behind it, asked for Saxon's work first-hand. D2 (deep vs shallow bi-) is the standing question.

2. Source

SOURCED Saxon, Leslie 2003. "On Two OSV Constructions in Navajo: Beyond Subject-Object Inversion." In L. Carmichael, C.-H. Huang & V. Samiian (eds.), Proceedings of the Thirtieth Western Conference on Linguistics (WECOL 2001), vol. 13, pp. 340–359. Fresno: Dept. of Linguistics, California State University, Fresno. Read in the public PDF of the whole volume on the CSU Fresno linguistics site (link). Printed p. 340 = PDF p. 347 (the volume's contents page gives 340, and the next paper starts on p. 360). The scan's text layer garbles Navajo ("yiyiiltsa", "biisxi"), so every Navajo form below was read off page images at 130, 300 and 500 dpi. Her footnote * says a version was to appear in the WSCLA 6 working papers. That version was not opened.

Saxon is an Athabaskanist (Dogrib/Tłı̨chǫ). Most of her Navajo data are cited from others: Hale 1973, Perkins 1978, Schauber 1979, Willie 1991, Willie & Jelinek 2000, and Young & Morgan 1987. None of those were re-opened here. The forms are quoted as Saxon prints them, with her source given. Willie 1991 has been used elsewhere on this site, but his p. 204, the source of Saxon's (7), was not re-checked this run.

3. The data SOURCED (as printed by Saxon)

The first gloss line is Saxon's, as printed. The second is our Leipzig-style line. It adds nothing her prose does not say: 3OBJ(yi)/3OBJ(bi) mark which object prefix appears. Her apostrophes and capitals are kept.

3a. The classic alternation and a question word under bi-

Saxon 2003:341, (2a, b), citing Perkins 1978:102; p. 343, (7), citing Willie 1991:204; p. 348, (16), citing Willie & Jelinek 2000:272 (page images checked)
(1a)ʼAtʼéédJáanyiztsʼǫs.
girlJohny-kissed
girlJohn3OBJ(yi).3SBJ.PFV.kiss
'The girl kissed John.' but not 'John kissed the girl.'
(1b)Jáanʼatʼéédbiztsʼǫs.
Johngirlb-kissed
Johngirl3OBJ(bi).3SBJ.PFV.kiss
'The girl kissed John.' but not 'John kissed the girl.'
(2)Háí-shJáanbiztsʼǫs?
who-QJohnb-kissed
who-QJohn3OBJ(bi).3SBJ.PFV.kiss
'Which one (of you) was it that John kissed?'
(3)Háí-shJáanyiztsʼǫs?
who-QJohny-kissed
who-QJohn3OBJ(yi).3SBJ.PFV.kiss
'Who kissed John?' or 'Who did John kiss?'

P. 343: (2) "is in conflict with the topic analysis of bi-, as question elements cannot be topics because of a conflict between semantic content and information structure." The question word under bi- gets a "discourse-linked" reading ('which one?'). P. 348: (3) "is particularly interesting because it is ambiguous … with neutral question interpretations". Under yi- a fronted question word can be the object.

3b. Focus Movement: object first, yi- kept

Saxon 2003:347, (14a, b), citing Schauber 1979:118, 117 (page images checked at 300 dpi)
(4a)Ashkiihaʼátʼííshyiyiiłtsą́?
boywhat-Qy-saw
boywhat-Q3OBJ(yi).3SBJ.PFV.see
'What did the boy see?'
(4b)Haʼátʼííshashkiiyiyiiłtsą́?
what-Qboyy-saw
what-Qboy3OBJ(yi).3SBJ.PFV.see
'What did the boy see?'

P. 347: "In both cases the object inflection is yi-." Pp. 346–347: Young & Morgan 1987 dictionary sentences (her (10)–(13), pp. 601, 640, 599, 219) also front an object while keeping yi-. Two of them have a fronted object marked with the indefinite léiʼ, which is impossible under bi- inversion (Willie 1991:77, her (5b), p. 342). "Three of the four examples … involve movement of an object which would be ineligible for SOI because of a conflict with the animacy hierarchy". The fronted object is "typically inanimate, preceding a human agentive subject which outranks it" (p. 347). Those four forms are not quoted here (only the 130 dpi images were read for them). Pp. 348–349: question words can also move out of an embedded clause ((17), (18), Schauber 1979:131, 181; not quoted).

3c. One bi- per clause when the doer is a person

Saxon 2003:343, (8a–c), citing Perkins 1978:116 (page image checked at 300 dpi; verb forms not quoted)

Schematic: [Ashkii 'boy'] [atʼééd 'girl'] [bike] [P 'with'] [V 'pull']. (8a) Both objects are under yi- (yił 'y-with', y-pull): 'The boy is pulling the bike with the girl [on it].' (8b) One is inverted (bił 'b-with', y-pull): 'The girl is pulling the bike with the boy [on it].' (8c) Both are inverted (bił 'b-with', b-pull): starred. P. 343: "it is only possible to invert one nominal; thus (8b) is possible but (8c), with two bi- forms, is strongly ruled out." The verb and the word 'bike' are left schematic because the accents on the verb could not be read with certainty (it is printed dziʼizi in (8a) and dziʼizí in (8b, c)).

3d. Thing-subjects: bi- in either order, and two bi-s in one clause

Saxon 2003:352, (36), citing Young & Morgan 1987:802; p. 353, (38), citing Y&M 1987:227; p. 354, (39), citing Y&M 1987:151 (page images checked at 500 dpi)
(5)... joołshéłtʼishńtʼééʼsitsilibéstał.
ball1s-hit fouland1s-y.brotherb-kicked
ball1SG.SBJ-hit.fouland1SG.POSS-younger.brother3OBJ(bi)-kicked
'I batted a foul ball and it hit [lit. kicked] my younger brother.'
(6)Biłbináʼeeł.
b-withb-float around
3(bi)-with3OBJ(bi)-float.around
'It (a boat) floats around it (e.g. an island) with him.'
(7)... shicheiiʼahbínídą́ą́ʼtsinaabąąsbi4 [sic]biníbą́ą́z lá.
1s-gr.fatherthis.morningwagonb-withb-catch up to   Emphatic
1SG.POSS-grandfatherthis.morningwagon3(bi)-with3OBJ(bi)-catch.up.to EMPH
'... this morning the wagon rolled along catching up to them with my grandfather [driving].'

P. 353 on (5): "the null subject of béstał 'hit' refers to the ball … Object inflection agreeing with sitsili 'my younger brother' is bi-." So bi- appears with the order subject before object. Pp. 351–352: Willie 1991:90's 'the lightning killed the horse' is bi- in both orders, and yi- is "objectionable". Pp. 353–354: "When we examine clauses where the subject is inanimate, we find examples in which one clause contains two instances of bi-, contrary to these predictions." In (6) and (7) the subject is a conveyance (boat, wagon), the rider is the object of 'with', and another object is also bi-. In (7) "bi4" is a misprint in the volume; her gloss is 'b-with', and the text calls it the postposition -ił 'with'. The spacing "bin íbą́ą́z" in the scan is read as one word. (40), with 'airplane', has the same pattern (Y&M 1987:475; not quoted).

Saxon 2003:353, (37a), citing Young & Morgan 1987:grammar 7–8 (page image checked)

ʼashkii 'boy', ʼashiiké 'boys'. P. 353: "Only animate (actually, human) nouns in Navajo can be specified for singular as opposed to plural number." This is the basis of her account of §3d.

4. Her analysis SOURCED (hers)

Her clause tree (9), p. 345, written out (heads on the right, as she draws them):

[DiscP  DP(1/2 subject; D-linked 3 object)
        [AspP  [NumP  DP(3 subject)
                      [AgrOP  DP(object)
                              [VP  DP(internal subject)  [V'  DP(internal object)  V ]]  AgrO° ]  Num° ]  Asp° ]  Disc° ]
(Focus Movement lands higher still, in Spec,CP.)
  • yi- needs a c-commanding 3rd-person subject (p. 345): "yi- is used when the object falls within the c-command domain of a subject in NumP. If the object is in [Spec, DiscP], it is not in the c-command domain of the third person subject. Therefore yi- is not licensed and the default third person object marker bi- must be used instead." Fn. 6 (p. 356): Hale 2000:70 describes inversion as "removing it from the domain in which obviation applies"; "Clearly my concept is identical."
  • DiscP (p. 344): named after Déchaine 2000. "discourse-linked DPs in Navajo raise to this [Spec, DiscP] in order to be licensed in the discourse". The position "is identical to Hale's conception of the inverted object as 'the DP appearing in the highest Spec or adjunct position in the inflectional portion of the extended projection' (Hale 2000:71, fn12)". It is the one slot for 1st/2nd-person subjects and D-linked 3rd-person objects. So with a 3rd-person subject there is at most one bi- (pp. 345–346). P. 345: Rice & Saxon 1994 proposed "a structure just like (9)" for northern Athabaskan, but "we called the highest functional category not DiscPhrase but AgrSP".
  • The definite restriction follows because léiʼ "necessarily introduces a new discourse referent" (p. 344, after Fernald et al. 2000). D-linked 'which one?' in (2) fits, and so does Willie's ban on generics (p. 344).
  • Focus Movement (p. 348): "object movement, as in (10), and question movement, as in (16), involve movement of a DP to [Spec, CP]. The use of yi- as object inflection represents a reconstruction effect, expected in A-bar movement." She leaves open whether non-question Focus Movement is iterable and unbounded (pp. 348–349). P. 349: yi- in Focus Movement "represents agreement with an empty object reflecting a moved element", which links it to null objects under yi- ((19)–(27)).
  • Thing-subjects stay in VP (p. 353): "in these examples the subject nominal remains inside the VP". "Assuming that features of Num° attract only a subset of nouns including those inflectable for number, we can derive the restriction against inanimate subjects in [Spec, NumP]." With no subject in Spec,NumP, every 3rd-person object is bi-, so two bi-s can co-occur (p. 354).
  • Against topic and one-proximate accounts: Clitic Left Dislocation "is not restricted to a single occurrence per clause" (p. 344, citing Cinque 1990:58), and question words can't be topics ((2)). She rejects Aissen 2000's proximate bi- (pp. 344, 353), but keeps yi- as an obviative: "it represents a second third person in a given domain" (p. 355). bi- is "the default form" (p. 355).
  • Framework label. Tree (9) has VP-internal subject and object positions, so it is already VPISH-compatible. It is not a flat-S tree and needs no pre-VPISH restatement. Its functional heads (AgrOP, NumP) are 1990s labels. Below, AgrO° is read as the head that probes the object for Agree, and Spec,NumP as the derived position of a 3rd-person subject.

5. What this does to T60c and D2

SOURCED Saxon separates two object-first constructions and treats only one as A′-movement. Focus Movement goes to Spec,CP and is A-bar, with yi- as its reconstruction effect. bi- inversion goes to Spec,DiscP, inside the inflectional layer, and there the moved object's own position decides the prefix.

HYPOTHESIS By her own reasoning, the bi- landing site behaves like an A-position. She never uses the words "A-position" for DiscP, so this is our reading, with three supports from her text:

  1. No reconstruction. Her test for A-bar movement is that it reconstructs, which is why Focus Movement keeps yi-. bi- inversion is the case where the prefix is computed from the landing site. That is the A-movement pattern.
  2. Same slot as subjects. Spec,DiscP is also where 1st/2nd-person subjects go in (9).
  3. Same slot as AgrSP. She equates it with what Rice & Saxon 1994 called AgrSP, a subject-agreement projection. Its specifier is standardly an A-position (standard terminology, not a claim of hers).

HYPOTHESIS Consequence for (70b) (possessor-command options (ii) vs (iii)). On Saxon's tree, the null bi- patient is in Spec,DiscP. That is above Spec,NumP, where the plural doer phrase [tʼáá ʼałtso Jáan ba-ʼáłchíní] sits, so the patient c-commands Jáan. If Spec,DiscP is an A-position, standard Condition C ("a name must not be bound from an A-position") gives the star directly. That is option (ii). Under option (iii), a null A′ topic, standard Condition C would not apply, since A′-binders don't count for it. The star would then need either reconstruction or Platero's string-based precede-and-command rule (PC-1), which does not distinguish A from A′. Saxon's arguments against the topic analysis are (2) and the two-bi- clauses. They weaken (iii) as an account of bi- itself. They do not rule out a separate null topic.

HYPOTHESIS Saxon is a sourced "deep" analysis, in Oxford's sense. The patient Moves above the agent's derived position, and the prefix follows from that. She differs from Oxford 2023's outer-Spec,v version in height: she lands the object above Spec,NumP, not just above Spec,v. She differs from Aissen in what the prefix means: bi- is the default, and only yi- is obviative.

OPEN T60c is partly answered. There is now a sourced analysis with a principled A-vs-A′ split, but still no datum: Saxon gives no reciprocal, weak-crossover or quantifier-binding sentence under bi-. D2 stays open. The deep side has gained a source and a structure, not a test. T59e is partly answered. Saxon's later Navajo account (yi- needs a c-commanding 3rd-person subject in Spec,NumP) fits Horseherder's report of Saxon's "disjoint anaphor" condition (possessor-yi). But fn. 1 sets possessive yi-/bi- aside, and Saxon 1984a/1986/1995 remain unopened.

6. Two bi-s: contradicting data for "one proximate per clause"

SOURCED (6) and (7) (Y&M 1987 via Saxon) have two bi- objects in one clause with a thing as subject. (8c) (Perkins via Saxon) bars two bi-s with a human subject.

HYPOTHESIS These are contradicting data for any account where bi- marks the (single) proximate (Aissen 2000) or the (single) topic (Speas & Yazzie 1996, as Saxon reports them; W&J 2000). They are data, not silence, so they count. What survives:

  • Aissen's ranking may still be right for clauses with an animate 3rd-person subject, which is all the data on possessor-binding, possessor-gap and possessor-yi. Those pages' results are not overturned. But "bi- = proximate" cannot be the whole story, and the possessor-gap conclusion "ranking alone explains it" now carries the caveat "for animate-subject clauses".
  • Saxon's own fix uses structure, not ranking: no subject in Spec,NumP → no yi- licensor → default bi- everywhere. Her rainbow pair (fn. 9, p. 356, Y&M 1987:225: yi- and bi- both listed with 'rainbow' as subject) is an exception. She writes "Neither Perkins 1978 nor I predict the grammaticality of (i)", and notes that Y&M list the yi- form as "a secondary option" (T61c).

HYPOTHESIS Proto-roles. The subjects that stay in VP (lightning, a ball, a boat, a wagon, a plane) are weak Proto-Agents in Dowty's sense. Saxon's own criterion is morphological: the noun can't inflect for number. The two criteria overlap on her data, and nothing here separates them (T61d). Note that in (6) and (7) the vehicle is the subject of a motion verb and the rider is an oblique 'with him'. This is the same -ił travel pattern documented on the -ił travel pattern page. Saxon adds a structural reason why the vehicle-subject leaves both objects as bi- (SX-4, HYPOTHESIS as applied there). None of these sentences has a distance phrase, so they don't bear on Grant's distance-as-incremental-theme question.

7. Rulesets A and B HYPOTHESIS

Saxon's yi-/bi- choice is a word-level spell-out (ruleset A: a prefix inside the verb word) that is decided by a clause-level configuration (ruleset B: whether a 3rd-person subject in Spec,NumP c-commands the object at the end of the derivation). In this site's terms, the changeover point carries information one way: B-level c-command is read by an A-level Agree. Focus Movement (Spec,CP) is the one B-level Move that the prefix does not see, because it reconstructs. So the prefix is a probe into the inflectional layer (DiscP and below) and is blind to CP. Saxon draws her heads on the right in (9) and does not derive the left-headed prefix order (she cites Hale 2001 on the "bipartite" verb for that). The A-side ordering is not tested here.

8. Rules

SX-1  yi- (3rd object) is licensed iff the object is c-commanded by a 3rd-person subject     SOURCED (Saxon p. 345, 353, 355)
      in Spec,NumP; otherwise default bi-.
SX-2  bi- inversion: a D-linked 3rd object Moves to Spec,DiscP (above NumP); the one slot     SOURCED (p. 344-346, tree (9))
      for 1/2 subjects and D-linked 3 objects → at most one bi- with a 3rd-person subject.
SX-3  Focus Movement: object or wh-phrase Moves to Spec,CP; A-bar; yi- kept as a              SOURCED (p. 348); iterability and
      reconstruction effect; indefinites and animacy-violating orders allowed.                   unboundedness OPEN (her words)
SX-4  Thing-subjects (no number feature) stay in VP; Num° attracts only number-inflectable   SOURCED (p. 353-354)
      nouns → no yi- licensor → bi- in either order, and two bi-s per clause possible.
SX-5  bi- under a wh-word is D-linked ('which one'); wh can't be a topic → against the        SOURCED (p. 343)
      topic analysis of bi-.
SX-6  Spec,DiscP behaves as an A-position: no reconstruction for the prefix, hosts            HYPOTHESIS (our reading; Saxon
      1/2 subjects, = Rice & Saxon's AgrSP.                                                  does not say "A-position")
SX-7  With SX-6, W&J (70b)'s star follows from standard Condition C (option (ii));          HYPOTHESIS; D2 OPEN
      an A′ null topic (iii) would need reconstruction or Platero's PC-1.
SX-8  (6), (7) two bi-s: contradicting data for one-proximate / one-topic per clause; ranking  HYPOTHESIS (on SOURCED data)
      results on animate-subject clauses unaffected.
      [Update 2026-10-26: checked against Aissen 2000 fn. 5, 15 → confirmed; see /one-proximate/ OP-5.]

9. Derivations, step by step

Labels: Σ role schema, A word, B clause. VPISH throughout: ρA first-Merged in Spec,V/v (Saxon's "internal subject"), ρP as complement. Saxon's DPs are written as slot bundles, and overt nouns are linked at B (nominal adjunction). Where her account needs a noun in a position, that is noted.

X-1. (1a) ʼAtʼééd Jáan yiztsʼǫs 'The girl kissed John' (direct)

  1. Σ Σ = {ρA: 3sg human (girl); ρP: 3sg human (John)}.
  2. B Merge V 'kiss' + ρP; Merge ρA in Spec,VP (VPISH). Merge AgrO°; Move ρP → Spec,AgrOP. Merge Num°; ρA has a number feature, so Move ρA → Spec,NumP.
  3. A Agree: AgrO° probes ρP (3sg). The licensing check (SX-1) asks whether a 3rd-person bundle in Spec,NumP c-commands ρP. It does, so the object prefix is spelled yi- → yiztsʼǫs.
  4. B Linear order, right-headed: Spec,NumP before Spec,AgrOP before the verb word → girl, John, verb ✓.

X-2. (1b) Jáan ʼatʼééd biztsʼǫs (bi- inversion)

  1. Σ as X-1.
  2. B As X-1 through Spec,NumP. Then Merge Asp°, Disc°. ρP (definite, D-linked) Moves → Spec,DiscP. Under SX-6 this is A-movement (HYPOTHESIS).
  3. A Agree: AgrO° probes ρP. The check is made at ρP's landing site (no reconstruction): ρP is above Spec,NumP, so it is not c-commanded by the subject, and yi- is not licensed. Default bi- → biztsʼǫs.
  4. B Order: Spec,DiscP before Spec,NumP before V → John, girl, verb ✓. The roles never change: the girl is still ρA.

X-3. (4b) Haʼátʼíísh ashkii yiyiiłtsą́? (Focus Movement, yi- kept)

  1. Σ Σ = {ρA: 3sg human (boy); ρP: 3, wh}.
  2. B As X-1: ρP → Spec,AgrOP; ρA → Spec,NumP. Merge C; Move the wh bundle → Spec,CP (A-bar).
  3. A Agree: the licensing check is made at the copy in Spec,AgrOP (reconstruction, SX-3). That copy is c-commanded by Spec,NumP, so yi- → yiyiiłtsą́ ✓. Contrast X-2, where the check is made at the moved position.
  4. B Order: Spec,CP first → what, boy, verb ✓. (4a) is the same without the Move to CP.

X-4. (6) Bił bináʼeeł 'It (a boat) floats around it with him' (two bi-s)

  1. Σ Σ = {ρA/S: 3, boat (no number feature); ρP: 3 (island); ρwith: 3 (rider)}. Saxon gives no theta labels. 'Around' and 'with' are her glosses.
  2. B Merge the VP with the boat bundle in Spec,VP. Num° finds no bundle with a number feature to attract, so the subject stays in VP (SX-4).
  3. A Agree on each 3rd-person object (the verb's, and the postposition's): no c-commanding subject in Spec,NumP → yi- not licensed → default bi- twice → bił, bináʼeeł ✓. Nothing limits bi- to one, because neither object had to Move to the single Spec,DiscP slot.

X-5. W&J (70b) on Saxon's tree HYPOTHESIS

  1. Σ Σ = {ρA: 3pl human, index c; ρP: 3sg, index j}; π inside the doer's phrase: 3sg, j. (W&J's string is carried over from possessor-gap §5, not re-opened.)
  2. B ρA → Spec,NumP. To give bi- with a human subject, ρP must Move → Spec,DiscP (SX-1, SX-2).
  3. A Agree: bi-, and da- counts the plural agent → dabiiłtsą́.
  4. B The doer's phrase [tʼáá ʼałtso Jáan ba-ʼáłchíní] is linked to ρA in Spec,NumP. ρP (null, index j) in Spec,DiscP c-commands it. If the doer's phrase counts at Spec,NumP (as assumption A3 on the Hale derivation page already requires), then ρP c-commands Jáan.
  5. Condition C, under SX-6 (A-position): j = John is out → only 'saw himk' ✓. That matches W&J. Under an A′ reading of the same slot, standard Condition C would not block it, and the star would need PC-1 or reconstruction.

10. Grant's ideas

GRANT'S CONJECTURE Feature bundles as arguments. Not contradicted. Saxon says outright that her assumption that yi-/bi- "are affixes marking agreement with DPs in argument positions" is "controversial in studies of Navajo (cf. Willie and Jelinek 2000, Hale 2001…)" (p. 341). It is an analytic choice, not contradicting data. One part of her account is bundle-friendly: what Num° attracts is chosen by a feature (number), not by a noun as such. Thing-nouns stay down because they lack the feature, which is exactly what a bundle-driven grammar would say. X-1 to X-5 run with bundles throughout.

GRANT'S CONJECTURE Strata (theta roles first, event structure next, lexical items later). OPEN, consistent. Saxon's spine reads as layers stacked on a fixed theta core: VP (roles) → AgrOP/NumP (person, number) → DiscP (discourse linking) → CP (focus). The roles never change in (1)–(7), only the layer the object reaches. Whether a noun is definite (léiʼ) matters at DiscP, and it is a lexical property that arrives late, as the strata idea expects.

GRANT'S CONJECTURE Merge/Agree sites inside the word. Not addressed by Saxon. §7's reading (an A-level Agree that reads B-level c-command) is a HYPOTHESIS of this site.

Chemistry: SPECULATIVE, flag only. "A default marker everywhere except where one specific licensor c-commands" is an elsewhere-condition. No chemical correlate is identified or claimed.

11. Bottom line

  • SOURCED Saxon 2003 (first-hand, page images checked): Navajo has two object-first constructions. bi- inversion (definite/D-linked, one per clause with a human subject) goes to Spec,DiscP. Focus Movement (yi- kept, indefinites and question words fine) goes to Spec,CP as A-bar movement, and yi- is its reconstruction effect. Thing-subjects stay in VP, giving bi- in either order and two bi-s in one clause ((6), (7)).
  • HYPOTHESIS By her own criteria the bi- landing site is A-like. That supports option (ii) (deep bi- + Condition C) for W&J (70b) over a null A′ topic. T60c is partly answered: analysis yes, binding/crossover datum no. D2 OPEN.
  • HYPOTHESIS The two-bi- clauses are contradicting data for "bi- = the one proximate/topic". The ranking results on this site's animate-subject pages stand, but only for animate subjects.

12. Fidelity notes

  • The text layer is badly garbled for Navajo. All quoted forms were read off page images: ʼAtʼééd, Jáan, yiztsʼǫs, biztsʼǫs, Háí-sh (130 dpi); Ashkii, haʼátʼíísh, yiyiiłtsą́ (300 dpi); jooł, shéłtʼish, ńtʼééʼ, sitsili (printed without accents), béstał, Bił, bináʼeeł, shicheii, ʼahbínídą́ą́ʼ, tsinaabąąs, biníbą́ą́z, lá (500 dpi).
  • Kept as printed: the "bi4" misprint in (7) (marked [sic]); no initial glottal stop on Ashkii in (4) (Schauber's spelling via Saxon); ʼashkii with one in (37a).
  • (8a–c) and (10)–(13), (17), (18), (40) are described, not quoted, because their verbs or diacritics could not be confirmed.
  • All of Saxon's data are second-hand from the cited works. None of those (Hale 1973, Perkins 1978, Schauber 1979, Young & Morgan 1987, W&J 2000) were re-opened this run.
  • "A-position" for Spec,DiscP is our reading (SX-6). Saxon labels only Focus Movement (A-bar).
  • The PDF is a public download of the whole proceedings volume from CSU Fresno. No restriction was bypassed.

13. Open questions

IDQuestionWhat would settle it
T61aThe datum T60c still lacks: does a Spec,DiscP (bi-) patient bind a reciprocal or a bound pronoun in the doer's phrase, or show weak crossover? Does a Focus-Moved (yi-) object fail to?A sourced pair: bi- vs Focus Movement with 'each other' or a quantified object. OPEN
T61bDoes Focus Movement reconstruct for binding as well as for the prefix? Is non-wh Focus Movement iterable and unbounded (Saxon leaves both open)?Sourced Focus-Moved objects containing an owner prefix; long-distance non-wh fronting. OPEN
T61cSaxon's fn. 9 rainbow pair (Y&M 1987:225): yi- with a thing-subject. Is "agency" (Willie 2000:365) the licensor instead of number?Y&M 1987:225, 237 first-hand; Willie 2000. OPEN
T61dNumber feature (Saxon) vs Proto-Agent strength (Dowty): which decides whether a subject leaves VP? Animals and collectives would separate them.Sourced thing/animal subjects with number-inflecting vs non-inflecting nouns. OPEN
T61eDoes Aissen 2000 say anything about inanimate subjects and two bi-s? Re-read her (on the box) against SX-8.Aissen 2000 first-hand ([local working copy, not in the repository]). [Update 2026-10-26 ~4:46am PT: answered on one-proximate. She is silent on such clauses, but her explicit "exactly one proximate" (fn. 15) and "nonsubject pronouns" (fn. 5) predict one yi- in (6)/(7); SX-8 confirmed. T62a–e.] SOURCED / HYPOTHESIS
T61fRice & Saxon 1994 (AgrSP), Hale 2000:70–71 (inversion as leaving the obviation domain), WSCLA 6 version, Saxon 1984a/1986/1995 first-hand.Those texts. OPEN

Sources: Saxon 2003 (WECOL 2001 vol. 13, pp. 340–359; page images pp. 341, 343, 345–349, 351–355). Carried over: W&J 2000 (70b), Platero 1978 PC-1, Horseherder 1998, Oxford 2023. See also possessor-command, possessor-yi, possessor-gap, deep or shallow inverse, inverse, wh-questions, open questions, sources.

One "main" third person per clause? Aissen 2000 re-read against the two-bi- sentences (T61e)

Working note. Kept as written (URL, data and status labels unchanged). Filed under Obviation: how many third persons; the consolidated argument is the essay obviation. Rule IDs: rules index.

Verbs on this page → dictionary: dabiiłtsą́ ‘see’ · yoolóós ‘lead (one animate)’ · nímááz ‘rolled; with yaa/baa ‘to’: rolled to’

Restated on the unified structure (2026-11-15). The inverse is restated as one extra Move on the clause spine: the patient moves (via an outer Spec,vP) above the agent, which has already raised to Spec,TP by the working default (H3, H6′, H21b). The event LF is the same as the direct; the trigger and landing site are OPEN. 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. Status on the spine: see unified structure, §4, inverse row. 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.

The last page (two ways to put the object first) found Young & Morgan sentences in which a thing is the subject and two other third persons are both marked bi-. T61e asked what Judith Aissen's obviation account says about such sentences. This run re-read her paper in full, footnotes included. In plain words: Aissen says outright that a clause has exactly one "main" (proximate) third person and that bi- marks it. She never discusses a clause with a thing as subject and two people or things after it. Her rules therefore predict that one of the two would have to be yi-, and the Young & Morgan sentences have bi- on both, so they contradict her as she states it. A smaller change fits all the data checked here. In it, yi- means "outranked by the subject", not "not the main one", and bi- is the default elsewhere. That is close to how Hale (2000, as Saxon quotes him) and Saxon put it. The sources also disagree with each other: when a ball rolls over a log, Perkins (via Saxon) has bi-, but when a rock rolls onto a tree, Creamer has yi-.

On this page: 1 Question · 2 Sources · 3 What Aissen says · 4 The data she doesn't cover · 5 Where her rules and the data part · 6 Four repairs tested · 7 A conflict between sources · 8 Rulesets A and B · 9 Rules · 10 Derivations · 11 Grant's ideas · 12 Bottom line · 13 Fidelity notes · 14 Open questions

Notation as on two-osv. Σ = the role schema. A = the word level (prefixal, left-headed; W = the verb or postposition word). B = the clause level (right-headed). ρA/ρP = agent and patient slots, ρwith = the object of 'with', ρG = another oblique object. [prx]/[obv1]/[obv2] = places on Aissen's obviation tier. Merge vs Agree: Merge builds structure: it puts two pieces together (Move is Merge again of something already built). Agree values features: it lets a head read person, number or rank off a bundle. This page is mostly about Agree. The question is which bundle the prefix on a word compares itself with to come out yi- or bi-.

1. Question

T61e (from two-osv §13): "Does Aissen 2000 say anything about inanimate subjects and two bi-s? Re-read her (on the box) against SX-8." SX-8 said that Saxon's two-bi- clauses are contradicting data for "one proximate per clause". That was stated against a summary of Aissen. This page checks it against her text.

2. Sources

SOURCED Aissen, Judith 2000. "YI and BI: Proximate and Obviative in Navajo." In A. Carnie, E. Jelinek & M. Willie (eds.), Papers in Honor of Ken Hale (MITWPL Endangered and Less Familiar Languages 1), pp. 129–150. Read in full in the author's RTF (people.ucsc.edu/~aissen/navajo.rtf, on the box since 2026-10-24). This time all 15 footnotes were read as well, pulled straight from the RTF, because the plain-text copy used on possessor-binding and possessor-gap drops them. The RTF has no page numbers, so examples, sections and footnotes are cited instead. The Navajo is in the SIL IPA93 font, decoded with the key used before (¬ = ł, 7 = nasal hook, @ = high tone), and the RTF's \lquote is printed as ’.

SOURCED Saxon 2003 (WECOL 2001 vol. 13, pp. 340–359; full reference on two-osv). Re-opened. Page images were checked again at 400 dpi for p. 352 (32) and at about 130 dpi for p. 354 ((39), (40) and the paragraph after them). Fn. 7 and fn. 9 (p. 356) were re-read in the text layer.

SOURCED Creamer 1974 ("Ranking in Navajo Nouns", Navajo Language Review 1(1), ERIC ED 101 549; full reference on eight ranks). Re-opened: printed pp. 36–37 page images were re-read for (c21)–(c26) and the group-7 rule.

Carried over, not re-opened: Willie 1991 (73)–(80), (85)/(86) (inverse §2.4–2.5); Perkins 1978 and Young & Morgan 1987 only as Saxon prints them.

3. What Aissen says SOURCED

3a. The tier, and only one proximate

  • The tier (§2.2). "Each obviation span consists of a set of ranked relations Proximate > Obviative 1 > Obviative 2 > …". "each 3rd person nominal associates to a relation of obviation in an obviation span", and "two 3rd person nominals associate to the same relation if and only if they are presupposed coreferential."
  • Exactly one proximate. In the main text (§4.1, on (13b)): "Since there can be only one proximate, the subject, 'man', would by force be obviative." Fn. 15: "within an obviation span, there is exactly one proximate. But obviation spans are possible with multiple 4th persons." Conclusion (§7): obviation selects "a single most prominent 3rd person in the span".
  • What the prefixes mean (§3.1, (8)): "a. BI is the object proximate pronoun; b. YI is the object obviative pronoun." Fn. 5: "(8) is misleading … YI and BI contrast in other functions beside direct object. It would be more appropriate to term them nonsubject pronouns. The analysis proposed here extends straightforwardly to many of these cases." So postposition objects such as yił/bił are covered.
  • Not 3rd-person subjects (fn. 7): "The 3rd person object marker with 1st, 2nd, and 4th person subject is not always BI. In fact, it is usually ∅. But in those contexts where an overt marker appears, it is always BI." 4th person (§6): 4th persons lie "outside the YI/BI alternation", and using them allows reference "outside the obviation system".
  • Constraints: (11) Proximate precedes Obviative; (12) Avoid a non-topic proximate; (14) Avoid an inanimate proximate; (20) Avoid a lexical proximate; (26)/(33) Avoid a possessum proximate; (36) Avoid a proximate object. Ranking (32): (14) » (20) » (12); (33) unranked against them; (36) "ranks below all the others discussed".
  • Stated for two. §5 sums up the account as: "given two 3rd person arguments, associate that argument which is most prominent with the more prominent relation of obviation, namely proximate." §2.1 likewise frames the animacy preference "in a context with only two 3rd persons".
  • Framework (§3.1): she says her assumption that the prefixes are incorporated pronouns follows Jelinek 1984 (flat S, pre-VPISH) and that it does "not play a crucial role in my analysis". The tier is "a structural dimension distinct from constituent structure or grammatical function" (§2.2). Restated below with vP/TP slots; nothing in her argument depends on the flat S.

3b. Her one thing-subject example

Aissen 2000, (13a, b) and (16a, b), both citing Hale 1973 (RTF, SIL IPA93 decoded)
(1)dinédziłyooį́.
manmountainYI-sees
manmountain3OBJ(yi)-see
'The man sees the mountain.' = her (13a); (13b), with the mountain first and BI, is starred.
(2)dibétóʼabííł ʼéél.
sheepwaterBI-swept off
sheepwater3OBJ(bi)-swept.off
'The water swept the sheep off.' = her (16a)
*(3)tódibéʼayííł ʼéél.
watersheepYI-swept off
watersheep3OBJ(yi)-swept.off
'(The water swept the sheep off.)' = her (16b)

On (16): "when the subject is inanimate and the object animate, only the BI form should be possible … In this case, (14) forces proximate status on the object because it outranks the subject in animacy. The appropriate object pronoun then is BI." This is the only clause in the paper with an inanimate subject; it also opens the paper as (2), cited to Hale 1973:302. (Her (22)–(23) have an inanimate, but as the unspoken object.) Every example has at most two 3rd persons, and none has a 3rd-person postposition object beside a 3rd-person direct object. §7: "The present paper scratches the surface, both with respect to the data it does discuss, and certainly with respect to data not discussed."

Answer to T61e, part 1 SOURCED: No. Aissen says nothing about clauses with a thing-subject and two other 3rd persons, or about two bi-s in one clause. What she does say, explicitly (fn. 15, §4.1, §7) and not as a reconstruction of ours, is that a span has exactly one proximate. Fn. 5 says BI/YI are nonsubject pronouns in general, so the prediction covers postposition objects.

4. The data she doesn't cover SOURCED (as printed by Saxon and Creamer)

Saxon's (38), (39) are quoted with full glosses on two-osv §3d as (6), (7). Here they are in short form, with the paragraph that follows them on p. 354 (page image re-read this run):

  • (4) = Saxon (38), Y&M 1987:227: Bił bináʼeeł. 'It (a boat) floats around it (e.g. an island) with him.'
  • (5) = Saxon (39), Y&M 1987:151: ... shicheii ʼahbínídą́ą́ʼ tsinaabąąs bi4 [sic, 'b-with'] biníbą́ą́z lá. '... this morning the wagon rolled along catching up to them with my grandfather [driving].'
  • (6) = Saxon (40), Y&M 1987:475 ('airplane'). Same pattern: 'with' + rider and 'down to it', both b-. Not quoted, because the 'with' word is faint in the scan.
  • Saxon p. 354: "In each of these examples the main predicate is a verb of motion whose inanimate subject is a conveyance (boat, wagon, plane) for the rider who is represented in Navajo grammar as the object of the postposition -ił 'with'. The clause in each case also includes another oblique object, which like the object of -ił is marked by bi-."

The comparison set (all SOURCED, carried over or re-read):

  • Perkins 1978:116 via Saxon (8a–c) (two-osv §3c): human subject 'girl'/'boy', bike, 'with'. (8a) yił + y-verb ✓, (8b) bił + y-verb ✓, (8c) bił + b-verb *.
  • Creamer 1974:36–37 (page images re-read): thing-subject with yi- on a postposition object: (c21) Tó dáʼdeestlʼin yinahjįʼ ninígoʼ 'The flood came to rest against the dam'; (c23) Tsé tʼiis yikʼi chʼínímááz 'The rock rolled upon the tree'; (c25) Tʼiis tsé yikʼiikę́ę́z 'The tree fell upon the rock'. The bi- counterparts (c22), (c24), (c26) are starred, with Creamer's comment that "one structure is not preferred". Group-7 rule (p. 36): the noun that the verb implies moves "takes higher status than the more stationary object". Glosses on eight ranks §3.3.
  • Saxon (32), Perkins 1978:55 (page image re-read at 400 dpi): Jooł tsin báhátis yiltsʼid ... 'ball log b-over 3s-rolled', 'The ball rolled over the log.' A thing-subject with bi- on a postposition object. Saxon cites Perkins 1978:136: "the /yi/ form is used with animate subjects and the /bi/ form is used with inanimate subjects."
  • Willie 1991 (79)/(80) yichʼįʼ/bichʼįʼ and (85)/(86) bá + yoolóós (inverse §2.5): postposition objects alternate in the same way, so they are on the subject's tier.

5. Where her rules and the data part HYPOTHESIS (on SOURCED data)

Apply Aissen's rules to (5), as she states them, with no added assumptions:

  1. Three 3rd persons: the wagon (subject), 'my grandfather' (object of 'with') and 'them' (object of 'catch up to'). They are three distinct referents, so by §2.2 they take three distinct relations: Proximate, Obviative1, Obviative2.
  2. All three are ordinary 3rd persons, not 4th: the prefixes print as b-, while 4th-person forms have their own prefixes (§6). So the fn. 15 escape ("multiple 4th persons") does not apply.
  3. (14) Avoid an inanimate proximate: the wagon is not proximate. One of the two humans is.
  4. Exactly one proximate (fn. 15), so the other human is obviative.
  5. (8) with fn. 5: the proximate nonsubject is BI, and the obviative nonsubject is YI.
  6. Printed: bił and biníbą́ą́z, i.e. BI, BI. ✗

(4) works the same way with boat, island and rider. Even if the island counts as inanimate, (14) still leaves one proximate, which must be the rider; the island is then obviative and should be YI. Printed: BI. ✗ So SX-8 stands against the text itself: these are contradicting data, not silence. Her account is right on (1), (2)/(3), Perkins (8a)/(8b), Willie (79)/(80), (86) and Creamer (c21)–(c28) (two inanimates with no topic: (36) makes the subject proximate, so the object is YI).

6. Four repairs tested HYPOTHESIS

RepairIdeaResult
R1A postposition phrase is its own obviation span, so bił's rider doesn't compete.✗ Perkins (8a) yił under a human subject shows that the 'with' object is obviated by the subject, and Willie (79)/(80) shows the same for 'to'.
R2One of the two humans is 4th person (off the tier).✗ The printed prefixes are 3rd-person b- (§5, step 2).
R3Inanimate subjects are off the tier (like 4th persons).✗ Two humans are still left on the tier, one of them obviative, so YI is still predicted on one.
R4 (update 2026-10-26 ~12:46pm PT: restated as Aissen 1997's direct/inverse comparison on ranked-obviatives §5)Subject-relative yi-: a 3rd-person nonsubject is YI iff the clause's 3rd-person subject is on the tier and outranks it. Every other 3rd-person nonsubject is BI (the elsewhere form). Ranking among the nonsubjects themselves has no effect on the prefix.✓ (4), (5): the thing-subject outranks neither human (it is below both by (14)), so both are BI. (2)/(3): sheep > water, so BI. (1): man > mountain, so YI. Perkins (8a): boy is subject and proximate, girl and bike are lower, so yił + y-verb. (8b): boy is proximate (topic), girl (subject) is Obviative1 and bike is Obviative2. The boy is above the subject, so bił; the bike is below the subject, so y-verb ✓. (8c): a b-verb would need the bike above the girl, against (14) ✗ *. Creamer (c21), (c23), (c25): the mover outranks the stationary thing (group-7 rule), so YI ✓; (c22), (c24), (c26) *. Willie (86): the patient is not above the subject, so yoolóós ✓.

What R4 changes. Aissen's tier and constraints stay as they are. Only the spell-out of (8) changes. The old version was BI = proximate, YI = obviative. The new version is YI = "below the subject on the tier", BI = elsewhere. For every clause with exactly two 3rd persons, the two say the same thing: if the subject is proximate the object is obviative (YI), and if the object is proximate it is above the subject (BI). They differ only when there are three or more 3rd persons, which is exactly the data Aissen doesn't discuss. [Update 2026-10-26 ~8:46pm PT: correction. Hale's text, read first-hand on hale-general-b §4, says that topicalizing the object removes the object from the obviation domain. He credits this to Speas and Jelinek, and it is one of two options he gives. The Saxon footnote is fn. 6, not 7. Hale's own rule (y- only under c-command by a clausemate 3rd-person subject) is subject-relative like R4 (SOURCED), with c-command in place of rank (HYPOTHESIS).] R4 matches Hale's phrase as Saxon quotes it (fn. 7, Hale 2000:70): inverting the subject removes it "from the domain in which obviation applies, forcing use of the general third person object prefix b-". It also matches Saxon's own structural rule SX-1 (yi- needs a c-commanding 3rd-person subject) and Aissen's fn. 7 (BI is the overt form whenever the subject isn't an ordinary 3rd person). In all three, bi- is the general form and yi- is the marked one that needs a subject.

A side effect. Under R4, Perkins (8c) is ruled out by animacy (the bike can't outrank the girl), not by "one bi- per clause". So (8c) no longer shows that a human-subject clause allows only one bi-. That limit is not independently shown by any datum on this site (T62a).

What R4 does not fix. In (4) the island, and in (6) the 'down to it' object, are things. Under Creamer's group-7 rule a moving boat or plane should outrank a stationary thing, so R4 predicts yi- on that object. Both are printed bi-. R4 and the mover rule together are contradicted there, unless the rider changes the count (all three Y&M clauses have a human rider and none of Creamer's three yi- clauses does) or the speakers differ (§7). OPEN (T62b).

7. A conflict between sources SOURCED data, HYPOTHESIS reading

Same kind of event, opposite prefix:

  • Creamer (c23) Tsé tʼiis yikʼi chʼínímááz 'The rock rolled upon the tree' (bi- version *, "not preferred").
  • Perkins via Saxon (32) Jooł tsin báhátis yiltsʼid ... 'The ball rolled over the log' (the yi- version is not given; Saxon: "typically only the bi- construction is permitted").

Each has a rolling thing as subject and a stationary piece of wood as the postposition object. Creamer's rule (p. 36) predicts yi- for both, and Perkins's (p. 136) predicts bi- for both. These sentences don't decide between them, because each source gives only its own form; neither has the other's form starred in the same sentence. Possible readings: speaker or dialect variation, a difference between 'over' and 'upon', or the trailing "..." in (32) (part of a longer sentence). OPEN (T62c). Until it is settled, "thing-subjects never give yi-" (Perkins; Saxon's SX-4) and "the moving thing outranks" (Creamer) should both be read as one source's pattern, not as the language's rule. Saxon's own fn. 9 exceptions (rainbow yi-, Y&M 1987:225; wind and water yi-, Y&M 1987:237) point the same way.

8. Rulesets A and B HYPOTHESIS

Under R4 the prefix choice needs only two things: the rank of the nonsubject's bundle and the rank of the subject's bundle. Both are present in Σ before any noun is chosen. In (4) Bił bináʼeeł, every participant is unspoken (Saxon p. 354: "all of the elements of the situation are expressed by pronominals"), yet the prefixes still depend on what kind of thing each participant is (a boat, an island, a person). So the rank features must be on the bundles themselves. The comparison can then be made word by word at A: each word with an object prefix (the verb, and each postposition) runs its own Agree with the subject bundle. The tier ranking itself (which human is proximate, by topic) is a clause-wide, B-level choice. The word reads that choice; it does not make it. The changeover point is the same as before: rank is set at B, and the prefix is spelled at A.

9. Rules

OP-1  Obviation tier: Proximate > Obviative1 > Obviative2 …; each ordinary 3rd person      SOURCED (Aissen §2.2)
      links to one relation; same relation iff presupposed coreferential.
OP-2  Exactly one proximate per span; 4th persons are off the tier.                          SOURCED (Aissen fn. 15, §4.1, §6, §7)
OP-3  BI = proximate nonsubject, YI = obviative nonsubject (direct and postposition objects). SOURCED (Aissen (8), fn. 5)
OP-4  Aissen discusses no clause with >2 3rd persons and only one thing-subject clause, (16). SOURCED (by inspection of the full RTF)
OP-5  OP-1 to OP-3 predict one YI in Saxon (38)/(39) (Y&M 1987:227, 151); both are BI.       HYPOTHESIS on SOURCED data
      → contradicting data for OP-2 + OP-3 together (SX-8 confirmed against the text).
OP-6  R1 (PP = own span), R2 (4th person), R3 (thing-subject off-tier): each fails.          HYPOTHESIS
OP-7  R4 subject-relative yi-: a 3rd nonsubject is YI iff an on-tier 3rd subject outranks    HYPOTHESIS (fits Hale 2000:70 via
      it; else BI (general form). Same as OP-3 in all two-3rd-person clauses.                  Saxon fn. 7; Saxon SX-1; Aissen fn. 7)
OP-8  Under OP-7, Perkins (8c) follows from animacy; "one bi- per clause" is not shown.      HYPOTHESIS; T62a OPEN
OP-9  Residue: (38) island, (40) 'down to it' are BI where OP-7 + Creamer's mover rule      HYPOTHESIS; T62b OPEN
      predict YI (rider effect or speaker variation).
OP-10 Creamer (c23) yik'i vs Perkins (32) báhátis: rival thing-on-thing patterns.            SOURCED data; reading OPEN (T62c)

10. Derivations, step by step

Labels: Σ role schema, A word, B clause. VPISH throughout: ρA/ρS is first-Merged in Spec,vP/VP, ρP as complement, and each postposition Merges with its own object. Aissen's flat-S (Jelinek 1984) assumptions are not used; she says they are not crucial. Overt nouns are linked at B (nominal adjunction). The tier is a separate layer read at B.

O-1. (5) Saxon (39), wagon, grandfather, 'them': Aissen as stated (fails)

  1. Σ Σ = {ρS: 3, wagon, inanimate; ρwith: 3sg human (grandfather); ρG: 3pl human ('them', the ones caught up to)}. The role labels are ours; Saxon gives glosses only.
  2. B Merge V 'catch up to' with its P-object ρG; Merge PP [ρwith -ił]; Merge ρS in Spec,VP.
  3. B Tier (OP-1): three referents → [prx], [obv1], [obv2]. (14) keeps the wagon off [prx]. Topic chooses between the two humans: say grandfather [prx], 'them' [obv1], wagon [obv2]. The other choice gives the mirror image.
  4. A Agree (OP-3): -ił reads grandfather [prx], so bi- → bił ✓. The verb reads 'them' [obv1], so yi-. Printed: biníbą́ą́z ✗. Either choice in step 3 puts one human at [obv] and so predicts one yi-.

O-2. (5) again, with R4 (works)

  1. Σ as O-1.
  2. B as O-1, steps 2–3. The wagon is at [obv2], below both humans.
  3. A Agree (OP-7) on -ił: does the subject bundle outrank ρwith? No ([obv2] < [prx]), so the general form bi- → bił ✓.
  4. A Agree (OP-7) on V: does the subject outrank ρG? No ([obv2] < [obv1]), so bi- → biníbą́ą́z ✓. The ranking between the two humans does not show on either prefix.
  5. B Linear order: as printed, 'my grandfather', 'this morning', 'wagon', then the PP word and the verb word (right-headed). Saxon's "inanimate subject precedes the (postpositional) object" holds for the wagon and bił.

O-3. Perkins (8b) 'The girl is pulling the bike with the boy [on it]' (bił + y-verb), with R4

  1. Σ Σ = {ρA: 3sg human (girl); ρP: 3, bike; ρwith: 3sg human (boy)}.
  2. B Merge as in O-1. Tier: the boy is the topic, so (12) gives boy [prx], girl [obv1], bike [obv2] ((14) keeps the bike low).
  3. A Agree on -ił: the subject [obv1] does not outrank the boy [prx], so bił ✓. On V: the subject [obv1] outranks the bike [obv2], so y-verb ✓.
  4. A (8c) would need a b-verb, i.e. the bike outranking the girl, which (14) excludes, so * ✓. No "one bi-" rule was used.

O-4. Creamer (c23) Tsé tʼiis yikʼi chʼínímááz and Perkins (32) Jooł tsin báhátis yiltsʼid

  1. Σ Σ = {ρS: 3, moving thing (rock / ball); ρG: 3, stationary thing (tree / log)}.
  2. B Tier, Creamer's speakers: the group-7 rule ranks the mover higher, so rock [prx], tree [obv]. Tier, Perkins's speakers: the thing-subject is not ranked above the other thing (in R4 terms it is not an on-tier subject; in Saxon's terms it stays in VP).
  3. A Agree (OP-7) on the postposition: Creamer, the subject outranks it, so yikʼi ✓. Perkins, no outranking subject, so báhátis ✓. One rule, two settings of a B-level ranking choice (OP-10, OPEN which is right or whether both are).

11. Grant's ideas

GRANT'S CONJECTURE Feature bundles as arguments. Consistent, with one datum that leans its way. Saxon (38) Bił bináʼeeł has no nouns at all, but its two prefixes still depend on what the unspoken participants are. So rank (animacy, motion) must be readable off the role bundles, with no noun needed (§8). R4 compares two bundles (subject and nonsubject), which is a bundle-to-bundle Agree. Aissen's own framework links "nominals" to the tier, but she says the pronoun-vs-noun question is not crucial to her. That is a framework choice, not contradicting data.

GRANT'S CONJECTURE Strata. Consistent. Aissen's tier is explicitly "a structural dimension distinct from constituent structure or grammatical function" (§2.2), i.e. a separate layer laid over the roles. The roles never change in (1)–(6). Only the tier ranking does, and topic, a discourse property that arrives late, can reorder it ((8b)).

GRANT'S CONJECTURE Merge/Agree sites inside the word. OPEN. Under R4 each word with an object prefix (the verb, -ił, -kʼi) runs its own Agree with the subject bundle (O-2, steps 3–4). If those prefixes sit inside the word, that is an Agree site inside the word. This is a HYPOTHESIS of this site, not addressed by Aissen or Saxon.

Chemistry: SPECULATIVE, flag only. Changing "the one top-ranked item gets the mark" to "each item is compared with one reference item" is the difference between a global maximum and a pairwise comparison against a fixed reference. No chemical correlate is identified or claimed.

12. Bottom line

  • SOURCED Aissen 2000 (full RTF with footnotes): exactly one proximate per span (fn. 15, §4.1, §7); BI/YI are nonsubject proximate/obviative pronouns (8), fn. 5; BI is the overt form with non-3rd subjects (fn. 7). She discusses no clause with more than two 3rd persons, and only one thing-subject clause ((16), sheep/water, BI). T61e answered: she says nothing directly about thing-subjects with two bi-s.
  • HYPOTHESIS Applied as stated, her rules predict one yi- in Saxon (38)/(39); both prefixes are bi-. SX-8 is confirmed against her text. Repairs R1–R3 fail. R4 (yi- = outranked by the on-tier subject; bi- elsewhere, Hale's "general" form) fits every datum checked except the residue OP-9. It agrees with Aissen on all two-3rd-person clauses, so earlier pages' results (possessor-binding, possessor-gap, possessor-yi) are unaffected.
  • HYPOTHESIS Perkins (8c) is explained by animacy under R4, so "one bi- per clause with a human subject" is not independently shown.
  • OPEN Creamer and Perkins give rival patterns for thing-on-thing clauses (OP-10). D2 (deep vs shallow bi-) is not affected.

13. Fidelity notes

  • Aissen's Navajo is printed as decoded from the RTF's SIL IPA93 codes. (1) is printed yooį́, with no glottal stop, as in her (13a). Her (13b) and (2) print booʼį́ with one; both are kept as printed. The RTF's \lquote glottal stop in (16) is printed as ’.
  • Aissen's prose and footnotes are quoted from the RTF. Page numbers in the printed volume (pp. 129–150) were not available, so citations are by example, section and footnote.
  • Saxon (32) was read at 400 dpi: Jooł tsin báhátis yiltsʼid .... Her (33) is described, not quoted, because 'rock' prints garbled ("tst") in the scan. (40) is described, not quoted (faint 'with' word).
  • Creamer's "*" marks the dispreferred form ("one structure is not preferred", pp. 36–37), not necessarily outright ungrammaticality.
  • Perkins 1978 and Young & Morgan 1987 were not opened. Their data are as Saxon prints them.
  • Role labels (ρwith, ρG) and tier rankings in the derivations are ours.

14. Open questions

IDQuestionWhat would settle it
T62aIs there any clean evidence for "one bi- per clause" with a human subject, once animacy is controlled? (Perkins (8c) is confounded.)A sourced clause with a human 3rd subject and two human 3rd nonsubjects, both topical: is bi- … bi- possible? OPEN
T62bThe rider effect: do (38)/(40) have bi- on the thing-object because a human rider is present?A minimal pair: a vehicle subject + thing object, with and without a 'with' rider (Y&M 1987 pp. 151, 227, 475 first-hand). OPEN
T62cCreamer (c23) yikʼi vs Perkins (32) báhátis: speaker variation, the postposition, or context?Perkins 1978 pp. 55, 136 first-hand; any source that judges both forms of one thing-on-thing sentence. OPEN
T62dDoes Aissen 1997 (the tier framework) or a later Aissen paper discuss spans with three 3rd persons, or a subject-relative spell-out?Aissen 1997 (Language 73) first-hand. OPEN Update 2026-10-26 ~12:46pm PT: answered on ranked-obviatives. 1997 has ranked obviatives for three or more 3rd persons, and a subject-vs-object direct/inverse comparison for Algonquian verbs, but no Navajo rule. R4 is that comparison (R5, HYPOTHESIS).
T62eHale 2000:70–71 first-hand: is his "general third person object prefix b-" the R4 idea?Hale 2000 (same volume family as T61f). OPEN

Sources: Aissen 2000 (author's RTF, full text and fn. 1–15); Saxon 2003 pp. 343, 352–354, 356 (page images pp. 352, 354); Creamer 1974 pp. 36–37 (page images). Carried over: Willie 1991 (79)/(80), (85)/(86); Perkins 1978 and Y&M 1987 via Saxon. See also two-osv, eight ranks, possessor-binding, possessor-gap, inverse, open questions, sources.

Three third persons on Aissen's 1997 tier: ranked obviatives, and a direct/inverse check made word by word (T62d)

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.

Working note. Kept as written (URL, data and status labels unchanged). Filed under Obviation: how many third persons; the consolidated argument is the essay obviation. Rule IDs: rules index.

Restated on the unified structure (2026-11-15). The inverse is restated as one extra Move on the clause spine: the patient moves (via an outer Spec,vP) above the agent, which has already raised to Spec,TP by the working default (H3, H6′, H21b). The event LF is the same as the direct; the trigger and landing site are OPEN. 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. Status on the spine: see unified structure, §4, inverse row. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

The last page (one "main" third person per clause?) found that Aissen's 2000 Navajo paper breaks down on sentences with three third persons, and it proposed a repair, R4: yi- means "outranked by the subject" and bi- is the default elsewhere. T62d asked whether Aissen's earlier, general paper (Language 1997, where the "obviation tier" comes from) says anything about three third persons, or about a rule that compares each person with the subject. This run read the relevant parts of that paper first-hand, from page images. In plain words: yes to both, but not for Navajo. In 1997 Aissen says a sentence can hold one "main" third person and several ranked "lesser" ones. She also says the Algonquian direct/inverse marker is chosen by comparing the subject with the object, not by asking which one is "main". R4 turns out to be that 1997 comparison, applied to Navajo. Her 1997 paper mentions Navajo only in passing. A simpler-looking fix that uses her ranked "lesser" persons without the comparison fails on a sentence she cites herself. One difference from Algonquian also comes out. An Algonquian verb makes the comparison once, but in Navajo every word that carries a yi-/bi- prefix seems to make its own.

On this page: 1 Question · 2 Sources · 3 What Aissen 1997 says · 4 What it says about Navajo · 5 R4 restated as her direct/inverse check · 6 A rival fix: ranked lesser persons alone · 7 One check per verb, or one per word? · 8 Rulesets A and B · 9 Rules · 10 Derivations · 11 Grant's ideas · 12 Bottom line · 13 Fidelity notes · 14 Open questions

Notation as on one-proximate. Σ = the role schema. A = the word level (prefixal, left-headed; W = a verb or postposition word). B = the clause level (right-headed). ρA/ρS/ρP = agent / sole / patient slots, ρwith = the object of 'with', ρG = another oblique object. [prx] > [obv1] > [obv2] = places on the obviation tier ("main" > first lesser > second lesser). Merge vs Agree: Merge builds structure: it puts two pieces together (Move is Merge again of something already built). Agree values features: a head reads person, number or tier rank off a bundle. This page is about Agree, specifically which two bundles a prefix compares.

1. Question

T62d (from one-proximate §14): "Does Aissen 1997 (the tier framework) or a later Aissen paper discuss spans with three 3rd persons, or a subject-relative spell-out?" The 1997 paper is on the box from the possessor-gap run (page images and OCR). That run used it for the owner/possessor gaps. This run reads the other parts: the tier itself (§2), the direct/inverse definitions (§3.1), ranked obviatives (§3.5), the minimal span (§3.2), and every mention of Navajo.

2. Sources

SOURCED Aissen, Judith 1997. "On the Syntax of Obviation." Language 73(4): 705–750 (JSTOR scan, on the box since 2026-10-24). Read this run: pp. 707–711 (§§1–3.1, examples (3)–(15), fn. 9–10), pp. 713–714 (§3.2, (22)–(24)), pp. 718–719 (§3.5, (34), fn. 16–17, §3.6 (35)), p. 734 ((94)–(95), fn. 30–31), p. 736 (§5.1) and p. 743 (§6, fn. 40–41). The quotations below from pp. 709, 711, 718 and 719 were checked against the page images, not just the OCR.

SOURCED Re-checked on the box for the Navajo forms used here: Saxon 2003 (WECOL 2001 vol. 13) p. 354, the page image of (38). Aissen 2000, author's RTF, (13a).

Carried over, not re-opened: Perkins 1978:116 (8a–c) as printed by Saxon 2003 p. 343 (left schematic, as on two-osv); Saxon (39); Creamer 1974 (c21)–(c26); Willie 1991 (79)/(80), (86). The analysis of all of these is on one-proximate.

Not opened: Frantz 1966 (Blackfoot); Dahlstrom 1991, 1995; LeSourd 1976; Thompson 1989, 1994; any later Aissen paper; Hale 2000.

3. What Aissen 1997 says SOURCED

3a. Several lesser third persons, ranked

  • The tier (§2, p. 709): "These relations have the names {Prox(imate), Obv(iative)1, Obv(iative)2, …}. Each third person nominal in a sentence bears an obviation relation to an obviation span. I assume that an obviation span contains at most one proximate, one obviative1, etc. This is consistent with descriptions in the literature which claim that within the relevant span there is only a single proximate third person, but possibly several obviatives …; I make the further assumption that the multiple obviatives correspond to distinct, ranked, relations (§3.5)."
  • Three third persons (§3.5, p. 718): "In obviation spans involving more than two third persons, only one third person can be proximate, the others must be obviative. … one would expect multiple obviatives within a span to be ranked." Her example is the English proposition 'that young man said to Scarface's brother-in-law' (from Frantz 1966, Blackfoot), with three third persons. With a direct verb the linkage is (34): young man → Proximate, Scarface → Obviative1, brother-in-law → Obviative2. "Blackfoot provides visible evidence that the two obviative relations in 34 are distinct, for they are marked by different morphology." Fn. 17: this "further obviative" is "marked only in genitive constructions (i.e. on the possessum when the possessor is obviative)" (Potawatomi and Plains Cree too). No Blackfoot forms are printed.
  • Two lesser persons in one clause (pp. 718–719): "the two participants in a transitive proposition can both be obviative (the proximate being some other participant in the local discourse), and the verb either direct or inverse, depending on which obviative is subject and which object" (Fox, Plains Cree). Citing LeSourd 1976, "the subject obviative is more discourse prominent than the object obviative when the direct verb is used, while the reverse is true with the inverse verb. Consistency requires that ranked obviatives be assumed …" Then: "While ranked obviatives will not be discussed further in this article, their existence is assumed as part of the overall framework proposed here."

So, unlike the 2000 paper, the 1997 framework explicitly allows three or more third persons on one tier. It is still "at most one proximate", the same as 2000 fn. 15.

3b. Direct vs inverse is a comparison between two arguments

  • (§3.1, p. 711): "the direct form of the verb is used in Algonquian when the subject outranks the object on the hierarchy …: when the object outranks the subject, the inverse form is used." (12) "local person > proximate > obviative". (13) "REL(ATIONAL) HIER(ARCHY) [tentative]: subject > primary object".
  • (14) "DIRECT ALIGN (X, Y) iff ¬∃α∃β | α > β on X and β > α on Y." (15) "INVERSE ALIGN (X, Y) iff ∃α∃β | α > β on X and β > α on Y." Note that (14) is a negative condition (no mismatched pair), while (15) is a positive one (a mismatched pair exists).
  • Fn. 10: in ditransitives "there are two nonsubject arguments"; the recipient-type "primary object" is the one that counts, and "properties of the primary object, but not the secondary object, are relevant to the distribution of direct and inverse in Fox." So in Algonquian the check is made once per verb, between the subject and one object.
  • Third persons only (fn. 41, p. 743): "Kutenai also restricts hierarchy effects to the third person (Dryer 1992). Like Tzotzil, …" So a system in which only 3rd-vs-3rd pairs are compared is an attested type in her survey.

3c. Where a span ends

(24) MINIMAL SPAN (p. 714) puts a head and its arguments in one span, so subject, object and a possessor share a span. She calls it "an initial (i.e. tentative) constraint" and asks: "is it correct to distinguish arguments from adjuncts, as intended here?" (94)–(95) and fn. 31 (p. 734) are stated "so as NOT to force a main clause and an adjunct into the same obviation span". 1997 therefore leaves room for R1 ("a postposition phrase is its own span"). R1 still fails on data: Perkins (8a) has yił under a human subject (one-proximate §6).

4. What it says about Navajo SOURCED

Navajo comes up twice and has no example sentences. On p. 736 (§5.1): "Some of the diagnostics in 97 have also been noted in Navajo (Athapaskan). A number of studies, starting with Hale 1973, have documented animacy-based gaps, and Navajo also has coreferential blockages involving object pro with genitive antecedent (Platero 1982; Hale et al. 1977). The clustering of these properties suggests the relevance of obviation, a proposal made for somewhat different reasons by Thompson (1989, 1994)." On p. 743 (§6), Navajo is listed among the head-marking languages discussed. There is no YI/BI spell-out rule, no example with three Navajo third persons, and nothing on a thing-subject.

Answer to T62d SOURCED: (i) Spans with three third persons: yes, in general (§2, §3.5). There is one proximate and the rest are ranked obviatives, Obv1 > Obv2, and two obviatives can share a clause. Her only worked example is from Blackfoot (via Frantz 1966) and has no printed forms. (ii) A subject-relative spell-out: yes, for the Algonquian verb (direct/inverse, (12)–(15)). It compares the subject with the primary object and is made once per verb (fn. 10). (iii) Navajo: none of this is applied. Navajo appears only as a language with obviation-like diagnostics (p. 736). The 2000 paper is where she maps the tier to Navajo, and there the mapping is absolute (BI = proximate, YI = obviative).

5. R4 restated as her direct/inverse check HYPOTHESIS

R4 from one-proximate said: a 3rd-person nonsubject is YI iff the clause's 3rd-person subject is on the tier and outranks it. Otherwise it is BI. Put in Aissen's 1997 terms:

  • Take X = the relational hierarchy (13), "subject > object", and Y = the tier ([prx] > [obv1] > [obv2]). Leave local persons off Y, as in Kutenai and Tzotzil (fn. 41). Navajo 1st/2nd/4th subjects never give YI (Aissen 2000 fn. 7; 4th persons are off the tier, 2000 §6).
  • R5 (= R4 restated). For a word W with a 3rd-person object prefix: YI iff ∃ subject α and W's object β, both on the tier, with α > β on Y. BI elsewhere.
  • R5 cannot simply be her DIRECT ALIGN (14), because (14) is negative. With a 1st-person subject and a 3rd-person object nothing is mismatched on a 3rd-only tier, so (14) is vacuously satisfied and would wrongly give YI. Navajo yi- needs the positive counterpart, the mirror image of her INVERSE ALIGN (15). This matches what the Navajo sources say: yi- is the marked form that needs a 3rd-person subject (Saxon's SX-1; Hale 2000:70 via Saxon fn. 7, "the general third person object prefix b-"), and bi- is the elsewhere form.
  • Whenever subject and object are two distinct 3rd persons on one tier, they hold different relations (her co-linking condition (10), p. 710). Then exactly one of "subject > object" and "object > subject" holds, and R5 gives YI = direct and BI = inverse. That is the traditional direct/inverse reading of yi-/bi- for 3-on-3 clauses.

So R4 is not an ad hoc patch. It is the comparison Aissen herself uses for Algonquian in 1997. What is new on this site is applying it to Navajo prefixes and restricting it to the 3rd-person tier. Aissen 2000 instead linked the prefix to the object's own tier rank. With only two 3rd persons the two readings give the same answer, and they differ only when there are three or more (one-proximate §6).

6. A rival fix: ranked lesser persons alone HYPOTHESIS on SOURCED data

1997's ranked obviatives suggest another fix that keeps Aissen 2000's absolute reading. Call it R6: BI = [prx] or [obv1], YI = [obv2] or lower ("only the second lesser person gets yi-"). It fixes the three-person sentence (39): grandfather [prx] → bił ✓, 'them' [obv1] → bi- verb ✓. But it fails on two-person sentences:

Aissen 2000 (13a), citing Hale 1973:302 (RTF on the box, SIL IPA93 decoded; as printed on one-proximate (1))
(1)dinédziłyooį́.
manmountain3OBJ(yi)-see
'The man sees the mountain.' (BI is starred.)

Two 3rd persons: man [prx], mountain [obv1] (2000 (14), Avoid an inanimate proximate). R6 predicts BI on the verb. Printed: YI ✗. Perkins (8a) via Saxon ('The boy is pulling the bike with the girl [on it]': the boy, the subject, is [prx]; the girl riding is [obv1]) gets the same wrong prediction: R6 gives BI on 'with', but it is printed yił.

R5 gets (1) right: man [prx] > mountain [obv1], so the subject outranks the object, giving YI ✓. So the 1997 ranked obviatives are needed for (39) (the two humans must hold distinct places), but they are not enough. The prefix has to compare the object's place with the subject's place, which is the 1997 direct/inverse idea. R6 is rejected.

7. One check per verb, or one per word? HYPOTHESIS on SOURCED data

In Algonquian the direct/inverse sign is a single suffix on the verb, and only the primary object counts (fn. 10). Navajo has a yi-/bi- prefix on the verb and on each postposition, and they can come out differently in one clause:

  • Perkins (8b) via Saxon p. 343 (schematic, verb and 'bike' left unprinted as on two-osv): 'The girl is pulling the bike with the boy [on it]': bił 'b-with' + y-verb. The postposition comes out inverse (the boy [prx] outranks the subject) and the verb comes out direct (the subject outranks the bike).
  • Saxon (38), page image re-read this run:
Saxon 2003 p. 354, (38), citing Young & Morgan 1987:227
(2)Biłbináʼeeł.
b-withb-float around
3OBJ(bi)-with3OBJ(bi)-float.around
'It (a boat) floats around it (e.g. an island) with him.'

Two words, two prefixes, each bi-. Saxon p. 354: "all of the elements of the situation are expressed by pronominals".

A single clause-wide value (one Algonquian-style theme sign) cannot give bi- on 'with' and yi- on the verb in (8b). A comparison made once for each word, between the subject and that word's own object, gives both. This is a real difference between Navajo and the Algonquian system Aissen models. It is not a defect in her 1997 account, which never claims Navajo works this way. What it shows is where the comparison lives: in the word.

Update 2026-10-26 ~8:46pm PT: Hale's own wording, read first-hand on hale-general-b: y- is "limited strictly to situations in which a third person object is in the c-command domain of a clausemate third person subject", and b- is "in fact the general third person object agreement marker" (SOURCED). That is R5 with c-command in place of tier rank, and the two are the same rule if rank = height (BR, HYPOTHESIS). Hale's causative (49)–(51) adds a one-slot-per-word limit that fits RO-8 (T63c partly answered).

8. Rulesets A and B HYPOTHESIS

Split R5 into its two parts:

  • B (clause, right-headed): build the tier. Which 3rd person is [prx], [obv1], [obv2] depends on topic, animacy and Creamer's mover rule, i.e. on the whole clause and its discourse. Aissen 1997 calls the tier "a dimension of linguistic representation" (p. 709) apart from constituent structure. It is set once for the span.
  • A (word, left-headed prefixal): compare and spell out. Each word with an object prefix (the verb, each postposition) runs one Agree: it reads the tier rank on its own object's bundle and on the subject's bundle, and spells out yi- if the subject is higher and bi- otherwise. The word is the domain of the comparison (§7).

So the word is where the changeover happens. The tier ranking is a B-level object, and each A-level word reads it once for itself. The same division was proposed on one-proximate §8. This page adds two things: a source-backed reason why the comparison is relative (Aissen 1997 (12)–(15)), and the (8b) evidence that it is computed per word rather than per clause.

9. Rules

RO-1  Tier relations {Prox, Obv1, Obv2, …}; at most one Prox; several Obv, ranked.        SOURCED (Aissen 1997 §2 p. 709; §3.5 p. 718)
RO-2  >2 third persons: one Prox, the rest ranked Obv (34); two Obv may be coarguments,      SOURCED (1997 pp. 718–719)
      with direct/inverse decided by Obv1 vs Obv2.
RO-3  Direct/inverse = comparison of subject and primary object on the participant hier.     SOURCED (1997 (12)–(15), fn. 10)
      (14) negative, (15) positive; made once per verb in Fox.
RO-4  Hierarchy effects restricted to 3rd persons are attested (Kutenai, Tzotzil).          SOURCED (1997 fn. 41)
RO-5  1997 mentions Navajo only as showing obviation diagnostics; no YI/BI rule, no example. SOURCED (1997 p. 736, p. 743)
RO-6  R5 = R4: YI on W iff ∃ on-tier 3rd subject outranking W's 3rd object; BI elsewhere.     HYPOTHESIS (built from RO-3, RO-4)
      Positive (15)-style, not (14): (14) would wrongly allow YI with 1st/2nd subjects.
RO-7  R6 (BI = Prox or Obv1; YI = Obv2+) fixes (39) but fails (1) Aissen 2000 (13a).          HYPOTHESIS on SOURCED data → rejected
RO-8  Navajo makes the comparison once per word, not once per clause: Perkins (8b).          HYPOTHESIS on SOURCED data
RO-9  Tier at B (clause); comparison and spell-out at A (word).                              HYPOTHESIS

10. Derivations, step by step

Labels: Σ role schema, B clause, A word. VPISH: ρA/ρS first-Merged in Spec,vP/VP, ρP as complement, each postposition Merges with its own object. Overt nouns are linked at B (nominal adjunction). Aissen's 1997 trees are not used. Her tier conditions are stated on grammatical relations, so nothing depends on a flat S. Tier assignments in the derivations are ours.

RD-1. (1) diné dził yooį́ 'The man sees the mountain'

ΣΣ = {ρA: 3sg human; ρP: 3, inanimate}.
BMerge V with ρP; Merge v; Merge ρA in Spec,vP. (Merge builds the structure.)
BTier: two distinct referents → two relations (co-linking (10)); the animate is [prx], the mountain [obv1] (2000 (14)).
AAgree on V: read ρA [prx] and ρP [obv1]. Subject > object, so RO-6 gives yi- → yooį́ ✓. (R6: [obv1] → BI ✗.)
BLinear order: nouns before the verb word, subject first (right-headed clause).

RD-2. (2) Bił bináʼeeł 'It (a boat) floats around it (an island) with him'

ΣΣ = {ρS: 3, boat, inanimate; ρG: 3, island; ρwith: 3sg human}. No nouns are chosen at all.
BMerge V 'float around' with ρG; Merge PP [ρwith -ił]; Merge ρS.
BTier: three referents → [prx], [obv1], [obv2] (RO-2). The rider is [prx] ((14) keeps both things off [prx]). The boat and the island fill [obv1]/[obv2]. The order between them is open; see RD-2′.
AAgree on -ił: subject (boat) vs ρwith [prx]. The subject does not outrank it → bił ✓.
AAgree on V: subject vs ρG (island). If the island is above the boat → bi- ✓. If the boat is above (Creamer's mover rule) → yi- ✗.

RD-2′. This is the residue OP-9 from one-proximate, still unsolved. The 1997 framework does not decide how two inanimate obviatives are ranked. LeSourd's "more discourse prominent" (via Aissen pp. 718–719) is a discourse notion, and (2) has no context. R5 is consistent with (2) only if the island outranks the boat here (T62b, rider effect, still OPEN).

RD-3. Perkins (8b), schematic: girl pulls bike, boy riding

ΣΣ = {ρA: 3sg human (girl); ρP: 3, bike; ρwith: 3sg human (boy)}.
BMerge as in RD-1, plus PP [ρwith -ił]. Tier: boy [prx] (topic), girl [obv1], bike [obv2].
AWord 1, -ił: girl [obv1] vs boy [prx]. The subject is lower → bił ✓.
AWord 2, V: girl [obv1] vs bike [obv2]. The subject is higher → y-verb ✓.
—A one-value-per-clause rule (fn. 10 style) would have to give both words the same value ✗. That is RO-8.

11. Grant's ideas

GRANT'S CONJECTURE Feature bundles as arguments. Consistent. R5 compares two bundles' tier ranks, and (2) has no nouns at all, so the ranks must sit on the role bundles. Aissen 1997's condition (7) is stated for "nominals" ("A nominal Nomn bears a relation … iff Nomn is third person"). That is how she frames it, not a datum against bundles. Silence is not counted.

GRANT'S CONJECTURE Strata. Consistent, and somewhat sharper than before. The roles come first (Σ), the tier is a separate layer (1997 p. 709, "a dimension of linguistic representation"), and spell-out reads both. In RD-3 the roles never change. Only the tier order (a topic choice) changes the prefixes, which fits a late, separate layer. Noun choice is not needed at any step.

GRANT'S CONJECTURE Merge/Agree sites inside the word. Some support, as a HYPOTHESIS. RO-8 puts one Agree site in each word that carries an object prefix. If the prefix is inside the word, which it is for the verb and for a postposition with its prefix, then the comparison is an intra-word Agree site: the word reads two bundles and spells out one prefix. No source claims this. It comes from (8b) as Saxon prints it.

Chemistry: SPECULATIVE, flag only. "One sign per verb, from one pair" vs "one sign per site, each comparing its own neighbour with a shared reference" is the difference between a single global label and a local, per-site comparison against a common reference. No chemical correlate is identified or claimed.

12. Bottom line

SOURCED Aissen 1997 allows spans with more than two third persons: one proximate and several ranked obviatives (§2 p. 709, §3.5 pp. 718–719, (34)), with two obviatives possible in one clause. Its direct/inverse is a comparison of subject and primary object ((12)–(15)), made once per verb (fn. 10), and some languages restrict it to 3rd persons (fn. 41). It gives no Navajo YI/BI rule or example (p. 736). T62d answered.

HYPOTHESIS R4 is her 1997 direct/inverse comparison applied to Navajo prefixes, on a 3rd-only tier, in its positive form (R5). The rival R6 (ranked obviatives with an absolute spell-out) fails on Aissen 2000 (13a). Perkins (8b) shows that Navajo makes the comparison once per word, not once per clause, which places it at ruleset A.

OPEN How two inanimate obviatives are ranked (RD-2′ = OP-9/T62b). The both-obviative test (T63a). Whether any Navajo form shows Obv1 vs Obv2 directly (T63b).

13. Fidelity notes

  • Aissen 1997 quotations on pp. 709, 711, 718, 719 were checked against page images. Those on pp. 710, 713–714, 734, 736, 743 are from the OCR of the same scan, with OCR noise fixed only where the page image or context left no doubt. The formulas (14)/(15) are as printed on p. 711.
  • Aissen 1997 has no Navajo forms. The Navajo forms here are Aissen 2000 (13a) (RTF, decoded as on one-proximate) and Saxon (38) (page image p. 354). Perkins (8b) stays schematic.
  • The Blackfoot example is cited only as Aissen gives it (English proposition and linkage (34)). Frantz 1966 was not opened.
  • R5 and R6 and all tier assignments in the derivations are this site's, not Aissen's.

14. Open questions

IDQuestionWhat would settle it
T63aThe both-obviative test: in a clause whose two 3rd persons are both non-topical (the proximate is someone else in the discourse), does the prefix track which of them is more prominent? Aissen 2000 predicts YI regardless (both obviative). R5 predicts BI if the object is the more prominent of the two.A sourced Navajo text passage with a third, topical 3rd person and a bi- clause between two non-topical ones (Y&M texts, Willie's or Thompson's discourse data). OPEN
T63bDoes Navajo show Obv1 vs Obv2 anywhere directly, e.g. a 'his' inside 'his' (owner of an owner), the way Blackfoot's further obviative does?A sourced Navajo double-possessive with three 3rd persons and the prefixes printed. OPEN
T63cPer-word comparison (RO-8): is there any Navajo clause where the verb's prefix and a postposition's prefix must agree (one clause-wide value)?Perkins 1978 first-hand, and Willie 1991's postposition chapters checked clause by clause. OPEN
T63dThompson 1989/1994 (cited by Aissen 1997 p. 736 for Navajo obviation): does he state a relative (subject-compared) rule?Thompson first-hand. OPEN

Sources: Aissen 1997 Language 73 pp. 707–711, 713–714, 718–719, 734, 736, 743 (page images pp. 709, 711, 718, 719); Saxon 2003 p. 354 (page image, (38)); Aissen 2000 RTF (13a). Carried over: Perkins (8a–c) via Saxon; Saxon (39); Creamer 1974. See also one-proximate, two-osv, possessor-gap, inverse, Algonquian inverse, open questions, sources.

Hale's "general" b-: the workshop text behind "Hale 2000:70–71", and whether it is the subject-relative yi- (T62e)

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.

Working note. Kept as written (URL, data and status labels unchanged). Filed under Obviation: how many third persons; the consolidated argument is the essay obviation. Rule IDs: rules index.

Verbs on this page → dictionary: yoolóós ‘lead (one animate)’ · nímááz ‘rolled; with yaa/baa ‘to’: rolled to’

Restated on the unified structure (2026-11-15). Voice, causatives and events are restated with every theta role assigned inside vP: the classifier spells a value of v⁰ (H59, distribution NOT UNDERSTOOD), a causer is an external argument in Spec,vP, and the passive keeps a silent indefinite agent (H10b). This page’s causative analysis has not yet been restated on the unified structure (OPEN). 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. Flat-S and other pre-VPISH notation below is labelled contrast. Status on the spine: see unified structure, §4, causatives and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

The last two pages repaired Aissen's account with R4/R5: yi- means "outranked by the subject", and bi- is the general form everywhere else. They kept quoting Hale's phrase "the general third person object prefix b-", but only second-hand through Saxon. T62e asked for Hale himself. Hale's 1997 Navajo workshop notes (hale023) have been on the box since 2026-10-15. This run shows they contain, word for word, the passages that Saxon 2003 quotes as "Hale 2000:70" and "Hale 2000:71, fn12". Read in full, Hale's rule is subject-relative. But his relation is structural ("in the c-command domain of a clausemate third person subject"), not a ranking.

In plain words: Hale says bi- is the ordinary "him/her/it" prefix. yi- is special: it shows up only when a real third-person doer sits above the object in the sentence's structure SOURCED. That matches the repair on the last two pages, provided "ranks higher" and "sits higher" turn out to be the same thing HYPOTHESIS. The prefix can't be read off the order of the nouns, though: two sources give the same noun order with opposite prefixes. So the "above" has to be between the hidden pronoun slots, not the spoken nouns HYPOTHESIS on SOURCED data. Hale's causative sentences add a limit: one verb word has room for only one object-agreement prefix of this kind SOURCED.

On this page: 1 Question · 2 Sources · 3 What Hale says · 4 A correction: who is "removed" · 5 Hale's rule vs R4/R5 · 6 Does noun order decide? · 7 The causative: one slot per word · 8 Rulesets A and B · 9 Rules · 10 Derivations · 11 Grant's ideas · 12 Bottom line · 13 Fidelity notes · 14 Open questions

Notation as on one-proximate and ranked-obviatives. Σ = the role schema (Dowty's proto-roles as the baseline; the Navajo slots are extensions of it). A = the word level (prefixal, left-headed; W = a verb or postposition word). B = the clause level (right-headed). ρA/ρS/ρP = agent / sole / patient slots, ρwith/ρfor = the object of 'with'/'for'. [prx] > [obv1] > [obv2] = tier places. H = Hale's rule as stated in §3. Merge builds structure (it puts two pieces together). Agree values features (person, number, obviation) between pieces that are already built. Hale's EP/MP/VP labels are his: E is his "edge" head, where object agreement sits, and M is Mode.

1. Question

T62e (from one-proximate §14): "Hale 2000:70–71 first-hand: is the 'general third person object prefix b-' the subject-relative idea?" The printed Hale 2000 (Fernald & Hale, eds., Diné Bizaad Naalkaah, MITWPL ELL 3: 55–97, as listed in Saxon's bibliography) is not on the box, and no open copy was found. What is on the box is Hale's 1997 workshop text with the same title. If Saxon's quotations occur in it verbatim, it can stand in for those pages. They do (§2).

2. Sources

SOURCED Hale, Ken 1997. "Remarks on the syntax of the Navajo verb", Parts I–III, NLA workshop, Tsaile, July 1997 (MIT papers hale023.pdf, 61 PDF pages, born-digital; on the box since 2026-10-15). Read this run: PDF pp. 16–18 (the obviation passage, fn. 9–11, tree (5)), pp. 21–22 (trees (10)–(11)), p. 27 ((17)–(18), fn. 17), and pp. 54–59 (the causative, (40)–(54)). Page images checked for pp. 16, 17, 27 and 58. The other pages come from the text layer, which matched the images wherever both were checked.

HYPOTHESIS (strongly supported) hale023 = the text Saxon cites as Hale 2000. Saxon 2003 fn. 6 quotes "removing it from the domain in which obviation applies, forcing use of the general third person object prefix b-" (Hale 2000:70). That string occurs verbatim at hale023 PDF p. 17. Saxon p. 356 quotes "the DP appearing in the highest Spec or adjunct position in the inflectional portion of the extended projection" (Hale 2000:71, fn12). That string occurs verbatim in hale023 fn. 11, PDF p. 18. The order and spacing also fit: printed p. 70 ≈ PDF pp. 16–17, and p. 71 ≈ PDF p. 18. The footnote number differs (12 vs 11), so the printed version is not identical. It probably has one more note earlier (T64a).

SOURCED Re-checked on page images this run: Willie 1991 PDF p. 101 = printed p. 95, (85)/(86) at 300 dpi. Saxon 2003 notes page (PDF p. 363 of the WECOL volume): footnote markers 5 (Broadwell & Duncan) and 6 (Hale).

Carried over, not re-opened: Aissen 2000 (13a), (16); Saxon (38), (39), (32) (= Perkins 1978:55), Perkins (8a–c) via Saxon p. 343 (schematic); Creamer 1974 (c23). The forms and pages are on one-proximate, ranked-obviatives and two-osv.

Not opened: Hale 2000 as printed; Thompson 1989, 1994; Young 1996; Young & Morgan 1987 (cited by Hale for the indefinite ’-).

3. What Hale says SOURCED

  • The rule (PDF p. 16, page image): "The prefix b- is in fact the general third person object agreement marker. The obviative y- is much more restricted, being limited strictly to situations in which a third person object is in the c-command domain of a clausemate third person subject. In all other situations, b- is the agreement marker which appears, where overt." y- is "glossed 3o by Young and Morgan, reflecting its obviative function which, like that of its parallel in Algonquian … regularly marks clausemate third persons as referentially distinct."
  • Other subjects (fn. 9, p. 16): "If a third person object appears in the c-command domain of a first, second, or fourth person subject, … If the third person object morphology appears in E, then it is realized as Ø; otherwise, it is realized as b-."
  • The fourth person (fn. 17, p. 27, page image): "The fourth person shares zero-AGRs with the third person. It does not behave like a third person in obviation, however—only the 'true' third person (that realized pronominally as bí, for example) requires that a clausemate object in its c-command domain show obviative (y-) agreement."
  • 3-on-3 clauses (p. 16): "the object cannot remain in the c-command domain of the subject if the object agreement is b-."
  • Order and height (p. 17, page image): "We have seen that, in general, if some element X is to the left of Y, it is higher than Y." Tree (5): the object DP tsídii in Spec,EP, the subject DP mósí inside VP, and b- under AGRo in E, "in accordance with the above mentioned Navajo principle of obviative agreement, which restricts the obviative y- to objects c-commanded by a clausemate third person subject."
  • Direct clauses (pp. 21–22): the object raises to Spec,MP to agree with E, "the closest c-governor". Then: "Since the third person obviative agreement morphology y- appears in E, we can at the same time satisfy the requirement that the subject c-command the object in the obviative by raising the subject into Spec of EP" ((10) → (11)). So in Hale's direct clause the raised object is above the unraised subject, and only the subject's further raise to Spec,EP makes y- possible.
  • Cause or result? (p. 18): "Is b- really the cause of the movement, or is it instead just a by-product of it? If the latter is true, then the appearance of b- would be a result, not the cause … This seems more reasonable". The motive he gives for moving the object is topic: speakers at the 1974 workshop said the Inverse puts the object in the position of topic, and Platero's relative tsídii mósí biní¬ʼínígíí 'the bird that the cat is looking at' (as printed, cf. Hale, Jeanne & Platero 1977). Fn. 11 defines "clausal topic" as "the DP appearing in the highest Spec or adjunct position in the inflectional portion of the extended projection of V (i.e., in Navajo, the position dominated by EP or a segment of EP)".

Hale's own pair (PDF pp. 14–15; (2) cited from Y&M 1987G:65), printed with glosses on inverse-agree §2: (2) Mósí tsídii yiní¬ʼ²í 'The cat is looking at the bird' and (3) Tsídii mósí biní¬ʼ²í 'The cat is looking at the bird' / 'The bird is being looked at by the cat' (inverse; both translations Hale's). These are as printed: ¬ and ² are font substitutions and are not decoded here.

Answer to T62e SOURCED + HYPOTHESIS: Yes, in shape. In Hale's rule, b- is the elsewhere form, and y- requires a clausemate "true" third-person subject standing in a relation to that particular object. That makes it subject-relative and assigned object by object, which is exactly what R4/R5 say (SOURCED that Hale says this). The difference is the relation. In R4/R5 it is tier rank; in H it is c-command, i.e. height in the tree. §5 shows that the two agree if rank and height line up (HYPOTHESIS).

4. A correction: who is "removed" SOURCED

Read in its source, the "removing it" passage (p. 17) runs: "Or it might be, to follow suggestions by Margaret Speas and Eloise Jelinek, that the Inverse involves topicalization of the object, removing it from the domain in which obviation applies, forcing use of the general third person object prefix b-." Three things follow:

  • "it" is the object. One-proximate §6 paraphrased this (following the wording of Saxon's fn. 6) as "inverting the subject removes it". That is the wrong referent, and the page now carries a correction note.
  • It is one of two options Hale gives, and he credits the idea to Speas and Jelinek. On p. 18 he leans toward "b- as a result", with topic as the motive.
  • The obviation footnote in Saxon is fn. 6 (page image), not fn. 7 as one-proximate had it.

5. Hale's rule vs R4/R5 HYPOTHESIS

  • H (from §3): for a 3rd-person object β, y- iff β is in the c-command domain of a clausemate "true" 3rd-person subject. Otherwise b-, or Ø in E under a 1st/2nd/4th-person subject (fn. 9).
  • R5 (ranked-obviatives §5): YI on W iff an on-tier 3rd-person subject outranks W's 3rd-person object. BI elsewhere.
  • Points where they already agree (SOURCED in each source): 1st/2nd-person subjects never give y- (H: fn. 9; R5: off the 3rd-only tier, Aissen 2000 fn. 7). 4th-person subjects never give y- (H: fn. 17; R5: 4th persons are off the tier, Aissen 2000 §6). In both, b- is the general form.
  • Bridge BR (HYPOTHESIS): "α outranks β on the tier" ⇔ "α c-commands β at the point where β's prefix is valued". In Hale's terms, the proximate is the "clausal topic" (fn. 11: highest Spec or adjunct of EP), and rank is read off height. With BR, R5 and H are the same rule stated in two vocabularies.
  • What H adds: a structural reason for the thing-subject bi-s. In Hale's direct derivation the object first raises to Spec,MP, above the subject in VP (pp. 21–22). A subject that does not go on to Spec,EP therefore stays below the object, and H gives b-. Saxon's SX-4 says thing-subjects stay in VP (two-osv). Together these give bi- on a verb's object under a thing-subject (Perkins p. 136 via Saxon; Aissen 2000 (16a) dibé tó ʼabííł ʼéél), and they give both bi-s in Y&M (38)/(39) with no ranking among three 3rd persons. In the R5 vocabulary, this is the OP-9 residue (how two things are ranked) recast as "did the subject raise?" (HYPOTHESIS).
  • And a prediction: if a thing-subject can raise to Spec,EP, H allows yi- with it. Saxon's footnote pair from Y&M 1987:225 ('Rainbows surround the mountains', printed both with yi- and with bi-; T61c) has that shape. The forms stay schematic here because the page image is faint (T64b).

6. Does noun order decide? HYPOTHESIS on SOURCED data

Hale's "if X is to the left of Y, it is higher than Y" is stated "in general". Take it literally for the overt nouns and combine it with H. The sourced sentences then come out like this:

Willie 1991 p. 95 (PDF p. 101, page image at 300 dpi), (85)/(86)
(1)ʼashkiiʼatʼéédłį́į́ʼyáyoolóós
boygirlhorse3.for3.3.lead [sing. +animate]
boygirlhorse3OBJ(yi)-for3OBJ(yi).3SBJ-lead
'The boy is leading the horse for the girl.' (85)
(2)ʼashkiiʼatʼéédłį́į́ʼbáyoolóós
boygirlhorse3.for3.3.lead [sing. +animate]
boygirlhorse3OBJ(bi)-for3OBJ(yi).3SBJ-lead
'The girl is leading the horse for the boy.' (86)

Willie p. 95: "Here the Inverse gives the agent role to the internal argument, and the Benefactive role to the external argument. Note that the theme argument does not take part in the inversion process." The second gloss line is ours (Leipzig style). Willie's own gloss is the first.

SentenceNoun order (left = higher)H predictsPrinted
Hale (2) / (3)cat > bird / bird > caty- / b-✓ / ✓
Aissen 2000 (13a) diné dził yooį́man > mountainy-✓
Aissen (16a) dibé tó ʼabííł ʼéél 'the water swept the sheep off'sheep (object) > waterb-✓
Aissen (16b) *tó dibé ʼayííł ʼéélwater > sheepy- (the form printed)form ✓, but the star is not from H: it needs a separate animacy filter (Hale 1973 via Aissen (14)) that bans this structure
Willie (1) = (85)boy (subj) > girl (for) > horseyá, y-verb✓
Willie (2) = (86)boy (for) > girl (subj) > horsebá, y-verb✓ (one clause, two values)
Perkins (8a) / (8b) via Saxon (schematic)boy > girl > bike(8a) yił, y-verb; (8b) bił, y-verb✓ / ✓
Perkins (8c)* (schematic)boy (with) > girl (subj) > bikeverb y- (bike is below girl), so b-verb out✓ (the star follows from order alone)
Creamer (c23) Tsé tʼiis yikʼi chʼínímááz 'The rock rolled upon the tree'rock > treeyikʼi✓
Saxon (32) = Perkins 1978:55 Jooł tsin báhátis yiltsʼid … 'The ball rolled over the log'ball > logyi- on 'over'✗ bá-
Y&M (38) Bił bináʼeełno nouns—H over nouns is silent

Result. H over overt nouns fits everything except one pair. (c23) and (32) have the same order (thing subject, then thing object of a postposition) and opposite prefixes. Two sources with one order and two outcomes are contradicting data for "H applied to the spoken nouns, with left = higher". Two readings survive (HYPOTHESIS):

  • H over slots. H holds between the pronominal slots (ρ bundles). Overt nouns are linked later, at B, as on nominal adjunction, so their order is not the slots' height. In (32) the subject slot stays low in VP (SX-4) and the 'over' PP is attached above it, giving bá-. In (c23) either the rock's slot raised or the PP is low, giving yikʼi. Under this reading the (c23)/(32) clash becomes a question of attachment or raising, i.e. speaker or construction variation (T62c, T64c). It is no longer a clash of rules.
  • "True third person". Fn. 17 makes only a "true" third person a y-trigger. If Perkins's speakers treat a thing-subject as not "true" in this sense, (32) follows. Nothing in Hale says things fall outside "true" third persons (his example is bí), so this reading is OPEN and unsupported.

Under "H over slots", (8c)'s star no longer follows from order alone. It needs Saxon's SX-4 the other way round: a human 3rd-person subject must raise to Spec,EP, so it c-commands the bike's slot and the verb must be y-. The order-only account and the slots account make different predictions for a reordered (8c) with the bike before the girl (T64d).

7. The causative: one agreement slot per word SOURCED data, HYPOTHESIS reading

T63c asked whether any clause forces the verb's prefix and a postposition's prefix to share one value. Hale's causative section (PDF pp. 54–59) shows a different kind of limit. Here the postposition is inside the verb word. Hale's analysis (p. 57): the causee 'boy' is the object of a causative postposition -y, and its "agreement needs … are met locally, internal to the PP". Then "the causative P intrudes into the upper and left-most structure of verb word (i.e., that portion projected by E)". His forms (page image p. 58) are printed as printed: ² and ¬ are font substitutions and are not decoded. The morpheme breakdowns are his.

Hale 1997, PDF p. 58, (50a, b); (49b), (51b) p. 58
(3)Shíashkiibiʼiiss²á.
1SGboyb-ʼ-y-Ø-sh-¬-y²á
1SGboy3OBJ(bi).AGRc-INDF.OBJ-CAUS.P-M-1SG.SBJ-CLF-eat
'I am feeding the boy (lit., making make the boy eat).' (50a)
(4)Hastiinashkiiyiʼiis²á.
manboyy-ʼ-y-Ø-Ø-¬-y²á
manboy3OBJ(yi).AGRc-INDF.OBJ-CAUS.P-M-3SBJ-CLF-eat
'The man is feeding the boy (i.e., making the boy eat).' (50b)
(5)*Hastiinashkiibááh yiyiis²á.
manboybread yAGRc-yAGRo-y-ØM-ØAGRs-¬-y²á
'The man makes the boy eat bread.' (49b), Hale's star; (51b), with one agreement prefix left out, is also starred.

Hale p. 58: "the AGRc agreement morphology associated with the causative usurps the function of AGRo … In effect, only one agreement morpheme of the relevant sort can appear in E." (49)/(51) are his "hypothetical" forms with his stars. (50) is good because the indefinite ’- "does not 'compete' with AGRc for the standard object agreement position in E" (p. 59). The Leipzig line is ours, built from Hale's segmentation and labels. CAUS.P = his causative postposition -y, INDF.OBJ = his indefinite object ’-, M = Mode.

  • The causee prefix follows H (HYPOTHESIS on SOURCED data): y- under the 3rd-person subject 'man' (4), b- under 'I' (3). This is a postposition's agreement obeying the same subject-relative rule, even though it sits inside the verb word.
  • But not fn. 9's Ø (HYPOTHESIS on SOURCED data): fn. 9 says a 3rd-person object prefix in E under a 1st-person subject is Ø. In (3) AGRc is in E under 'I' and comes out b-, not Ø. It keeps the postposition's "otherwise b-". So the Ø/b- choice follows where the agreement came from (the P, valued inside the PP), not where it ends up. The y-/b- choice follows the subject (H). Hale does not comment on this. The reading is ours. [Correction 2026-10-27 ~4:46am PT: Hale & Platero 1996 (paragraph after their (10)) do state it. Causee markers are overt bi- under a 1st/2nd-person subject, "as if attached to an incorporated postposition". See causee-bi, which tests it on 10 forms (T64f).]
  • T63c, partly (HYPOTHESIS): no case of a forced shared value was found. (5) is a limit on number instead: one verb word holds one referential object agreement in E, and the causative P's agreement takes that one slot. Separate words (Willie (2): bá + yoolóós) keep their own values. The word, not the clause, is the domain of the limit. This fits RO-8 (one comparison per word) and gives it a second, independent kind of evidence.

8. Rulesets A and B HYPOTHESIS

  • B (clause, right-headed): everything H looks at is clausal. This covers Merge of the subject in VP, the object's raise to Spec,MP, the EPP raise to Spec,EP (subject or object), and where a PP attaches. Hale's heads are on the right (VP before M, MP before E: trees (5), (10), (11)). So this is right-headed structure, and c-command is computed here.
  • A (word, left-headed prefixal): the prefix itself. Head Movement takes V to M to E, and a P's agreement "intrudes" into E. This turns the right-hand clausal heads into a word whose prefixes stand to the left of the stem. Agree on AGRo (or on a P's AGR) reads the B-level fact "is my object below a 3rd-person subject?" and spells out y- or b-.
  • Where the changeover happens: at Head Movement into E. The relation is computed at B and stored in the word at A. Each word with its own agreement (V or P) makes the comparison once (RO-8). A P that intrudes into E keeps the result it already had (§7). This refines RO-9 (tier at B, spell-out at A) by naming the step that does the handover.

9. Rules

HG-1  b- = general 3rd-person object agreement; y- only if the 3rd object is in the c-command   SOURCED (Hale 1997 PDF p. 16)
      domain of a clausemate 3rd-person subject.
HG-2  Under 1st/2nd/4th subjects: Ø if the object morphology is in E, else b-.                    SOURCED (fn. 9, p. 16)
HG-3  Only a "true" 3rd person (pronoun bí) triggers y-; 4th persons do not.                     SOURCED (fn. 17, p. 27)
HG-4  Direct: object → Spec,MP (Agree with E); subject → Spec,EP to c-command it.               SOURCED (pp. 21–22, (10)–(11))
      Inverse: object → Spec,EP ("clausal topic", fn. 11); b- a result, not a cause (p. 18).
HG-5  "removing it" (p. 17) = topicalized OBJECT; Speas & Jelinek's suggestion, one of two.       SOURCED (correction to one-proximate)
HG-6  hale023 = the text Saxon cites as Hale 2000:70–71 (verbatim matches; fn. 11 vs 12).        HYPOTHESIS (strong)
HG-7  BR: on-tier rank = c-command height at valuation; then R5 ≡ HG-1.                          HYPOTHESIS
HG-8  HG-1 over overt nouns with "left = higher" is contradicted by Creamer (c23) vs Perkins (32);  HYPOTHESIS on SOURCED data
      HG-1 over pronominal slots (nouns linked at B) survives.
HG-9  One referential object agreement per verb word: a causative P's AGRc takes it.              SOURCED (Hale (49)–(51), p. 58)
HG-10 AGRc's y-/b- follows HG-1 ((50a)/(50b)); its b- under 'I' ignores HG-2's Ø                 HYPOTHESIS on SOURCED data
      → spell-out keyed to the P origin.
HG-11 Changeover B→A at Head Movement into E; one comparison per agreeing word.                 HYPOTHESIS (refines RO-8, RO-9)

10. Derivations, step by step

Labels: Σ role schema, B clause, A word. Restating Hale's trees: his trees are not flat-S trees. They already put the subject DP inside VP (tree (5), (10)), which is VPISH. What is older is the transitional notation (d-/s-structure, Move Alpha, government, "c-governor"). Below, MP and EP are kept as his labels for the two heads above VP. Movement is written as internal Merge, and agreement as Agree. Overt nouns are linked at B (nominal adjunction), so in the derivations the slots move, not the nouns.

HD-1. (2) = Willie (86) 'The girl is leading the horse for the boy'

ΣΣ = {ρA: 3sg human (girl); ρP: 3sg animate (horse); ρfor: 3sg human (boy)}. Proto-Agent / Proto-Patient as in Dowty; ρfor is a Navajo P-slot (an extension).
BMerge V 'lead' with ρP. Merge ρA in Spec,VP. Merge M, then E. (Merge builds the structure.)
BInternal Merge ρP → Spec,MP. Agree with E (AGRo). ρA → Spec,EP (EPP). ρA now c-commands ρP.
BPP [ρfor P] Agrees inside itself (p. 57 "locally"). The PP adjoins to EP as the clausal topic (fn. 11 "adjunct"), above ρA.
AWord 'for': is ρfor c-commanded by a 3rd-person subject? No → bá ✓. Word V: is ρP below ρA? Yes → yoolóós ✓.
BLink nouns: 'ashkii → ρfor, ʼatʼééd → ρA, łį́į́ʼ → ρP. Order boy > girl > horse ✓. The theme does not invert (Willie), since ρP never leaves Spec,MP.

HD-2. (4) = Hale (50b) 'The man is feeding the boy'

ΣΣ = {ρA: 3sg human (causer); ρcausee: 3sg human, object of CAUS.P; Proto-Patient of 'eat' = the indefinite ’-, non-referential}.
BMerge V2 'eat' with ’-. Merge PP [ρcausee -y]. Merge V1 ł-causative, then ρA, M, E (Hale (47)/(48)).
BAgree inside PP: ρcausee with P (AGRc). ’- raises to Spec,MP and incorporates (p. 59), so it takes no agreement slot. ρA → Spec,EP and c-commands the PP.
AHead Movement V2→V1→M→E. CAUS.P and AGRc intrude into E (the changeover into the word). AGRc: a 3rd-person subject is above → y- → yiʼiis²á ✓. With 'I' as subject (50a): not 3rd → b- (P origin, not fn. 9's Ø) ✓.
—Add a referential 'bread' (49b): it would also need AGRo in E, but AGRc already holds the one slot → * ✓ (HG-9).

HD-3. Saxon (32) = Perkins 1978:55 'The ball rolled over the log' (the HG-8 case)

ΣΣ = {ρS: 3, ball, inanimate; ρG: 3, log, object of 'over'}.
BMerge V 'roll' with ρS. ρS stays in VP (SX-4, a thing-subject). The PP ['over' ρG] adjoins above VP.
AWord 'over': is ρG below a 3rd-person subject? No (the subject slot is lower) → bá- ✓.
BLink nouns: Jooł 'ball' first, tsin 'log' second (topic order, not slot height) ✓.
—Creamer (c23) needs the same B steps with the subject slot raised or the PP low → yikʼi. Which one, and why the speakers differ, is OPEN (T62c, T64c).

11. Grant's ideas

GRANT'S CONJECTURE Feature bundles / proto-words as arguments. Consistent, with support from new data. The one contradicting pair in §6 is contradicting only if H holds between the spoken nouns. Stated between the role-bearing slots, it fits. That is the bundles view (HYPOTHESIS). Hale's indefinite ’- is "utterly non-referential", has "all the syntactic and semantic characteristics of an incorporated element", "essentially removes a referential DP object from the argument roster", and in (53)/(54) stands for "the 'prototypical patient' of the event denoted by the verb" (SOURCED, pp. 58–59). An argument position filled by a role-only element that is not a word is close to Grant's "proto-word". Hale frames ordinary arguments as DPs ("canonical DP arguments must enter into the agreement relation", p. 58). That is his framework, not a datum against bundles, so it is not counted.

GRANT'S CONJECTURE Strata. Consistent. Hale's derivation has the same layers in the same order. Argument structure comes first, in VP (roles). Agreement comes next (object to Spec,MP, Agree with E). Topic comes after that (EPP/Spec,EP, "clausal topic"). The prefix is "a result, not the cause" of the topic layer (p. 18), so it is read late. Noun choice is not needed until linking (HD-3).

GRANT'S CONJECTURE Merge/Agree sites inside the word. Supported, as a HYPOTHESIS. The causative gives a clear intra-word case. A postposition's Agree is completed inside its own phrase, the P is then merged into the verb word (Head Movement), and it keeps its own value and its own spell-out rule (§7). It also uses up the word's single object-agreement place (HG-9). So within one word there are two Agree sites, one of them carried in from a smaller word.

Chemistry: SPECULATIVE, flag only. A marker that depends on whether a site lies within the reach of a particular neighbour (a geometric domain), rather than on a global ranking, plus a host with only one place for such a marker. No chemical correlate is identified or claimed.

12. Bottom line

SOURCED Hale's workshop text (hale023, PDF pp. 16–18, 21–22, 27) states that b- is the "general" 3rd-person object marker and y- needs a 3rd-person object "in the c-command domain of a clausemate third person subject". It excludes 1st/2nd/4th-person subjects (fn. 9, fn. 17), and treats the inverse's b- as a result of the object's rise to "clausal topic". The "removing it" passage concerns the object and is credited to Speas and Jelinek. A causative P's agreement takes the verb word's only object-agreement slot ((49)–(51)).

HYPOTHESIS hale023 is the text Saxon cites as Hale 2000:70–71. T62e answered: Hale's rule is subject-relative and assigned object by object, like R4/R5. It differs in using c-command where they use rank, and the two coincide under BR. Applied to spoken-noun order, H is contradicted by Creamer (c23) vs Perkins (32); applied to pronominal slots, it survives. The B→A changeover is Head Movement into E.

OPEN The printed Hale 2000 pages (T64a). Whether thing-subjects can raise (the rainbow pair, T64b). PP attachment in (c23) vs (32) (T64c). The reordered (8c) test (T64d).

13. Fidelity notes

  • Hale quotations on PDF pp. 16, 17, 27 and 58 were checked against page images, and the others come from the text layer. Navajo forms from hale023 are printed as printed: ¬ and ² are font substitutions in the PDF itself, the same on the page image, and are not decoded. The morpheme breakdowns in (3)–(5) are Hale's. The Leipzig lines are ours.
  • The PDF page numbers are hale023's. Its printed folios and the printed Hale 2000 pagination were not checked (T64a).
  • Willie (85)/(86) were read at 300 dpi: the verb is printed yoolóós with a plain l. Several site pages (pp-by-verb, postposition-units, procliticized-p) print yoołóós. Not yet corrected there; flagged as T64e.
  • Saxon's footnote markers were read from the page image (PDF p. 363): Hale's remark is fn. 6. The rainbow note follows fn. 7 (Potter) and so is likely fn. 8 rather than the "fn. 9" used on earlier pages; the marker itself is faint (T64b).
  • Perkins (8a–c) stay schematic, as on two-osv. The noun-order column in §6 is Saxon's printed order.
  • H, BR, the slot reading, and all slot positions in HD-1 to HD-3 are this site's, not Hale's. Hale did not discuss (32), (c23) or Willie (85)/(86).

14. Open questions

IDQuestionWhat would settle it
T64aIs the printed Hale 2000 (MITWPL ELL 3: 55–97) pp. 70–71 the same text as hale023 PDF pp. 16–18? Which footnote is added (11 → 12)?The printed volume first-hand. OPEN
T64bCan a thing-subject raise to Spec,EP and so license yi-? Saxon's rainbow pair (Y&M 1987:225, yi- and bi- both printed) is the H-shaped test (= T61c).Young & Morgan 1987 p. 225 first-hand (page image), plus any context that separates the two. OPEN
T64c(c23) yikʼi vs (32) báhátis under HG-8: is the difference the subject's height or the PP's attachment? Does the 'over'/'upon' phrase behave like a topic (e.g. in relatives, as in Platero's p. 18 example)?Perkins 1978 pp. 55, 136 first-hand; Creamer 1974 group 7 context. OPEN
T64dOrder-only vs slots: a version of Perkins (8c) with 'bike' before 'girl' (bił + b-verb). The order reading predicts it is good, and the slots reading with SX-4 predicts it is bad.Perkins 1978 p. 116 first-hand, or a native-speaker judgment. OPEN
T64eWillie (85)/(86) verb: yoolóós (page image) vs yoołóós (printed on pp-by-verb, postposition-units, procliticized-p). Correct site-wide after one more check.Willie pp. 95–96 and (29) page images; a dictionary entry for 'lead (animate)'. OPEN
T64fHG-10: is a causative or other incorporated P's agreement under a 1st/2nd-person subject always b- and never Ø? Is the Ø of fn. 9 limited to E's own AGRo?Young & Morgan causative paradigms; Hale 1997 pp. 54–59 against other P-incorporations. OPEN [Update 2026-10-27: answered for causatives on causee-bi (10/10 b-, 0 Ø; see T65a–e).]

Sources: Hale 1997 workshop notes (hale023) PDF pp. 14–15 (via inverse-agree), 16–18, 21–22, 27, 54–59 (page images pp. 16, 17, 27, 58); Willie 1991 p. 95 (page image, (85)/(86)); Saxon 2003 notes page (image) and pp. 343, 352, 354, 356. Carried over: Aissen 2000 (13a), (16); Creamer 1974 (c23); Perkins (8a–c), (32) via Saxon. See also ranked-obviatives, one-proximate, two-osv, inverse-agree, open questions, sources.

Fourth person: escape from the animacy scale

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.

Working note. Kept as written (URL, data and status labels unchanged). Filed under The fourth person; the consolidated argument is the essay fourth-person. Rule IDs: rules index.

Verbs on this page → dictionary: yiztał ‘kick’ · yishxash ‘bite’ · hashshish ‘sting’ · dabiiłtsą́ ‘see’ · yítʼiid ‘do’ · yidzaa ‘happen to’ · hóótʼiid ‘happen’ · ní ‘say’

Restated on the unified structure (2026-11-15). The fourth person is restated as a person value on an argument merged in Spec,vP or VP, with nothing extra moving: the optative fourth bound by the wisher (H51), the respect/report/story ji- as an Agree mark (H57a, OPEN against H57b), the areal ho- as a personless area argument (H54a, H60). 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. 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, §4, fourth-person rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

When a lower-ranked agent acts on a human patient, Navajo has two responses among third persons: flip to the inverse (bi-), or re-categorize the human as fourth person and keep the direct form with the agent as subject. This page states that choice as rules, compares it with Grant's ranking-edge shape, and derives the sourced pairs step by step. The fourth person is the next open item after the inverse page (T5). It is open question T8.

On this page: 1 Question · 2 What fourth person is · 3 Data · 4 Rules · 5 Vs Grant's edge · 6 Where the choice is made · 7 Derivations F1–F5 · 8 Verdict · 9 Open

1. The question

GRANT'S CONJECTURE The project is argument ranking. The natural alignment is that the higher proto-agent argument is the subject. The edge case is when a lower argument outranks it. Then a marked expression is preferred, and some permutations are impossible. Open question after T5: is the fourth person (ji- / ha- / ho-) another ranking-driven inverse, or something else?

HYPOTHESIS (preview; argued below). It is not a second inverse. It is the scale's escape hatch: fourth person, like first and second, is off the animacy hierarchy. Taking the human patient (or agent) out of the ranked set lets the default (direct) alignment survive a clash that would otherwise force bi-. Roles stay fixed (C2 / INV-1). Only the person category of one slot changes.

2. What the sources say the fourth person is SOURCED

  • Voice vs obviation. Willie 1991, abstract and Ch. 9: the yi-/bi- alternation is "a voice contrast confined to sentences with all third person arguments"; the "fourth" person is "used for obviation and for deference." Concluding remarks: "the yi-/bi- alternation is a voice contrast, while the fourth person marks obviation. First, second, and fourth person have properties in common as opposed to third person."
  • Forms (Willie §4.1). Fourth-person subject ji-, fourth-person object ho-, kept distinct from a homophonous pleonastic/"areal" subject ho- that she sets aside (now read in full: pleonastic & psych). Printed object examples in §4.2.7 begin ha- (haztał, hashxash); the two-fourth form is hojishxash. The ha-/ho- shapes are Young & Morgan's phonology (not opened); this page does not derive them.
  • W&J 2000:277. "The fourth person in Navajo, like first and second, cannot enter into the direct/inverse alternation, and is employed in a variety of discourse contexts. It is often used as an impersonal, and does not ordinarily cooccur with an NP, but may do so in certain construction types that resolve ambiguities or force coreference." p. 278: "The fourth-person pronouns, both subject and object, are exclusively definite and refer primarily to humans, as with the bi-pronoun." Note 4 sends the reader to Willie 1991.
  • W&J 2000:264. The voice alternation "is excluded where there is a first-, second-, or 'fourth'-person argument" (schema C4).

Update 2026-10-10 (evening): the discourse uses W&J mention (respect, reportative, story, impersonal, optative) are now page-checked from Willie §4.2.1–4.2.6 on the fourth person uses page, together with the ban on attaching nouns to a fourth person and its three exceptions.

3. Data SOURCED

Sources opened this run (2026-10-08 evening). (a) Willie, MaryAnn 1991, Navajo Pronouns and Obviation, PhD dissertation, University of Arizona (UMI 9210298 scan already on the box). Page images of printed pp. 122–130 (= PDF pp. 128–136) re-read; hi-res crops of pp. 122, 125, 127–130. Navajo strings checked against the images; the repository OCR text layer is garbled for Navajo and was used only to locate sections. (b) Willie & Jelinek 2000, "Navajo as a Discourse Configurational Language," in Fernald & Platero (eds.), The Athabaskan Languages (Oxford), via Carnie & Harley (eds.) 2012 Jelinek compilation: printed pp. 275–278 at 200–300 dpi (PDF pp. 155–158). (c) Cross-check against the 2026-10-05 research note; every form below was re-read on the images this run.

Glosses are Willie's (3.3.bit, INV3.3.bit, 4.3.bit, 3.4.bit, 4-3-sting) or W&J's (4POSS, PL-4ACC-3NOM-saw), as printed. Where Willie's typescript prints the same kick form with and without a bar on the l ((28b) on p. 127 vs p. 130), the barred form from the clearer page is given, matching the inverse page's practice for yiztał/biztał.

3.1 All-third baseline: inverse required (Willie 20)

Willie 1991:122, (20)
a.ʼashkiiłééchąąʼíyishxash
boydog3.3.bit
‘The boy bit the dog.’
b.*łééchąąʼíʼashkiiyishxash
(The dog bit the boy.)
c.ʼashkiiłééchąąʼíbishxash
INV3.3.bit
‘The boy was bitten by the dog.’

p. 122: "when a lower-ranked agent acts upon a higher-ranked patient, and all arguments are third person, then the Inverse voice must be used." Same page: "Fourth person, like first and second, is definite, human, and not involved in the animacy hierarchy." Spelling note: earlier site pages print the dog as łééʼchąąʼí (with an apostrophe after éé); the hi-res image of p. 122 this run is read as łééchąąʼí. Both are recorded; neither is repaired (T8f).

3.2 Fourth-patient rescue: keep direct (Willie 21–22)

Willie 1991:123, (21)–(22)
(21)łééchąąʼíshishxash‘The dog bit me.’
dog1.3.bit
(22)łééchąąʼíhashxash‘The dog bit that person.’
dog4.3.bit

p. 123: "With first, second, and fourth person patients, it is possible to have a lower-ranked agent (third person) acting upon a human patient." "When the agent is third person and the patient is human, the speaker can use a fourth person patient pronoun and avoid having to use an Inverse form." And: "In (21–22), the lower ranked third person agent is the subject, the external Nominative argument, and not an internal (Ergative) agent, as we see in the Inverse."

3.3 Bare fourth forms; deictic opener (Willie 23–25)

Willie 1991:124–125, (23)–(25)
(23a)hashxash4.3.bit‘It bit that person.’
(23b)jishxash3.4.bit‘That person bit it.’
(24)Yíh,haztał‘Oops, It kicked him [that person].’
4.3.kicked
(25)*Yíh,biztał‘It kicked him [third person].’

pp. 124–125: third-person pronouns without a nominal need a discourse antecedent; the fourth "is simply deictic." Rodeo opener: (24) is acceptable as an initial remark; (25) is not. Also p. 125: with a 1st, 2nd or 4th argument present, "There is no question about the theta role of the third person argument, since the theta role of the 1, 2 or 4 argument is clearly marked in the pronoun incorporated into the verb."

3.4 Indefinite low agent allowed under fourth-direct (Willie 26)

Willie 1991:126, (26)
séígoʼhashshish
scorpion4-3-sting
‘A scorpion stung her.’

p. 126: Inverse external arguments must be definite; "This limitation does not apply to any Direct construction, including fourth person constructions. Therefore, it is possible to have an indefinite, lower ranked agent acting upon a human patient in a Direct construction by using the fourth person." Compare INV-4 on the inverse page.

3.5 Question–answer pairs track which participant is presupposed (Willie 27–31)

Willie 1991:127–130, (27)–(31)
Question (as printed)PresupposedReply (as printed)Form
(27a) haash yítʼiid ‘What did he do?’human agent(27b) mósí yiztał ‘He kicked the cat.’direct
(28a) haash yítʼiid ‘What did it do?’non-human agent(28b) haztał ‘It kicked him [that person].’fourth (patient)
(29a) haash yidzaa ‘What happened to him?’human patient(29b) mósí bishxash ‘He was bitten by the cat.’inverse
(30a) haash yidzaa ‘What happened to it?’non-human patient(30b) jiztał ‘That person kicked it.’fourth (agent)
(31a) haash hóótʼiid ‘What happened?’neither (event focus)(31b) ʼashkii mósí yishxash; (31c) ʼashkii mósí bishxash; (31d) mósí hashxash; (31e) mosi jishxash; (31f) hojishxashtwo nominals, two fourths, or a mix

Translations as Willie prints them: (31b) ‘The boy bit the cat.’; (31c) ‘The boy was bitten by the cat.’; (31d) ‘The cat bit that person.’; (31e) ‘That person bit the cat.’; (31f) ‘That person bit that person.’ As printed: (31e) has mosi without the accent that (27b)/(31b–d) have on mósí. On (31), p. 129: when neither referent is presupposed, "the reply must include either two nominals, two fourth person pronouns, or a combination of the two." On third-person alternatives to (28b)/(30b) (ʼashkii yishxash ‘The boy bit it’; ʼashkii yiztał ‘The boy kicked it’): they "show incorrect word order, according to the principles of the voice alternation and the animacy hierarchy," yet are "permitted in context"; "Use of the fourth person forms makes it possible to avoid this problem."

(28b) cannot be inverse (p. 130): haztał cannot be read as (32) ‘It was kicked by him.’

3.6 Coreference rescue (W&J 70, 77)

Willie & Jelinek 2000:276–278, (70), (71), (77)
(70)tʼááʼałtsoJáanba-ʼáłchínídabiiłtsą́
justallJohn3POSS-childrenPL-3ACC-3NOM-saw
a. ‘[All John’si children]j saw himk.’   b. * ‘[All John’si children]j saw himi.’
(77)Jáantʼááʼałtsoníha-ʼáłchínídahoołtsą́
Johnjustall4POSS-childrenPL-4ACC-3NOM-saw
‘Johni, just all that person’si childrenj saw that personi.’ (All John’si childrenj saw himi.)

Statement (71), p. 276: coreference between possessive arguments internal to a complex NP and the pronominal arguments of a following verb-sentence is excluded ("impossible" for most speakers, "strongly dispreferred" for others). The reading that (70b) stars is expressible once the possessor and object are fourth person (77). W&J also discuss a photograph/please pair (75)–(76) on p. 277 where "The deictic 'fourth' person can be used to force coreference"; the exact Navajo strings of (75)–(76) are hard to read reliably on this scan (prior note and this run's image readings disagree), so they are not reprinted (T8c). The prose claim on p. 277 is used.

4. The rules

Notation as on the schema page and the inverse page. Σ = {ρA, ρX}. r(ρ) is animacy rank. EXT(Σ) is the external argument (subject). R(Σ) = {ρ ∈ Σ : π(ρ) = 3} (C4). Fourth person is written π(ρ) = 4.

EXIT     If π(ρ) ∈ {1, 2, 4}, then ρ ∉ R.                                    [= C9 for 4]
         Fourth (like 1st/2nd) is definite, human, and off the animacy scale.

KEEP-DIR When r(ρ_A) < r(human patient) would force EDGE / bi- under INV-0,
         valuing the patient as 4 bleeds INV-0 (R shrinks), so VOICE is undefined
         and NAT survives: EXT = ρ_A, spelled out as direct (no bi-).

4-DEF    π(ρ) = 4 → the referent of ρ is definite (and primarily human).
         (W&J 2000:278; Willie: "definite, human")

4-INDEF-A  Under KEEP-DIR (direct + 4th patient), the external agent may be
         indefinite. Inverse EXT must be definite (INV-4); that ban does not apply.

4-NO-INV Fourth cannot enter the yi-/bi- alternation; a 4th-patient direct form
         cannot be read as inverse ((28b) ≠ (32)).

4-DEIX   Fourth needs no discourse antecedent (deictic); bare 3rd does.
         Hence (24) ✓ vs (25) * as a conversational opener.

4-COREF  Fourth (possessor and/or object) can force a coreference that all-3rd
         inverse bans ((70b) * rescued by (77)).

4-LINK   With a 1st/2nd/4th argument present, a lone nominal links to the 3rd-person
         slot (external or internal). With all third persons, a lone nominal links
         only to the internal slot (Willie p. 126; INV-6 for the all-3rd case).
RuleStatusBasis
EXITSOURCED (restated as C9)Willie 1991:122; W&J 2000:264; schema C4/C9.
KEEP-DIRSOURCED as Willie's description; the bleed formalization is HYPOTHESISWillie pp. 122–123, (21)–(22); "instead of Inverse voice forms"; "avoid having to use an Inverse form."
4-DEFSOURCEDWillie p. 122; W&J 2000:278.
4-INDEF-ASOURCEDWillie p. 126, (26).
4-NO-INVSOURCEDW&J 2000:277; Willie p. 130, (28b) ≠ (32).
4-DEIXSOURCEDWillie pp. 124–125, (24)–(25).
4-COREFSOURCEDW&J 2000:276–278, (70), (71), (77).
4-LINKSOURCEDWillie pp. 125–126.

The cleanest restatement HYPOTHESIS. Navajo has two responses to the same ranking clash (low agent, high human patient): (i) flip alignment (EDGE / bi-, inverse page); (ii) take the high argument off the scale (EXIT / fourth) and keep NAT. Inverse changes EXT. Fourth changes π of one slot and leaves EXT alone. That is why fourth is a boundary condition on Grant's edge, not another instance of it.

5. Against Grant's ranking-edge shape

Edge property (Grant, 2026-10-04)Navajo fourth person, as sourcedVerdict
Fixed roles under the marked expression(22), (26), (28b), (30b): agent and patient unchanged; only the person category of one argument changesMatch
A ranking scale independent of the rolesSame animacy hierarchy as the inverse — but fourth is outside it ("not involved in the animacy hierarchy")Different relation to the scale. Fourth does not rank higher; it exits the comparison set
Marked form preferred when the lower argument outranksWith a human patient and a lower-ranked agent, fourth is an alternative to the inverse, and it keeps the default alignment (agent = external subject, direct)Not the same move. Inverse flips; fourth leaves alignment alone and removes the clash
Some permutations impossible(20b) *; (25) * as an opener vs (24) good; fourth cannot be inverse ((28b) ≠ (32)); fourth cannot enter the alternation; (70b) * rescued by (77)Match (categorical bans present)
No demotionBoth stay pronominal arguments; Willie: the lower-ranked agent is "the external Nominative argument" in (21)–(22)Match
Obviation as the ranking device (Algonquian-style expectation)Willie splits the jobs: voice (yi-/bi-) is the ranking-sensitive device; fourth is the obviative person categoryFourth ≠ a second inverse. Added 2026-10-09: the cross-linguistic check (Bruening 2005 on Passamaquoddy; Lybech 2020 on West Greenlandic) is on the fourth person compared page. Algonquian obviation corresponds to yi-/bi-, and the Greenlandic “fourth person” is a bound third person, the opposite of Navajo's on reference.

HYPOTHESIS Fine-event grain. Willie's Q–A pairs (27)–(31) show the choice among direct, inverse and fourth tracking which participant is already established as the event unfolds (human agent presupposed → direct; human patient presupposed → inverse; non-human participant presupposed with a human co-participant → fourth; event-only → two nominals, two fourths, or a mix). Roles are the same across all replies (C2). This is at the event-structure grain Grant asked for; it is not a Dowty proto-role effect.

6. Where in the derivation the choice is made

6.1 π-valuation before voice HYPOTHESIS (from SOURCED facts)

  1. Order of operations. If VOICE were computed before ρP's π-value was fixed, a human patient with a dog agent would force bi-, as in (20c). Fourth-person valuation bleeds the inverse (D5; KEEP-DIR). So Σ-valuation of π precedes the step that fixes voice. Same as schema §5 and D5.
  2. Inside the word. Bare (23a) hashxash / (23b) jishxash already encode agent vs patient fourth with no nominals. The person choice, like voice (inverse §5.1), is fixed at Σ and spelled out by rule set A. Rule set B only links the optional 3rd-person nominal (4-LINK).
  3. Two paths, one clash. Low agent + human patient, all else equal: path INV applies EDGE (EXT = patient, bi-); path 4 applies EXIT on the patient (EXT = agent, direct). Both leave Eρ untouched (INV-1 / C2).

6.2 What VPISH predicts

  • HYPOTHESIS Under V-P (VPISH raising) with KEEP-DIR: ρA raises Spec-to-Spec to Spec,M/SP as in the ordinary direct voice. Nothing crosses. That is Willie's "external Nominative" agent in (21)–(22). Contrast the inverse (I1), where ρX raises past ρA.
  • HYPOTHESIS Under V-H (Hale 2003 in-situ licensing): no movement; EXT is Nominative on the agent pronoun. Same string as V-P.
  • HYPOTHESIS + GRANT'S CONJECTURE What changes under fourth is a role-slot's person value, not a lexical item. Nominals are adjuncts (4-LINK; W&J: fourth "does not ordinarily cooccur with an NP").
  • OPEN Whether V-P and V-H diverge for fourth-person binding/scope into a co-argument (T8a). The inverse's sharper test remains T5c.

7. Derivations F1–F5, step by step

Conventions as on the derivations page and I1–I5. Tags: Σ schema, A word-internal, B clause level. V-H = in-situ; V-P = VPISH raising. Verb words are not segmented beyond the sources' glosses. Entailment assignments are ours, from the English translations (HYPOTHESIS): for ‘bite’/‘kick’/‘sting’, EρA ⊇ {vol, caus, mov} and EρP ⊇ {aff}, matching D3/D5.

F1. Baseline clash → inverse: Willie (20c) ʼashkii łééchąąʼí bishxash ‘The boy was bitten by the dog.’

  1. Σ Σ(√BITE) = {ρA, ρP}. posρA = Spec,v; posρP = Comp,V. C5: ρA is the agent slot. Entailments fixed (INV-1 / C2).
  2. Merge as in I1 (V, v, ρA, M/S). A.
  3. Σ π(ρA) = π(ρP) = 3. INV-0 holds. Rank: boy > dog, so r(ρP) > r(ρA) if the dog is agent and the boy is patient. INV-3 / EDGE: EXT = ρP; INV-2: VOICE = bi-.
  4. Agree M/S ⟶ EXT. Move V-P: ρP to Spec,M/SP, crossing ⟨ρA⟩ (internal Ergative agent). V-H: no movement.
  5. Spell-out A bishxash INV3.3.bit.
  6. B INV-6: łééchąąʼí (next to V) = internal = ρA; ʼashkii = EXT = ρP. INV-4: boy definite ✓.

Contrast: (20b) * tries NAT with the same referents (EXT = dog = ρA) and crashes INV-3.

F2. Fourth-patient rescue: Willie (22) łééchąąʼí hashxash ‘The dog bit that person.’

  1. Σ Same Σ and entailments as F1 (C2). Referents: ρA = dog, ρP = human.
  2. Merge as in F1, steps 1–2.
  3. Σ π-valuation first. π(ρP) = 4; π(ρA) = 3. EXIT / C9: ρP ∉ R. R = {ρA}, |R| = 1, so INV-0 fails and VOICE is undefined (C4). KEEP-DIR: NAT survives — EXT = ρA.
  4. Agree M/S ⟶ EXT = ρA (Willie's "external Nominative"). Move V-P: ρA Spec-to-Spec to Spec,M/SP (ordinary direct). V-H: no movement. Nothing crosses; contrast F1 / I1.
  5. Spell-out A hashxash 4.3.bit (unsegmented; Willie names 4-object ho-; ha- shape not derived here).
  6. B 4-LINK: łééchąąʼí links to the sole 3rd-person slot ρA. No DP for ρP (W&J: fourth "does not ordinarily cooccur with an NP").

Result. Same clash as F1; different path. Roles unchanged. Agent stays external. Fourth bleeds the inverse. Same bleed for (21) with π(ρP) = 1.

F3. Indefinite low agent under fourth-direct: Willie (26) séígoʼ hashshish ‘A scorpion stung her.’

  1. Σ Σ(√STING) = {ρA, ρP}. π(ρP) = 4, π(ρA) = 3. EXIT → VOICE undefined → EXT = ρA (KEEP-DIR), as in F2.
  2. Spell-out A hashshish 4-3-sting.
  3. B séígoʼ links to ρA. 4-INDEF-A: the external agent may be indefinite ✓. Under the inverse path, INV-4 would ban an indefinite EXT; that ban does not apply here because the construction is Direct.

F4. Fourth agent; two fourths: Willie (30b), (31f)

  1. Σ (30b) jiztał — π(ρA) = 4, π(ρP) = 3. EXIT removes ρA from R. VOICE undefined. EXT = ρA (NAT; the fourth agent is external). Spell-out A jiztał (Willie: subject ji-). Q–A: replies to (30a) ‘What happened to it?’ with a human agent.
  2. Σ (31f) hojishxash — π(ρA) = π(ρP) = 4. R = ∅. VOICE undefined. Both slots off-scale. Spell-out A hojishxash. Used when (31a) presupposes neither participant.
  3. Σ 4-NO-INV: (28b) haztał cannot be read as (32) ‘It was kicked by him.’ Fourth-patient direct is not inverse.

F5. Coreference rescue: W&J (70b) * → (77)

  1. Σ (70) is an inverse verb-sentence (W&J tree: Topick–Focusj–V INVERSE) with all third-person arguments. Statement (71) bans coreference between a possessor inside the complex NP and a pronominal argument of the following verb. Reading (70b) *.
  2. Σ (77): possessor and object valued π = 4 (ha-ʼáłchíní, dahoołtsą́ PL-4ACC-3NOM-saw). 4-COREF / 4-NO-INV: fourth is outside the voice alternation; the banned all-3rd coreference becomes available. Translation: (All John’si children)j saw himi.
  3. B Jáan may cooccur with fourth here (W&J: fourth "may do so in certain construction types that … force coreference"). Ordinary fourth "does not ordinarily cooccur with an NP" (p. 277).

What F1–F5 show HYPOTHESIS

  1. Roles never change. F1 and F2 share Σ entailments. Only π and EXT differ.
  2. π before voice. F2's EXIT step must precede INV-0, or the inverse would be forced. Same order as D5.
  3. Fourth keeps NAT; inverse flips EXT. Two responses to one clash. Fourth is not a second EDGE.
  4. Bans that inverse carries (definiteness of EXT; all-3rd coreference (71)) need not apply on the fourth-direct path (F3, F5).
  5. Under V-P, fourth-direct is ordinary agent raising; inverse is patient raising past the agent. Same string under V-H. The sharper VPISH test stays on the inverse (T5c).

8. Verdict on T8

  • Escape hatch, not a second inverse. SOURCED Willie 1991 §4.2.7: fourth is "not involved in the animacy hierarchy"; the speaker can use it "instead of Inverse voice forms" and keep the agent as "the external Nominative argument." W&J: fourth "cannot enter into the direct/inverse alternation."
  • Grant's edge. HYPOTHESIS Fixed roles, impossible cells, and no demotion match. The relation to the scale does not: fourth exits the ranked set rather than flipping alignment when the lower argument outranks. So fourth is a boundary condition on the edge, not another instance of it.
  • Where. HYPOTHESIS π-valuation at Σ bleeds voice (KEEP-DIR); spell-out is rule set A; rule set B only links. Same architecture as the inverse page's "choice inside the word," with a different lever (person vs EXT).
  • Event grain. SOURCED Q–A (27)–(31) tie the choice to which participant is presupposed. Roles stay fixed.

9. Open questions this page adds

#QuestionWhat would settle it
T8aUnder V-P vs V-H, does a fourth-person argument show any binding/scope asymmetry with its co-argument that the inverse's patient/agent pair (T5c) does not?Elicitation; Willie's quantifier pairs are aimed at inverse first. OPEN
T8bWillie 1991 Chs. 5–6: fourth person marking disjoint reference across adjoined and relative clauses; how it interacts with W&J's statement (38) (obligatory coreference across all-third-person transitive clauses).Partly answered 2026-10-09 (evening): §5.1 (pp. 132–140) is page-checked on the =go chains page. Across transitive =go pairs the cast is fixed (GO-PP). π = 4 introduces a new agent only from the first clause, and the patient must stay (GO-4-AGT, GO-PAT). §5.2–5.4 and Ch. 6 are still unread. OPEN 2026-10-10: §5.2–5.4 read on the reported speech page: under ‘say’, all-3rd = different subjects, 4/4 = same; respect vs quotative jiní; person focus. 2026-10-10 (afternoon): Ch. 6 read on the relatives page: in relatives one 4 marks a different person and 4/4 the same, nouns and 4 both block coreference, and a 4 cannot head a restrictive relative.
T8cExact Navajo strings of W&J (75)–(76) (photograph/please coreference). Prior note and this run's scan readings disagree; not reprinted.Cleaner copy of Fernald & Platero 2000 pp. 277–278, or Young & Morgan for the verb. OPEN
T8dAkmajian & Anderson 1970 and Young & Morgan 1987 for the fourth-person paradigm itself (subject ji-, object ho- as Willie names them), and the ha-/ho- phonology.Open those sources. OPEN
T8eGrant's multiple-wh / casino test: a clause with a fourth-person argument has no yi-/bi- ranking to read (4-NO-INV). Does that change fronting or -sh placement vs D7 / T5e?Needs Grant's texts and native intuitions. Do not pre-fill. OPEN
T8fDog spelling: łééchąąʼí (this run's p. 122 image) vs łééʼchąąʼí (earlier site pages for the same examples). Also mosi vs mósí within Willie (31).Young & Morgan; speaker check. OPEN
T8gGradience: Young (1977; 1991) found "only about half of the speakers obeying the animacy hierarchy" (via Willie Ch. 1). Does fourth-person use correlate with hierarchy obedience?Speaker survey; Young not opened. OPEN

One fourth person, five uses: respect, reportative, story, impersonal, wish, and when it may carry a noun

2026-11-11 audit pass 6. §12 (derivations P1–P6) is now labelled HYPOTHESIS. Its steps, including “the A → B changeover”, are our analysis, not sourced facts. See the pass 6 report.

Working note. Kept as written (URL, data and status labels unchanged). Filed under The fourth person; the consolidated argument is the essay fourth-person. Rule IDs: rules index.

Verbs on this page → dictionary: yiztał ‘kick’ · naa-ghá ‘walk around (one)’ · yi-sh-cha ‘cry’ · doogááł ‘arrive, will go / come (future)’ · ní ‘say’ · joolyé ‘be called (named)’ · nábiʼdiiʼnil ‘pick up (plural objects); translated as ‘were picked up’’ · yiʼash ‘walk (two)’ · yiildlosh ‘walking (as glossed; ‘there is a coyote around’, ‘walking along’)’ · ʼajoołháásh ‘sleep (optative form only printed)’ · nizin ‘think; wish (with an optative clause)’

Restated on the unified structure (2026-11-15). The fourth person is restated as a person value on an argument merged in Spec,vP or VP, with nothing extra moving: the optative fourth bound by the wisher (H51), the respect/report/story ji- as an Agree mark (H57a, OPEN against H57b), the areal ho- as a personless area argument (H54a, H60). 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. 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, §4, fourth-person 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 rest of Willie's (1991) §4.2, the sections on what the fourth person is used for: §4.2.1 the ban on nouns (pp. 109–111), §4.2.2 the polite fourth (pp. 111–113), the end of §4.2.3 on respect vs reportative (pp. 114–116), §4.2.4 narratives (pp. 116–117), §4.2.5 the Impersonal (pp. 117–119) and §4.2.6 the Optative (pp. 120–122). Read with this site's rules, the pages give four results. (1) Normally a noun cannot be attached to a fourth person, because nouns count as third person (5c), (6f). Willie names exactly three contexts where it can: respect for one's mother (8), the reportative (11), which also needs the indefinite léiʼ, and the opening of a traditional story (14). (2) Willie treats respect, reportative, story, impersonal and optative as uses of one fourth person, whose shared job is to avoid naming or pointing at anyone (pp. 113, 121–122). That is her answer to T13a. (3) The optative fourth breaks a rule this site drew from Chapters 5–6. Under ‘think’ (nizin), a fourth-person optative subject is the third-person thinker: ‘My father wishes he could buy the horse’ (18). So “a fourth person never picks up a third person” (PI-MATCH i) is not general. (4) In that optative, what fixes the referent is a role link (the one who wishes is the one who would buy), not a noun and not a matching person value. This is new support for Grant's “role slots, not lexical items”.

On this page: 1 Question · 2 Source · 3 The noun ban · 4 Polite fourth · 5 Respect vs reportative · 6 Stories · 7 Impersonal · 8 Optative · 9 Rules · 10 T13a · 11 PI-MATCH revised · 12 Derivations P1–P6 · 13 Open

1. The question

The reported speech page kept two jobs of jiní ‘4.said’ apart, respect and quotative, and left T13a open: one morpheme with two uses, or two morphemes with one form? Willie lists further uses (narrative, Impersonal, Optative) that had not been read. Across the =go, ‘say’ and relative pages this site also built PI-MATCH: a fourth-person slot never takes a third person's referent, and two fourth persons take the same one. For Grant's two phenomena this page asks:

  • Theta roles as arguments (phenomenon 2). When the person value of a slot is 4, what decides who fills it: the discourse use, a noun, or a link to another role slot?
  • Rule set A vs rule set B (phenomenon 1). A noun attaches at clause level (B), to a slot built inside the word (A). When is that attachment blocked, and is the block decided inside the word or outside it?

2. Source opened this run (2026-10-10, ~8:50pm PT)

  • Willie, MaryAnn 1991. Navajo Pronouns and Obviation. PhD dissertation, University of Arizona (the scan on the box). Fresh page images of printed pp. 109–122 (PDF 115–128; PDF = printed + 6) at 150 dpi. Every Navajo line of (5)–(8) and (11)–(19) checked on 300 dpi crops ((9)–(10) were checked in the morning run). No OCR used for any form.

Transcription and glosses. Navajo as printed, with the glottal stop as ’ and the nasal hook as ą/į. Willie's gloss lines are copied exactly, misprints included and marked [sic]. Her gloss order for transitive verbs is object.subject (confirmed on the reported speech page). (5)–(6) have translations only, no gloss line, and we add none. Where a gloss line has a blank under a word, the blank is kept.

Cited by Willie in these sections, not opened: Uyechi 1990 (claims nouns can link to fourth persons); Young & Morgan 1987 (fourth person as “a deictic pronominal”); Witherspoon 1977:84 (the impersonal and Navajo ideas of control).

3. The noun ban (§4.2.1, pp. 109–111) SOURCED

P. 109: “Third and fourth person pronouns do not match in phi (person) features, so coreference is blocked. Furthermore, nouns are syntactically third person, so a nominal cannot be adjoined to a fourth person pronoun. In semantic terms, the referent of the fourth person, like first and second, is so specific that it does not leave reference open to any person previously mentioned in the discourse, as the third person normally does.”

Willie 1991:109–111, (5)–(6)
(5a)*ʼatʼééd,yicha [sic?]
[I, as a girl, I cried]
(5b)*ʼatʼééd,yínícha
[You, as a girl, you cried].
(5c)*ʼatʼééd,jíícha.
[That person, as a girl, cried]
(6a)Shíyícha. [sic?]
‘I, I cried.’
(6b)Níyínícha
‘You, you cried.’
(6c)Bí,yícha.
‘Him, he cried.’
(6d)ʼashkiiyícha
‘The boy, he cried.’
(6e)Hó,jíícha
‘That certain person, he cried.’
(6f)*ʼashkiijíícha.

P. 110: “It is possible to adjoin an independent free pronoun that agrees in person with the pronominal argument in the verb.” P. 111: “In the next section, I will identify certain special contexts that are exceptions to this generalization.” On Uyechi 1990 (p. 110): “This does not occur in my speech, or other speakers I have consulted. These speakers explain that where there are both fourth person forms and nominals, there are now too many participants!”

[sic?] in (5a) and (6a): the verb is printed yicha / yícha under a first-person translation. That is the spelling of the third-person forms in (6c, d), and the first-person form on the dictionary entry for ‘cry’ (from the NLA) is yi-sh-cha. We copy Willie as printed and do not correct it (T15g). Tone also varies as printed: yicha (5a) vs yícha (6).

4. The polite fourth (§4.2.2, pp. 111–113) SOURCED

P. 111: “Two basic tenets of politeness in discourse in Navajo society are: first, never point to anyone in his presence; second, never call anyone by his name in his presence. The device that Navajo speakers employ to single out a particular third person referent in the discourse context is a discourse function of the fourth person.” Then: “In the presence of a third person referent, ji-/ho- are used to identify the subject and object respectively.”

Willie 1991:112–113, (7)–(8)
(7a)Jíícha
4. cried
‘That certain person cried.’
(7b)Johnjiztał
3.4.kicked
‘She kicked him, John.’
(7c)Johnhaztał
4.3.kicked
‘John kicked that certain person.’
(8)shimáʼéíBaaʼjoolyé
1.mother4.called
‘My mother is called Baa’.’

(7) is introduced with “If the referent is in hearing distance, then:” (p. 112). In (7b, c) John links to the 3 slot in both. In (7b) that slot is the object, in (7c) the subject, and the 4 is the other participant. On (8), p. 113: “In certain discourse contexts, where the speaker wants to express respect, it is possible to adjoin a nominal to a fourth person pronoun. One frequently hears this when the speaker is referring to his mother. … This usage does not seem to occur with other kin terms.” And p. 113: “In my experience, fourth person is a strategy to avoid using the name of person or to point at him. In this sense, fourth person is a deictic pronominal as stated in Young & Morgan (1987).” Willie adds (p. 112) that the older mother-in-law / son-in-law use of the fourth person as an address form is one “I do not find … in existence today.”

5. Respect vs reportative (end of §4.2.3, pp. 114–116) SOURCED

The quotative pair (9)–(10) is on the reported speech page. Willie then contrasts two sentences with nouns:

Willie 1991:115, (11)–(12)
(11)ʼashkiiléiʼʼatʼéédbiłjooʼash=go
boyagirls3.with [sic]4s.walking=when
náhodiiʼnilłájiní.
4s.picked up(PL)EMP4s.said
‘(I heard) some boy was walking with a girl when they were picked up (presumably by the police).’
(12)ʼashkiiʼatʼéédyiłyiʼash =go
boygirl3s.with3DUAL.walk=when
nábiʼdiiʼnilłájiní
3.picked up(PL)EMPj4s.said [sic]
‘That person told me that a boy and girl were picked up while walking.’

P. 115: in (11) “the fourth person appears in each of the inflected verbs”; in (12) “fourth person appears only on the final verb, the verb of speech.” Pp. 115–116: “The difference between the two consists in the fact that the former is a Reportative, while the second is a use of the fourth person in the verb jiní to indicate politeness or respect. With the Reportative (11), the person uttering the sentence may have read of the event via the paper, heard about it on the radio, or was privy to a conversation where the incident was discussed, but does not want to take responsibility for the truth of what was said. In (12) the fourth person pronoun in the final verb is not a reportative, but refers to a specific person who is present or nearby, and is used in deference. Note that in this usage, the verbs in the preceding clauses do not show any fourth person pronouns.”

P. 116, on léiʼ: “The indefinite is needed to give that particular interpretation and its omission would produce an ungrammatical sentence, since the referents are not presupposed in the discourse. Aside from these special contexts, the normal course of events is that only third person will permit nominal adjunction.”

Print. (11) glosses bił “s3.with”, where (12) has “3s.with” for yił. (12) glosses jiní “j4s.said”, where (11) and p. 114 have “4s.said”. Both copied with [sic]. (12) has =go spaced off from yiʼash, as printed. Willie spells the typo “shwon” on p. 115. “4s.walking” (11) vs “3DUAL.walk” (12) for the same pair of walkers is copied as printed. In (11) the postposition bił keeps a third-person gloss while both verbs are fourth (T15b).

6. Traditional stories (§4.2.4, pp. 116–117) SOURCED

Pp. 116–117: “There are certain conventions employed in narrative texts where the fourth person permits nominal adjunction. This is a device to signal that the speaker is telling a story. It is comparable to the Quotative use cited in the preceding section.”

Willie 1991:117, (13)–(14)
(13)mąʼiiyiildloshjiní
coyote3.walking4.said
‘(I hear) that there is a coyote around here.’
(14)mąʼiijooldlooshjiní
4.walking4.said
‘Once upon a time there was this coyote walking along...’

P. 117: “While (13) is a Quotative, (14) introduces a traditional story, and again we see that the subordinate verb is inflected for fourth person as well as the verb of speech, jiní.” Print: the gloss line of (14) has no gloss under mąʼii, kept blank. yiildlosh (13) vs jooldloosh (14) differ in vowel length as well as person, as printed.

7. The Impersonal (§4.2.5, pp. 117–119) SOURCED

Pp. 117–118: “A much desired attribute sought in accordance with Navajo philosophy is to have control over one's own actions, and to speak only for oneself. The fourth person allows the speaker to remain impartial and detached from the events under discussion and thus avoid any possible hint of criticism or accusation (Witherspoon, 1977:84).” P. 118: “These gentle instructions are conveyed in fourth person Impersonal sentences. The speaker has not accused the hearer of behaving wrongly, nor does he presume to tell him what to do.”

Willie 1991:119, (15)–(16)
(15)Doohatisnjigháada
NEG4.over4.walksNEG
‘One does not step over people.’
(16)ʼabíní=gohojitaał=goyáʼátʼééh
Morning=while4.sings=while3.good
‘It is good for one to sing in the morning.’

P. 119: “Since these sentences are for instruction or advising they often contain a value judgment or a negative. That is, some modal feature. (15) and (16) are "indirect speech acts" -- that is, instructions intended to benefit some hearer (as well as anyone else who may be within hearing), but without addressing a second person directly or openly criticizing anyone's behavior.” (16) “could be used when the intended hearer was still asleep in the morning”. Willie's “neccessarily” and “safely reasons” (p. 118, for “safety”) are as printed.

HYPOTHESIS (from the translation) In (15) there are two π = 4 slots in one clause: the walker (njigháa “4.walks”) and the object of ‘over’ (hatis “4.over”). The translation gives them different referents, ‘one’ and ‘people’. So “two fourth persons take the same referent” (PI-MATCH ii) is a cross-clause default, not something the value does inside a clause.

8. The Optative (§4.2.6, pp. 120–122) SOURCED

P. 120: “Another use of the fourth person to express politeness is the use of the fourth person in an Optative mode. Since it is not considered polite to speak openly of one's own desires or wishes, a speaker will use fourth to express that wish.”

Willie 1991:120–121, (17)–(19)
(17a)ʼajoołhááshlaanaa
4.O?T.sleep [sic: damaged type; cf. 4.OPT in (17b)]wishfully
Lit. ‘That certain person wants to sleep.’ (‘I want to sleep.’)
(17b)ʼajoołhááshlágo
4.OPT .sleepNEG
‘Don’t let him go to sleep.’
(18)Shizhéʼéłį́į́ʼnahijólniihnizin
1.fatherhorse3.4.OPT.buy3.thinks
Lit.: ‘My father thinks that that person should buy a horse.’ ‘My father wishes he could buy the horse.’
(19)ʼajoołhááshlaanaanínízin=ísh
4.OPT.sleepwishfully2.thinks=Q
‘Do you wish to go to sleep?’

P. 120: “Ex. (17.b) shows that with lágo and the verb in an optative fourth person, it gives a prohibitive reading.” P. 121: “The literal interpretation for (18) indicates that only the main verb is coreferent with 'my father' and the agent referent of the verb 'buy' is a polite fourth; that is, someone outside the sentence. But in the actual meaning the fourth person in optative mode always refers to the speaker of the clause indicating that reference outside the clause is not possible.” On (19): “Here again the person in the main verb is coereferent with the optative fourth. We see that the fourth person can be used to refer to any person: it is linked to the speaker in (17) by default because there is no subordinate clause; it is linked to third person human in (18), and in (19) linked to second person, which is marked on the main clause verb. All of these uses of the fourth person in indirect speech indicate that the fourth is used in order to avoid reference to any first, second or third person; depending on the context, it can refer to any person.” (pp. 121–122; “coereferent” as printed.)

Print: (17b) spaces “4.OPT .sleep” as printed. On (17), Willie writes “because there is no subordinate clause”; (17) has no main ‘think’ clause either. Quoted as printed. Gloss order “3.4.OPT.buy” = object 3 (the horse), subject 4 (the would-be buyer).

9. Rules

Notation as on the reported speech page. π(ρ) is the person value of a slot, M the main clause, K the complement. “Adjoin N to ρ” means a noun is linked to slot ρ at clause level (B). Each rule restates a statement of Willie's. Added parts are labelled.

NOM-4      A nominal cannot be adjoined to a π = 4 slot: nouns are 3rd person.  (p. 109; *(5c), *(6f))
PRO-MATCH  A free pronoun may be adjoined if it matches π: Shí/1, Ní/2, Bí/3, Hó/4. (p. 110; (6a–e))
X-MOTHER   Respect exception: a nominal may adjoin to π = 4 when the speaker refers
           to his mother; "not ... with other kin terms."                   (p. 113; (8))
X-REPORT   Reportative exception: 4 on each inflected verb; a nominal may adjoin to a
           π = 4 slot, and needs indefinite léi' ("referents are not presupposed").
                                                                            (pp. 115–116; (11))
X-STORY    Narrative exception: in a story opening, 4 on the subordinate verb and on
           jiní; a nominal may adjoin to the π = 4 slot.                    (pp. 116–117; (14))
RESP-JINI  Respect on 'say': 4 only on jiní, a specific person "present or nearby";
           preceding verbs show no 4; nominals link to their 3 slots.       (pp. 115–116; (12))
IMP-4      Impersonal: π = 4 = generic 'one'; used for instruction, usually with a
           negative or value judgment ("some modal feature").               (pp. 118–119; (15), (16))
OPT-4      Optative: π = 4 in the optative "always refers to the speaker of the clause":
           the speaker by default (17a); the subject of a matrix 'think', whatever its
           person value: 3 (18), 2 (19).                                    (pp. 120–122)
OPT-PROH   Optative 4 + lágo: prohibitive reading (17b).                   (p. 120)
ONE-4      [Willie's unifying statement] All uses: the fourth "is used in order to avoid
           reference to any first, second or third person; depending on the context, it
           can refer to any person."                                        (pp. 121–122; cf. p. 113)
RuleStatusBasis
NOM-4, PRO-MATCH, X-MOTHER, X-REPORT, X-STORY, RESP-JINI, IMP-4, OPT-4, OPT-PROH, ONE-4SOURCEDWillie's statements and examples as cited. The rule names and the slot notation are ours.
OPT-4 stated as a link to the thinker slot of ‘think’ (OPT-LINK, §11)HYPOTHESISWillie says “the person in the main verb”. We restate it as a role slot (ρExp of ‘think’), because the link holds whatever that slot's person value is.
The three noun exceptions share a property (§10)HYPOTHESISWillie lists them; she does not unify them.

10. T13a: one fourth person or two?

  • Willie's own position: one item with several uses. SOURCED Her section titles call each one a use (“Another use of the fourth person …”, pp. 117, 120). P. 113 calls the fourth “a strategy to avoid using the name of person or to point at him”, “a deictic pronominal”. Pp. 121–122 say it is “used in order to avoid reference to any first, second or third person”. On these pages she never proposes two morphemes, and every use is written with the same subject/object shapes she names as ji-/ho- (p. 111).
  • What tells the uses apart is distribution and context, not form. SOURCED Respect: 4 on jiní only, a specific present person (12). Reportative: 4 on every verb, nominal with léiʼ (11). Story: 4 on both verbs, nominal (14). Impersonal: a modal or negative, generic ‘one’ (15)–(16). Optative: the optative mode, linked to the clause's own speaker or thinker (17)–(19).
  • The “4 on every verb” test is one-way. HYPOTHESIS Willie tells (11) from (12) by whether the 4 spreads. But her own quotatives (9) yicha jiní and (13) mąʼii yiildlosh jiní have 3 in the complement, the same person pattern as respect (12). And the respect reading of Ch. 5 (26a) jidoogááł jiní has 4 on both verbs. So spreading points to a reportative or story, but a quotative need not spread, and a respect form can co-occur with a second 4 (T15c).
  • A shared property of the three noun exceptions. HYPOTHESIS Two of the three (reportative, story) are contexts where the speaker is not vouching for the events and the referents are “not presupposed” (p. 116). There the 4 is not pointing at anyone present, which is the job NOM-4's reason (“so specific”, p. 109) rests on. The mother exception (8) does not fit this and is lexically limited (“not … with other kin terms”). So we list it separately.
  • Verdict. T13a is narrowed. Willie's answer, one fourth person with discourse uses, is now SOURCED, and no form difference appears in pp. 109–122. An independent morphological test, such as a paradigm where respect and quotative jiní differ in any prefix, is still OPEN (Young & Morgan 1987, not opened).

11. PI-MATCH revised

PI-MATCH (reported speech §6, extended on the relatives page): (i) a π = 4 slot never takes the referent of a π = 3 slot; (ii) two π = 4 slots take the same referent by default; (iii) the 3/3 default is set by the construction.

  • (i) fails under the optative. SOURCED In (18), the π = 4 buyer of nahijólniih is the π = 3 thinker of nizin, ‘my father’ (“linked to third person human in (18)”, p. 121). Compare ‘say’ (25) jidoogááł ní ‘4.arrive 3.said’, ‘He said that person will come’: there a 4 in the complement and a 3 on the main verb read as different people. So (i) holds for =go, ‘say’, relatives and ‘believe’, but not for an optative complement of ‘think’. Revised (i): outside the optative, a π = 4 slot never takes a π = 3 slot's referent.
  • (ii) is cross-clausal and needs a referent. HYPOTHESIS Inside one clause, (15) has two π = 4 slots with different referents (translation). Across clauses, a quotative 4 on jiní has no specific referent, so in (11) it does not match the 4s in the complement (the boy and girl are not the reporters). That fits S4 reading b on the reported speech page.
  • OPT-LINK (ours). HYPOTHESIS In [KOPT] nizin, the π = 4 subject slot of K takes the referent of the thinker slot ρExp of ‘think’, whatever π(ρExp) is: 3 (18), 2 (19). With no ‘think’, it defaults to the speaker (17a). The link goes from role slot to role slot. A noun cannot be doing it: Shizhéʼé in (18) is linked to the thinker, and NOM-4 keeps it off the buyer slot. A matching person value cannot be doing it either, since 4 ≠ 3 and 4 ≠ 2. What is left is the role relation: the one who wishes is the one who would act. That is a clear case of Grant's “theta roles, not lexical items, as the arguments”.
  • A resemblance, not a merger. HYPOTHESIS OPT-LINK makes the Navajo optative 4 behave like the West Greenlandic reflexive third person (3r), which must be the main-clause subject (fourth person compared §3.4). Only the optative does this. The other Navajo uses do not, so the Navajo fourth person as a whole is still not the Greenlandic one.

12. Derivations P1–P6, step by step HYPOTHESIS

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). Verbs are not segmented beyond Willie's glosses. 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).

P1. (6d) ʼashkii yícha, (6e) Hó, jíícha, *(6f) ʼashkii jíícha

  1. Σ Σ(√CRY) = {ρ1}. π(ρ1) = 3 (6d) or 4 (6e, f).
  2. Merge v + ρ1; Merge M/S; Agree; Move (V-P) ρ1 to Spec,M/SP. A Spell-out: yícha / jíícha. The word is now closed. This is the A → B changeover.
  3. B Adjoin a nominal or free pronoun to the left, linked to ρ1. The link checks π. ’ashkii (3) + π = 3 ✓ (6d). Hó (4) + π = 4 ✓ (6e, PRO-MATCH). ’ashkii (3) + π = 4 ✗ (6f, NOM-4).
  4. HYPOTHESIS The check needs only the person value the word already carries (A) and the noun's person (B). So the ban applies at the changeover point, after the word is built and when a noun is linked to a slot inside it. Under V-P and V-H alike the word is the same, so P1 does not decide VPISH.

P2. (12) respect: ʼashkii ʼatʼééd yił yiʼash =go nábiʼdiiʼnil łá jiní

  1. Σ K1: Σ(√WALK.DU) = {ρ1}, π = 3 (“3DUAL.walk”). Postposition ‘with’: object π = 3 (“3s.with”). A yiʼash, yił. B ʼatʼééd links to the object of yił, ʼashkii to ρ1. Merge C = =go.
  2. Σ K2: ‘pick up (PL)’, the picked-up slot π = 3 (“3.picked up(PL)”). A nábiʼdiiʼnil. The picker is not expressed (translation ‘were picked up’). Which slot that is, we do not segment (T15b).
  3. Σ M: Σ(√SAY) = {ρA, ρTh = K}, π(ρA) = 4. RESP-JINI: ρA is a specific present person. A jiní.
  4. B [K1=go] K2 łá M. Every noun links to a π = 3 slot, so NOM-4 is never violated.

P3. (11) reportative: ʼashkii léiʼ ʼatʼééd bił jooʼash=go náhodiiʼnil łá jiní

  1. Σ X-REPORT: a reportative feature on the root clause (B, our placement, HYPOTHESIS) requires π = 4 on each inflected verb.
  2. K1: ρ1 of ‘walk’ π = 4 (“4s.walking”). ‘With’ keeps object π = 3 (“s3.with [sic]”). A jooʼash, bił. K2: picked-up slot π = 4 (“4s.picked up(PL)”), náhodiiʼnil. M: ρA of ‘say’ π = 4, quotative (no specific speaker), jiní.
  3. B ʼatʼééd links to the π = 3 object of bił (normal). ʼashkii léiʼ links to the π = 4 ρ1 of jooʼash. NOM-4 would block this. X-REPORT lifts the block, on condition that the noun is indefinite (léiʼ); without it the sentence is ungrammatical (p. 116).
  4. Reference: the quotative ρA of jiní has no specific referent, so PI-MATCH (ii) does not link it to the 4s in K (§11).
  5. HYPOTHESIS Where the reportative is marked: the 4s are spelled inside each word (A), but the trigger is the whole report (B). One way to say it is that a clause-level choice reaches down into each word's person value. Because the postposition stays 3, the feature reaches verbs, not every person slot (T15b). That is the only place on this page where a B-level choice shows up inside A-level words.

P4. (15) impersonal: Doo hatis njigháa da

  1. Σ Σ(√WALK) = {ρ1}, π = 4 (“4.walks”; that this is the singular ‘walk’ of the dictionary's naa-ghá is our guess). Postposition ‘over’, object π = 4 (“4.over”). IMP-4: both generic.
  2. A njigháa, hatis. B The postpositional phrase precedes the verb (head-final). The negation doo … da frames the clause.
  3. HYPOTHESIS Two π = 4 slots in one clause with different referents (‘one’, ‘people’), so (ii) is cross-clausal (§11). The negation is the modal feature Willie says these sentences “often contain”. Whether IMP-4 needs it is OPEN: (16) has ‘good’ instead.

P5. (18) optative: Shizhéʼé łį́į́ʼ nahijólniih nizin ‘My father wishes he could buy the horse.’

  1. Σ K: Σ(√BUY) = {ρA, ρP}, mode OPT. π(ρP) = 3 (the horse), π(ρA) = 4 (gloss “3.4.OPT”). Merge V + ρP, v + ρA, M/S; Agree; Move (V-P) ρA. 4-NO-INV: no yi-/bi- choice. A nahijólniih.
  2. B łį́į́ʼ links to the π = 3 ρP, to the left inside K.
  3. Σ M: Σ(√THINK) = {ρExp, ρTh = K}, π(ρExp) = 3 (“3.thinks”). A nizin. B K precedes nizin, and Shizhéʼé (linked to ρExp) comes first.
  4. Reference: OPT-4 / OPT-LINK. ρA(K) = ρExp(M) = my father, although π differs (4 vs 3). Shizhéʼé cannot link to ρA(K) directly (NOM-4). It reaches it only through ρExp.
  5. HYPOTHESIS VPISH: the link is between two role slots in different words, made at clause level (B), after each word is built (A). That is the same place the =go and ‘say’ reference rules apply. It needs no movement out of K, so V-P and V-H do not differ here. What it does show is that the arguments the clause-level rule sees are role slots (ρA, ρExp), not nouns.

P6. (19) ʼajoołháásh laanaa nínízin=ísh and (17b) ʼajoołháásh lágo

  1. K: ‘sleep’, OPT, π(ρ1) = 4: ʼajoołháásh, followed by laanaa ‘wishfully’ (B).
  2. M (19): ‘think’, π(ρExp) = 2: nínízin, then =ísh (Q) at the right edge (B, head-final). OPT-LINK: ρ1(K) = the hearer. ‘Do you wish to go to sleep?’
  3. (17b): no ‘think’. lágo (NEG) follows. Willie: prohibitive. The translation ‘Don’t let him go to sleep’ gives the 4 a referent that is neither the speaker nor a thinker. OPEN (T15d): OPT-4 says the optative fourth “always refers to the speaker of the clause”, and (17b) is a counterexample unless lágo changes the link.

13. Open questions

IDQuestionWhat would settle it
T15aWhy does the indefinite léiʼ make a noun on a reportative 4 grammatical (11), and does a story opening need anything similar? (14) has none.More reportative and narrative sentences with nouns; Grant's texts (traditional stories). OPEN
T15bIn (11), the 4 spreads to both verbs but the postposition bił stays 3 (“s3.with [sic]”). Does the reportative 4 reach only verbs, or only some slots (e.g. not the girl)? Which slot of náhodiiʼnil is the 4?A paradigm for ‘pick up (PL)’; more reportatives. OPEN
T15cIs “4 on every verb” only a sign of reportative/story, given quotatives with 3 in the complement ((9), (13)) and respect with 4/4 (Ch. 5 (26a))?Willie gives no minimal set. A speaker. OPEN
T15dOptative 4 “always refers to the speaker of the clause” (p. 121), but (17b) ʼajoołháásh lágo is ‘Don’t let him go to sleep’. Does the prohibitive lágo change the link?More optative + lágo sentences. OPEN
T15ePI-MATCH (i) fails in (18). Is the optative the only case, or does ‘think’ (nizin) let a 4 take a 3's referent with non-optative complements too? This is part of T14a: ‘say’, ‘believe’ and now ‘think’ all differ.‘think’ with non-optative complements in 3/3, 3/4, 4/3. OPEN
T15fUyechi 1990 claims nouns can link to fourth persons. Willie and her consultants reject this, and Robert Young questioned Uyechi's examples (p. 110). Dialect, register, or the X-REPORT/X-STORY contexts?Uyechi 1990 (not opened). OPEN
T15g(5a) yicha and (6a) yícha are printed under first-person translations, in the spelling of the third-person forms. Misprint, or a form we do not understand?Another Willie example of 1st-person ‘cry’; Y&M. OPEN

Not read this run: Willie §4.1 and the start of §4.2 (before p. 109); Young & Morgan 1987 on the fourth person paradigm; Witherspoon 1977; Uyechi 1990. §4.2.7 (p. 122ff.) was already on the fourth person page.

Three things called “fourth person”: Navajo, Algonquian, Inuit

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.

Working note. Kept as written (URL, data and status labels unchanged). Filed under The fourth person; the consolidated argument is the essay fourth-person. Rule IDs: rules index.

Verbs on this page → dictionary: yiztsʼǫs ‘kiss’ · yishxash ‘bite’

Restated on the unified structure (2026-11-15). The fourth person is restated as a person value on an argument merged in Spec,vP or VP, with nothing extra moving: the optative fourth bound by the wisher (H51), the respect/report/story ji- as an Agree mark (H57a, OPEN against H57b), the areal ho- as a personless area argument (H54a, H60). 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). Status on the spine: see unified structure, §4, fourth-person rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

The label “fourth person” and the label “obviation” are used for several different grammatical devices. Grant's backlog item asks for the Navajo fourth person to be compared with other Athabaskan languages, with Algonquian obviation (proximate vs obviative), and with the Inuit “fourth person”. Each comparison has to be sourced, and different phenomena that only share a name must not be merged. This page does that with four opened sources. It has two results. (1) The labels cover three different jobs: ranking two third persons by position (Algonquian obviation), a separate person value that sits outside the ranking scale (Navajo fourth), and a third person that is bound by the matrix subject (West Greenlandic “3r”). Navajo's counterpart to Algonquian obviation is the yi-/bi- voice system, not its fourth person. (2) Bruening's Passamaquoddy work is a sourced case where an inverse is VPISH-style A-movement of the object over the subject. It gives the site's open scope test (T5c / T10a) a template. It also shows that one of its three diagnostics cannot be built in Navajo with the rules already on the site. New open questions are T11a–T11f.

On this page: 1 Question · 2 Sources · 3 Data · 4 Features compared · 5 Three jobs, not one · 6 The Algonquian inverse and VPISH · 7 Derivations X1–X4 · 8 Verdict · 9 Open

1. The question

GRANT'S CONJECTURE (backlog, 2026-10-08) The Navajo fourth person should be compared cross-linguistically with other Athabaskan languages (Apache, Slave), with Algonquian obviation, and with the Inuit/Yupik “fourth person”, which is a coreferential third person. Every comparison must be sourced. Phenomena that only share a label must not be merged.

The site's previous answer (fourth-person page, T8) was that Navajo's fourth person is not a second inverse. It is an exit from the animacy scale. Willie divides the work between voice (yi-/bi-) and fourth person (“obviation and deference”). The questions here are these.

  • Q1. Is Navajo's fourth person the same device as Algonquian obviation, or as the Inuit fourth person, or as neither?
  • Q2. Does any of these systems give sourced evidence about VPISH: does an inverse move an argument, and how would we tell?

2. Sources opened this run (2026-10-09, afternoon)

  • Willie, MaryAnn 1991. Navajo Pronouns and Obviation. PhD dissertation, University of Arizona (UMI 9210298; the scan already on the box). We read page images of printed pp. 18–20 (PDF pp. 23–25), pp. 96–100 (PDF pp. 102–106) and pp. 216–217 (PDF pp. 222–223). The Navajo lines of (28)–(29) in Ch. 8 were checked on 300 dpi crops. The OCR text layer was used only to find pages.
  • Bruening, Benjamin 2005. “The Algonquian Inverse is Syntactic: Binding in Passamaquoddy.” Manuscript, University of Delaware, dated July 21, 2005, 27 pp. Its first footnote says it is “a substantially revised version of chapter 2 of Bruening (2001)”, his MIT dissertation Syntax at the Edge. Author's PDF, udel.edu/~bruening/Downloads/SyntacticInversion2.pdf. The PDF is born-digital, so the page numbers below are printed pages = PDF pages. Pp. 13–14 were also checked as images.
  • Lybech, Stian L. 2020. An Introduction to West Greenlandic. Self-published PDF (oqa.dk), XeLaTeX, 269 PDF pages. Printed pp. 115–116, 135, 153, 156, 176 (PDF = printed + 16). P. 156 was also checked as an image.

Secondhand only. Slave (Rice 1986, 1989) and Sarcee (Cook 1984) are reported here only through Willie 1991:18–19. Neither was opened. No Apache source was opened this run. Sandoval & Jelinek's paper on Apachean yi-/bi- (in Cook & Rice eds.) is cited by Willie and is summarized in the annotated bibliography of the Jelinek compilation already on the box. It is not a source for any claim here. No Yupik source was opened. The Inuit column is West Greenlandic (Kalaallisut) only.

Transcription and glosses. Navajo is given as printed, with the glottal stop as ’ (site practice) and Willie's glosses copied exactly. Passamaquoddy is given in Bruening's orthography, with his glosses and translations copied exactly (his footnote 1: ’ = initial h, o = schwa, q = [kʷ], c = alveopalatal affricate). Bruening leaves direct, independent, proximate and singular unglossed when they are the unmarked value. Lybech gives his Greenlandic examples without interlinear glosses. The segmentation and Leipzig labels in §3.4 are ours. They are read off his ending tables (Figure 9.3, p. 153) and his text, and are labelled as ours.

3. Data SOURCED

3.1 Navajo: Willie on fourth person, obviation and the inverse

Willie 1991:18: “Prefixes that are obviously cognate with the Navajo fourth person forms seem to be present in most Athabaskan languages, but its function differs from language to language.” Same page: fourth person is “reserved for humans and also for personified animals” (agreeing with Akmajian & Anderson 1970), and it “has properties that are very similar to first and second pronominal”. P. 122 (already on the fourth-person page): “Fourth person, like first and second, is definite, human, and not involved in the animacy hierarchy.”

Willie 1991:96, after the postpositional pair (87)–(88): “Note that this same interpretative principle applies with Inverse marking on the verb; it signals that the immediately preceding noun is not the object of the verb. In this respect, the Inverse marking functions like obviative marking. However, note that the Inverse has a further property: it always makes the argument with an agent theta role the internal argument.” (p. 97; underlining as printed.)

Willie 1991:216–217, Ch. 8, (28)–(29)
(28a)Johnháíshbaghan=diyiztsʼǫs?
who3POSS:home=at3.3.kissed
‘Who did John kiss at his house?’
(28b)Johnháíshyighan=diyiztsʼǫs?
who3POSS:home=at3.3.kissed
‘Who did John kiss at her house?’
(29)Johnháíshhaghan=diyiztsʼǫs
who4POSS:home=at3.3.kissed
‘Who did John kiss at that other person’s house?’

As printed: (29) has no question mark, and Willie's glosses fuse the possessor and the noun (3POSS:home), so they are not segmented here. The three words differ only in their first consonant, b- / y- / h-. That is an observation about the strings, and the morpheme analysis is Willie's gloss. P. 216: “Note that it is possible for the string háísh bighan-di to undergo the yi-/bi- alternation.” After (28): “This contrast in Possessor marking is not one of voice, but shows the use of this alternation to mark obviation. It is occurring with a Lexical postposition, not a Grammatical postposition. Speakers vary in the extent to which they use yi-/bi- in this way. Some use it with kin terms, others do not.” Then: “Since Navajo has fourth person forms, which are used to mark obviation, there seems to be less reliance on yi-/bi- for this purpose.” Willie does not say whose house ‘his’ and ‘her’ refer to in (28). We do not fill that in.

3.2 Other Athabaskan: secondhand, through Willie

Willie 1991:18–19, (5); Sarcee from Cook 1984
(5)ʼatsʼinilaa‘someone said’
No gloss line in Willie. Not segmented here.

pp. 18–19: “In Slave, a cognate form is a third person object when the subject is third person and disjoint reference is intended (Rice, 1986). In Sarcee, Cook (1984:195) identifies an unspecified third ts’i and a yi- prefix which he considers a fourth person. This fourth person seems to function similarly to the Slave prefix: it refers to an object when the subject is third, and also if the object NP is not overt. But Cook points out that yi- can be a subject in some cases (1984:201). And ts’i ‘…is chosen instead of zero or gi-, not necessarily because such an identity of the subject in question is unknown or unspecified, but because such an identity is irrelevant or redundant in a given context’.” (5) “could be used in texts when the speaker's identity is not in question.” P. 19: Young & Morgan 1987 call Navajo fourth subject ji- “a deitic pronoun” and say it “is used to distinguish between two third persons”.

Label warning. In Willie's report of Cook, the Sarcee prefix that Cook calls “fourth person” has the shape yi-. In Navajo, yi- is the direct third-person object of the voice system. This page does not say which Sarcee prefix is cognate with which Navajo one. Willie does not say either, and no Sarcee source was opened (T11a).

3.3 Algonquian (Passamaquoddy): Bruening 2005

Bruening 2005:1–2, (1)–(3)
(1a)K-ucem-a-k.
2-kiss-Dir-3P
‘You kissed them.’
(1b)K-ucem-ku-k.
2-kiss-Inv-3P
‘They kissed you.’
(3a)Pesq muwinʼ-toli-nuhsuphoqal-amahtoqehsu.
one bear3-Prog-chase-Dir.ObvPrabbit.ObvP
‘One bear (Prox) was chasing some rabbits (Obv).’
(3b)Mahtoqehsʼ-toli-nuhsuphoqal-ku-lmuwinuw-ol.
rabbit3-Prog-chase-Inv-Obvbear-Obv
‘A rabbit (Prox) was being chased by a bear (Obv).’

p. 2, (2): “Participant Hierarchy: 1,2 > 3 > Obv > Inan” (from Goddard & Bragdon 1988). “The proximate-obviative distinction is one that is made only for animate third persons. In any syntactic context (usually coextensive with the clause) one third person is distinguished as proximate …, while all others must be obviative”. On (3b): “it is important to note that this is not a passive. There is no loss of an external argument here, and no change in valence.”

Bruening 2005:12–14, (30), (33), (36b), (37a), (39a)
(33)[NP Skitap musqitaham-ac-il] ʼ-koti-tqon-a-l psi=te wen-il.
man hate-3Conj-PartObv 3-Fut-arrest-Dir-Obv all=Emph someone-Obv
‘A man that he∗1 hates will arrest everyone1.’ (direct: an object quantifier cannot bind into the subject)
(36b)Yatte wen pilsqehsisʼ-kis-cem-ku-lw-ikuwoss-ol.
each who girl3-Perf-kiss-Inv-Obv3-mother-Obv
‘Her1 mother kissed each girl1.’ (inverse: the object quantifier now binds into the subject)
(37a)Wen pihce w-itapihi-l nekol-iht kcihku-k?
who long.ago 3-friend-Obv IC.leave-3ConjInv forest-Loc
‘Who1 did his1 friend abandon in the forest a long time ago?’ (inverse avoids weak crossover; compare the starred direct (34b))
(39a)Ma=te keq utomeya-ku-w-on [NP tepelto-k ].
Neg=Emph what 3.bother-Inv-Neg-N IC.own.TI-3Conj
‘Nothing1 bothers the one who owns it1.’

p. 12, (30): “Syntactic Movement a. Direct: Subject > Object b. Inverse: Object > Subject (> tObj)”. P. 14, after (37): “the syntactic hierarchy has reversed in the inverse voice, such that the object now dominates the subject.” P. 14, §3.2.3: “the inverse does not just lead to inverse scope, it is actually ambiguous. The object can take scope over and bind a variable in the subject in the inverse, but the reverse is also possible”. In (39a) the subject quantifier ‘nothing’ binds into the raised object. P. 15: “Scopal reconstruction is only a property of movement chains”. Pp. 15–16, (40)–(42): it is “not surface word order that matters” for binding. P. 3: “There are exactly three A-positions in a simple transitive”: A1 (Spec,IP), A2 (subject, Spec,VoiceP) and A3 (object, Comp,V). P. 6: “The final will agree with the object in its low position. Whatever moves to the higher position will control agreement in the prefix.” P. 17: “the prefix always indexes the highest argument, and the final suffix the lowest.” P. 21, §4.1: “proximate NPs always c-command obviative NPs”, and “Possessors are always proximate while the possessed noun is always obviative.” Also p. 21: “there was one piece of the account above that was optional: giving Voice an [EPP] feature to drive movement of the object”, and discourse can exploit that option. P. 7: Condition C “simply appears not to apply”. (14a) ‘He1 hid all of the governor1’s money’ is the default reading.

3.4 Inuit (West Greenlandic): Lybech 2020

Lybech 2020:156
a.Piitaqoqar-poqkaffisor-toq
Petersay-IND.3SGdrink.coffee-PTCP.3SG
‘Peter said that he drank coffee.’ (he ≠ Peter)
b.Piitaqoqar-poqkaffisor-luni
Petersay-IND.3SGdrink.coffee-CONT.3R.SG
‘Peter said that he drank coffee.’ (he = Peter)

The strings, the translation and the “(he ≠ Peter)” / “(he = Peter)” notes are Lybech's. Lybech prints the examples unglossed, and the hyphens and the gloss line are ours. We read them off his statements that the endings are “Vb{ðuq} or Vb{(l)luni}” (p. 156), off Figure 9.3 (p. 153: indicative 3p.sg Vb{vuq}, participial 3p.sg Vb{ðuq}, contemporative 3r.sg Vb{(l)luni}, with no 3p.sg contemporative and no 3r participial or indicative), and off {uqaq}Vb ‘say s.t.’ (p. 156). kaffisor- is glossed from his translation only.

p. 115: Greenlandic “has a special, reflexive third person, which I abbreviate ‘3r.sg’ and ‘3r.pl’, to denote that the ‘he’ doing something in the first sentence (or ‘main clause’) is the same as the one who is doing something in the second sentence (or ‘subordinate clause’)”. Footnote 4: “Some authors, e.g. Bjørnum (2003) call the reflexive third person the ‘fourth person’ instead. I have decided not to do this, because to me, a ‘fourth person’ sounds like it is someone else, rather than the same person”. P. 156: “the participial … can only be used when there is no co-reference between the Subject in the subordinate clause and the Subject (or Agent) in the main clause”; “the contemporatives can (generally) only be used when there is co-reference”; the indicative “always forms main clauses, and thus we obviously never need the reflexive third person here”. P. 176: the indicative has no combination with 3r “neither as Agent nor as Patient”. P. 135: possessors too. ‘To his (own) daughter’ is panimminut, with the allative ending N{minut} built on the 3r.sg ergative marker {mi}. P. 115: there is no gender distinction, and “The 3.sg can be a man, a woman, a dog, a house etc.”

Update 2026-10-10 (evening): one Navajo use comes close to 3r. The optative fourth under ‘think’ must be the main-clause thinker (Willie 1991:120–121, (18)–(19); fourth person uses §11). The other Navajo uses do not behave this way, so this is a resemblance in one construction, not the same device.

4. Six features, compared

We compare the devices on six features that the sources can decide. Each cell cites its source. “—” means the opened source does not say.

FeatureNavajo fourth (ji-/ha-/ho-)Navajo yi-/bi-Passamaquoddy obviativeWest Greenlandic 3rSlave / Sarcee (secondhand)
F1 Anchor: what fixes the referenceDeictic. It needs no antecedent (Willie 124–125). It is disjoint from a 3rd person in (29) ‘that other person’ and in Y&M (via Willie 19). It can force coreference with another 4 (W&J (77)).Linking: which slot is EXT (inverse page, INV-6). On possessors, it is used “to mark obviation” (Willie 216).Rank among animate 3rds. Exactly one proximate per context, and all others are obviative (Bruening 2).Coreference with the matrix subject/agent (Lybech 115, 156).Slave: disjoint from a 3rd-person subject (Rice via Willie 18).
F2 DomainClause and discourse (Q–A pairs, Willie 127–130). Across complex NPs (W&J (77)).Clause; possessor inside a PP (Willie 216).“usually coextensive with the clause” (Bruening 2). Also inside NPs, where the possessor is proximate (Bruening 21).Matrix + dependent clause, and possessors. Never a main-clause (indicative) argument (Lybech 156, 176).—
F3 Relation to the ranking scaleOff the scale, like 1/2 (Willie 122).The scale's device: it is chosen by animacy rank (Willie 122).On the scale, as its lowest animate rank: 1,2 > 3 > Obv > Inan (Bruening 2).— (no ranking discussed by Lybech)—
F4 Fixed by structural position?No: a 4 can be agent or patient in the direct ((23a) hashxash, (23b) jishxash, Willie 124). 4 cannot enter yi-/bi- (W&J 277).Yes: EXT vs internal (Willie 97: inverse “always makes the argument with an agent theta role the internal argument”).Yes: “proximate NPs always c-command obviative NPs” (Bruening 21).Bound by a higher-clause subject (Lybech 156). Whether this is c-command is not stated (—).Sarcee yi-: object when the subject is 3rd, but “can be a subject in some cases” (Cook via Willie 19).
F5 Which referentsHumans and personified animals; definite (Willie 18, 122; W&J 278).Third persons ranked by animacy.Animate third persons. In (3) both bear and rabbit are animals.Third persons. Lybech does not restrict 3r by animacy (—). Ordinary 3sg covers “a man, a woman, a dog, a house” (115).—
F6 Tied to a direct/inverse voice?No. It is excluded from the alternation (W&J 264, 277).It is the voice.Yes. Direct iff the subject is proximate, inverse iff the object is (Bruening 2, 21).— (no voice system in the pages opened)—

5. Three jobs, not one

RANK-OBV   (Passamaquoddy obviation; Bruening 2005)
           In a domain D with two animate 3rd persons x, y: if x c-commands y at the
           A-position level, x = proximate and y = obviative.  Direct iff x = subject.
PERSON-4   (Navajo fourth; Willie 1991, W&J 2000)
           π(ρ) = 4 is a person value parallel to 1 and 2: definite, human, deictic,
           not in R (the ranked set).  Its value does not depend on ρ's position.
BOUND-3R   (West Greenlandic 3r; Lybech 2020)
           A 3rd-person argument or possessor of a dependent clause is 3r iff it is
           coreferent with the matrix subject/agent.  The indicative has no 3r slot.
RuleStatusBasis
RANK-OBVSOURCED as Bruening's generalization; it is his analysis, not a consensusBruening 2005:2, 12–14, 21. Pp. 2–3 set it against Aissen 1997's hierarchy-based account and inverse-linking accounts.
PERSON-4SOURCED (facts); “value independent of position” is our HYPOTHESIS from (23a, b)Willie 1991:18, 122–125; W&J 2000:264, 277–278; site rules EXIT, 4-NO-INV, 4-DEIX.
BOUND-3RSOURCEDLybech 2020:115, 135, 153, 156, 176.

What matches what HYPOTHESIS (from the SOURCED cells of §4)

  1. Algonquian obviation corresponds to Navajo yi-/bi-, not to the Navajo fourth person. Both are ranking devices within the third person, read off position, and tied to a direct/inverse alternation (F3, F4, F6). Willie draws the parallel herself (p. 96: the inverse “functions like obviative marking”), and she uses yi-/bi- on possessors “to mark obviation” (p. 216). There are two sourced differences. (a) Navajo nominals do not carry the mark. The verb carries it, and the nominals are linked by order (INV-6). Passamaquoddy marks the obviative on the noun too (muwinuw-ol ‘bear-Obv’). (b) Navajo's possessor use is optional and varies by speaker (p. 216). Bruening's possessor obviation is obligatory (p. 21).
  2. The Navajo fourth person is the opposite of an obviative on F3. The obviative is the lowest animate rank. The Navajo fourth is off the scale and patterns with 1 and 2 (definite, human). The two share only “marks a referent as not the one you'd default to” (F1). When Willie says the fourth person “marks obviation”, she means disjoint-reference marking, not a rank. On F3, F4 and F6 it does not behave like Bruening's obviative.
  3. The Greenlandic “fourth person” is the opposite of the Navajo one on F1. 3r requires coreference with a matrix subject. Navajo 4 needs no antecedent, and in (29) it means ‘that other person’. Lybech refuses the label for exactly this reason (p. 115 n. 4). BOUND-3R is a binding dependency and a kind of switch reference: participial vs contemporative = disjoint vs same subject. PERSON-4 is a person value.
  4. Slave (secondhand) looks more like a disjoint-from-subject object marker (F1) than like Navajo's free-standing person value. On Willie's report, Sarcee yi- is in between. We hold these as reported and draw no conclusion until Rice and Cook are opened (T11a).

For Grant's ranking-edge picture HYPOTHESIS: only RANK-OBV and Navajo yi-/bi- are instances of the edge (“a lower argument outranks the agent, so the marked form is used”). PERSON-4 is the boundary condition already found on the fourth-person page: take the argument off the scale. BOUND-3R has nothing to do with ranking. It is an event-to-event coreference device: is the participant in the dependent event the matrix event's subject? That is still at the fine event-tracking grain Grant cares about. But it is a different grain from argument ranking.

6. The Algonquian inverse and VPISH

6.1 A sourced inverse that is A-movement SOURCED

Bruening's analysis of Passamaquoddy is the V-P analysis that this site posits for Navajo bi- (inverse §5, number & scope §5). Arguments are first-Merged by role, with the subject in Spec,VoiceP and the object in Comp,V. In the inverse an optional [EPP] on Voice moves the object to the highest A-position (A1), past the subject. His evidence is three diagnostics.

DiagnosticDirectInverseExample
D-BIND: an object quantifier binds a pronoun inside the subject✗✓(33) * vs (36a–c)
D-WCO: a questioned object binds a pronoun in the subject✗ (weak crossover)✓(34b) * vs (37a, b)
D-AMB: is scope ambiguous?rigid (subject > object)ambiguous (by reconstruction of the raised object)(39a)

Bruening's argument against base-generating the reversed order is D-AMB. “NPs may not lower where they have not previously raised” (p. 15). He also shows that word order is not the variable (§3.3).

6.2 What carries over to Navajo, and what does not HYPOTHESIS

  1. The prediction table for T10a gets sharper. For Navajo (58) tʼáá ʼałtso ʼatʼééké ʼashkii łaʼ biztsʼǫs vs (59) … yiztsʼǫs (number & scope §2.3), Bruening's pattern lets us state the three hypotheses as predictions:
    (59) yi-: readings(58) bi-: readings
    V-P as in Passamaquoddy (raising + reconstruction)girls > boy (the girls are ρA and raise; nothing crosses)ambiguous: girls > boy (raised ρP), and boy > girls by reconstruction
    Inverse linking (base-generate the reversed order; Bruening's rejected option)girls > boyrigid: girls > boy only
    V-H (Hale: nothing moves; ρA in Spec,v > ρP in Comp,V in both voices)girls > boyrigid: boy > girls only (the boy is ρA)
    All three agree on (59). The informative cell is (58): two readings (V-P), only girls > boy (linking), or only boy > girls (V-H). As on the number & scope page, there is a caveat. The internal nominal is an indefinite (łaʼ), and in many languages indefinites can take wide scope from anywhere. A clean test needs a non-indefinite internal quantifier. No source prints one, and none is invented here (T11c).
  2. D-BIND and D-WCO cannot be built in Navajo with the rules already on the site. Both need a pronoun (a possessor) inside one argument's nominal to covary with the other argument. In Passamaquoddy that is unproblematic (Condition C “simply appears not to apply”, p. 7; (14), (36b)). In Navajo, W&J's statement (71) excludes coreference between a possessor inside a complex NP and the pronominal arguments of the following verb when all are third person ((70b) *). The rescue is to make the possessor and the object fourth person ((77)). A fourth-person argument cannot enter yi-/bi- (4-NO-INV). So the rescue takes away the voice contrast that the test was meant to probe. If (71) also covers bound-variable readings, Navajo has a structural blind spot. The two binding diagnostics that show the Algonquian inverse is syntactic cannot be constructed for the Navajo inverse, and only D-AMB (scope) remains. That is one reason the site's VPISH question keeps landing on T10a. OPEN: does (71) cover bound variables (‘each girl … her mother’) as well as coreference with a name (T11b)? W&J state it for coreference, and that (71) is “strongly dispreferred” rather than impossible for some speakers is also W&J's.
  3. What Navajo adds that Passamaquoddy lacks. In Passamaquoddy the obviative mark on the noun lets nominals appear in any order (Bruening 15–16). In Navajo the nominals are unmarked, so order is the linking cue (INV-6, rule set B). That fits the site's split: the voice is decided inside the word (rule set A / Σ), and the nominals are placed by rule set B. Passamaquoddy shows the same A-position hierarchy with free order. HYPOTHESIS: the A/B changeover in Navajo is where the morphology hands the linking job to word order, because the nominals carry no rank mark.
  4. Possessors. Bruening: possessors are always proximate and c-command the possessed noun, so there is no inverse inside an NP (p. 21). Willie's (28a)/(28b) show a yi-/bi- contrast on a possessor, which she reads as obviation, not voice (p. 216). OPEN (T11d): in (28a)/(28b), is the b-/y- choice decided by c-command (possessor vs a clause-mate), as RANK-OBV would predict, or by reference to John vs the kissee? Willie's translations ‘his’/‘her’ do not say whose.
  5. GRANT'S CONJECTURE + HYPOTHESIS Roles, not lexical items. Bruening's Passamaquoddy finals “agree with the object in its low position”. The TA/TI final tracks the logical object's animacy in both voices (pp. 5–6). That is the same pattern as Navajo da- tracking ρA across the inverse (DIST-1). In both languages one affix reads the first-Merge (role) slot, and another reads the derived (prominence) position. Two unrelated families each split “role-indexed” from “position-indexed” morphology inside the verb. That supports the site's claim that the verb's arguments are role slots, which movement can re-rank without changing them.

7. Derivations X1–X4, step by step

Conventions as on the derivations page. Σ schema step, A word-internal linearization, B clause level. Rule sets A and B are this site's analysis of Navajo. For Passamaquoddy and Greenlandic we write only Merge / Agree / Move / spell-out in the source author's terms, and we make no claim about headedness in those languages. Proto-role entailments are ours (HYPOTHESIS), from the translations.

X1. Passamaquoddy (36b) Yatte wen pilsqehsis ʼ-kis-cem-ku-l w-ikuwoss-ol ‘Her1 mother kissed each girl1.’ (Bruening's analysis)

  1. Σ (our notation) Σ(√KISS) = {ρA, ρP}. ρA = the mother (vol, sent, mov), ρP = each girl. These are the same entailments the site gives Navajo ‘kiss’ (dictionary).
  2. Merge V with the object (ρP) in A3 = Comp,V. Agree the final (TA, animate) with the object in its low position (Bruening 6).
  3. Merge Voice. Merge the subject (ρA, ‘her mother’) in A2 = Spec,VoiceP.
  4. Merge Infl. Voice carries the optional [EPP] (Bruening 21), so this derivation is the inverse.
  5. Move the object from A3 to A1 (Spec,IP), crossing the subject in A2 (Bruening (12), (38)). The copy stays in A3.
  6. Obviation (RANK-OBV). The object in A1 c-commands the subject, so object = proximate and subject = obviative (w-ikuwoss-ol ‘3-mother-Obv’).
  7. Binding. The quantifier in A1 c-commands the possessor ‘her’ inside A2 and binds it. ✓ In the direct (33) the object stays below the subject, and the binding fails.
  8. Spell-out. The prefix ’- (3) indexes A1, the highest argument. -ku = Inv. -l = Obv (Bruening 17: “the prefix always indexes the highest argument, and the final suffix the lowest”).
  9. HYPOTHESIS Navajo counterpart. Steps 1–5 are the site's V-P derivation of bi- (inverse I1; number & scope N2, N4) with A1 = Spec,M/SP. The step with no Navajo counterpart is 7, because of the §6.2 blind spot.

X2. Passamaquoddy (39a) ‘Nothing1 bothers the one who owns it1.’: how the inverse becomes ambiguous

  1. Merge the object (the relative ‘the one who owns it’) in A3. Merge the subject (keq ‘what’ under negation, i.e. ‘nothing’) in A2.
  2. [EPP] on Voice → Move the object to A1 (inverse -ku-). There are now two copies of the object: A1 and A3.
  3. LF, reading 1. Interpret the object in A1, so it scopes over the subject. (This is the reading of (36).)
  4. LF, reading 2. Reconstruct the object to A3. Now the subject in A2 c-commands it, and ‘nothing’ binds ‘it’ inside the object. This is the reading in Bruening's translation of (39a).
  5. HYPOTHESIS Navajo counterpart. In (58), V-P puts ρP (‘all the girls’) in Spec,M/SP with a copy in Comp,V. If Navajo is like Passamaquoddy, both readings of (58) should be available. If only one is, that is evidence against V-P-with-reconstruction for Navajo (§6.2 table). T10a / T11c.

X3. Navajo (29) John háísh haghan=di yiztsʼǫs ‘Who did John kiss at that other person's house?’ (Willie 217)

  1. Σ Σ(√KISS) = {ρA, ρP}. ρA = John, ρP = háísh ‘who’. Both are 3rd person. Rank is equal (human/human), so INV-3 leaves voice free. EXT = ρA, and INV-2 gives yi-.
  2. Σ The possessor slot of ‘home’ is valued π = 4. By PERSON-4 this needs no antecedent and is not ranked. The reading is ‘that other person’, distinct from both clause arguments in Willie's translation.
  3. Merge the possessor with the noun. A The prefix precedes the stem: h- + ghan, spelled haghan (the segmentation follows Willie's 4POSS:home only as far as the gloss allows).
  4. Merge the postposition =di ‘at’ with its complement. B The enclitic follows its complement (head-final), so we get haghan=di. HYPOTHESIS One PP shows both rule sets: the possessor prefix is A, and the postposition is B. Whether =di is inside or outside the word boundary (B1–B3, headedness §3) is not settled by Willie's “=”.
  5. Merge the verb word yiztsʼǫs (3.3.kissed) as in N4, with V-H or V-P raising of ρA. Nothing crosses.
  6. B Order: John (EXT-linked) · háísh (internal-linked wh, in situ before the PP) · PP · V. Under INV-6 the noun next to the verb links to the internal slot. Here a PP intervenes, and Willie gives the order without comment. (How INV-6 counts across an adjunct PP is OPEN.)
  7. HYPOTHESIS Contrast with (28a)/(28b). There the possessor is 3rd person, and b-/y- splits the 3rd-person possessor into two values, which is Willie's obviation use of the voice markers. In (29), π = 4 takes the possessor out of the third-person system entirely. This is the possessor-level version of the fourth-person page's two responses (EDGE vs EXIT): re-rank within the 3rd person (b-/y-, like RANK-OBV) vs exit it (h-, PERSON-4).

X4. West Greenlandic Piitaq oqarpoq kaffisorluni (he = Peter) vs … kaffisortoq (he ≠ Peter) (Lybech 156)

  1. Σ (our notation) Matrix Σ(√SAY) = {ρA, ρTh}, where ρTh is the reported clause. Embedded Σ(√DRINK.COFFEE) = {ρS}, a single argument (Lybech's “Subject”, pp. 115–116).
  2. Merge the embedded clause and select its mood. Contemporative {(l)lu} or participial {ðu} (Lybech 156: the participial follows verbs of “utterance, perception etc.”, and both are possible here).
  3. Agree / bind. With the contemporative, the embedded subject must be coreferent with the matrix subject and is valued 3r. With the participial, it must not be coreferent and is valued 3p (Lybech 156).
  4. Merge the matrix verb oqar- ‘say’ with the embedded clause. Merge Piitaq. The indicative ending -poq is 3p.sg. There is no 3r in the indicative (Lybech 176).
  5. Spell-out. kaffisor-luni = CONT.3R.SG; kaffisor-toq = PTCP.3SG (our segmentation, Figure 9.3).
  6. HYPOTHESIS Navajo counterpart: none. Nothing in X4 involves ranking two co-arguments or moving one over the other. The dependency is between a matrix argument and an embedded one, so it is a binding or switch-reference relation. The Navajo fourth is not anchored to any antecedent (X3, step 2). Navajo's own cross-clause disjoint reference with third vs fourth person (Willie Chs. 5–6) is now partly read: see the =go chains page §7, which finds the reverse of BOUND-3R (T8b). That is where a real comparison with BOUND-3R would have to be made.

What X1–X4 show HYPOTHESIS

  1. “Fourth person” names two near-opposite devices. Navajo's is an unranked, antecedent-free person value. Greenlandic's is a third person bound by the matrix subject.
  2. Algonquian obviation is a ranking read off c-command. Navajo's counterpart to it is yi-/bi-, not ji-/ho-.
  3. The Passamaquoddy inverse is a sourced instance of VPISH-type raising, and it is diagnosed by binding, weak crossover and scope ambiguity. In Navajo only the scope test can be built (if W&J (71) covers bound variables), so T10a is the site's decisive VPISH test.

8. Verdict

  • Main finding. SOURCED (each cell of §4) + HYPOTHESIS (the grouping). The labels hide three jobs. Algonquian obviation ranks two third persons by structural height. The Navajo fourth person takes a human referent off the ranking scale. The Greenlandic “fourth” (3r) marks a third person bound by the matrix subject. Navajo's counterpart to Algonquian obviation is the yi-/bi- voice, as Willie herself notes (p. 96). It is not the fourth person.
  • For VPISH. SOURCED Bruening 2005 shows that an inverse can be A-movement of the object over the subject, with reconstruction. HYPOTHESIS That turns T10a into a three-way prediction (ambiguous / rigid girls-first / rigid boy-first for (58)). It also shows why the Navajo test has to be a scope test: the binding diagnostics probably cannot be built (§6.2).
  • For “theta roles, not lexical items”. HYPOTHESIS In both families, one verb affix reads the role slot across the inverse (Navajo da-; the Passamaquoddy final) and another reads the derived position. That is the same split in two unrelated families.
  • Athabaskan. OPEN Slave and Sarcee are secondhand only, and no Apache source was opened. The cognate claim is Willie's.

9. Open questions this page adds

#QuestionWhat would settle it
T11aSlave and Sarcee “fourth person”: function, shape and cognacy. Which Sarcee prefix (yi-, tsʼi) corresponds to which Navajo prefix? Is Slave's form an obviative (RANK-OBV) or a disjoint-reference person (PERSON-4)?Rice 1989 A Grammar of Slave; Cook 1984 (Sarcee); for Apache, Sandoval & Jelinek on Jicarilla yi-/bi-. OPEN
T11bDoes W&J's (71) ban cover bound-variable possessors (‘each girl … her mother’) as well as coreference with a name? If not, D-BIND can be built for Navajo after all.Speaker judgments; W&J 2000 §11.4.6 re-read; Willie Ch. 6. OPEN
T11c(= T10a, sharpened by §6.2) Is (58) bi- scopally ambiguous and (59) yi- rigid, as in Passamaquoddy? A version with a non-indefinite internal quantifier is needed (schematic only; no source prints one).Speaker judgments; Perkins 1978 (not obtainable online; listed in Fernald's bibliography). OPEN
T11dIn Willie (28a)/(28b), is the possessor's b-/y- fixed by c-command (RANK-OBV) or by which clause participant it refers to? Whose house is it in each?Willie Ch. 7 (possessor raising and yi-/bi-); speakers. Partly answered 2026-10-11: Willie Ch. 7 (p. 185, (5)) states a linear rule for possessors: bi- = not coindexed with the preceding nominal, yi- = coindexed. Applied to (28), with háísh as the preceding nominal: baghan = John's, yighan = the kissee's (our application, HYPOTHESIS; possessor raising §7). Linear vs c-command is still not separated. OPEN
T11eDoes Navajo have a cross-clause same/different-subject device comparable to BOUND-3R (e.g. 3rd vs 4th in adjoined clauses)?Answered for =go clauses 2026-10-09 (evening): no. The Navajo =go chain marks disjoint reference with π = 4, and coreference is the default, which is the reverse of BOUND-3R (=go chains §7). Relative clauses (Ch. 6) are still open.
T11fInuit/Yupik beyond West Greenlandic: is Yupik's “fourth person” the same BOUND-3R device? Is 3r restricted by animacy (F5)?A Central Alaskan Yupik grammar (not opened); Lybech does not say. OPEN

Builds on the fourth-person page (T8), the inverse page (T5) and the number & scope page (T10). Dictionary entry updated: kiss (Willie (28)–(29)).

Two clauses, one cast: =go chains and the fourth person

Working note. Kept as written (URL, data and status labels unchanged). Filed under The fourth person; the consolidated argument is the essay fourth-person. Rule IDs: rules index.

Verbs on this page → dictionary: yiztał ‘kick’ · néidííłtsʼin ‘hit with the fist’ · jizloh ‘rope’ · neezdá ‘sit down; make sit; keep (somewhere)’ · neezdá ‘sat down; with yikáá’ ‘on it’: sat down on it’

Restated on the unified structure (2026-11-15). The =go clause is restated as a CP with =go in C⁰, adjoined at the left of the main clause at the TP level or higher (H47), with participant pairing H45 (core NOT UNDERSTOOD); reported speech as two CPs, H40a against H40b (OPEN). 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. 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, §4, =go and reported-speech rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

Willie (1991, Ch. 5 §5.1, pp. 132–140) shows how Navajo keeps track of who is who across a “when/if” clause marked with the enclitic =go and the clause that follows it. Read with this site's rules, her data give three results. (1) When both clauses are transitive and every argument is third person, the second clause must reuse the same two referents. Only their roles may change, and the yi-/bi- voice is what lets them change. (2) The fourth person is the only way to bring in a new participant, and it works only from the first clause and only for the agent. The patient has to stay the same unless the new one is indefinite. (3) This is a device for marking disjoint reference, which is the reverse of the West Greenlandic “3r”, a form that marks coreference (T11e). There is one fidelity problem: the printed indices in Willie's (9) do not match her prose. It is left open (T12a), not fixed.

On this page: 1 Question · 2 Source · 3 Data · 4 Rules · 5 Referents fixed, roles free · 6 Derivations G1–G5 · 7 Compared with BOUND-3R · 8 Open

1. The question

Two items were left open. T8b on the fourth-person page asks how the fourth person marks disjoint reference across adjoined clauses, and how that fits W&J's obligatory coreference across all-third-person transitive clauses. T11e on the fourth person compared page asks whether Navajo has a cross-clause device comparable to the Greenlandic BOUND-3R. This page covers Willie's first construction type, the adjoined temporal/conditional =go clause (§5.1). Event vs person focus (§5.2) and reported speech (§5.3) are on the reported speech page (added 2026-10-10). Relative clauses (Ch. 6) are not covered yet.

The reason this matters for Grant's second phenomenon (theta roles rather than lexical items as arguments) is that the constraint Willie states is about referents, while the voice and person prefixes reassign roles. If both clauses must share their cast but may swap who does what, then what the second verb word chooses is a mapping of roles onto a cast that is already fixed. It is not choosing new nominals.

2. Source opened this run (2026-10-09, evening)

  • Willie, MaryAnn 1991. Navajo Pronouns and Obviation. PhD dissertation, University of Arizona (the scan already on the box). We read page images of printed pp. 132–140 (PDF pp. 138–146; for this chapter PDF = printed + 6). The Navajo lines of (7)–(14) were checked on 250 dpi crops. The OCR text layer was used only to find pages. It garbles the barred ł and the tones, so no form here is taken from it.
  • Willie credits “parallel processing” to Platero 1976, Speas 1990 and Hale, Jelinek & Willie 1991 (p. 135). None of these was opened. The principle is quoted only as Willie states it.

Transcription and glosses. Navajo is given as printed (glottal stop as ’, as is usual on this site). Willie's glosses are copied exactly where she gives them. Her ordering is object.subject: compare (13) jizloh “3.4.roped” ‘that person roped him’ with (14) hazloh “4.3.roped”. That ordering is our reading of the pair, and it matches the site's earlier rows, where haztał is “4.3.kicked”. Examples that she prints without a gloss line are shown without one. We do not add a Leipzig line that she does not print. Subscript indices are hers.

3. Data SOURCED

3.1 Intransitive pairs: coreference is presumed but not forced (pp. 133–134)

Willie 1991:133–134, (2)–(5)
(2)yah ʼííyáa=godah neezdá
3.came in=when3.sat down
‘When hei came in, hei,j sat down.’
(3)díníyáa=goshídóʼ dooleeł
2-go =if1=too will
‘If you go, I will go too.’
(3′)ʼashkii yah ʼííyáa=godah neezdá
boy 3.came in=when3.sat down
‘When the boyi came in, hei,j sat down.’
(4)yah ʼííyáa=goʼatʼééd dah neezdá
‘When he came in, the girl sat down.’
(5)yah ʼííyáa=godashneezdá
3-came in=when4-sat down
‘When hei came in, hej sat down.’

The source prints two examples numbered “3)” (pp. 133 and 134). We call the second one (3′). Willie on (2): “a strong presumption of coreference.” On (5): “Thus the fourth person pronominal has the same syntactic function as the nominal in (4).” (p. 134)

3.2 Transitive pairs: parallel processing (pp. 135–136)

Willie 1991:135, (6): “Parallel Processing. 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.” Also: “With clauses linked by the enclitic =go, coreference is obligatory unless disjoint reference is overtly marked.”

Willie 1991:135–136, (7)–(9)
(7)yiztał=gonéidííłtsʼin
3.3.kicked=COMP3.3.fist-hit
‘As hei kicked himj, hei hit himj.’
(8)biztał=gonábiidííłtsʼin
INV.3.3.kicked=COMPINV3.3.fist-hit
‘As hei was kicked by himj, hei was hit by himj.’
(9a)yiztał=gonábidííłtsʼin
‘As hei kicked himj, hej was hit by himi.’
(9b)biztał=gonéidííłtsʼin
‘As hei was kicked by himj, hej hit himi (back).’

P. 136: “The yi/bi- alternation can be used to mark a change in thematic roles across clauses.” In (9a) and (9b) the source underlines bi and éi in the second verb. The inverse fist-hit is printed nábiidííłtsʼin in (8) and nábidííłtsʼin in (9a). We copy both spellings as printed.

3.3 The fourth person in transitive pairs (pp. 136–140)

Willie 1991:137–140, (10)–(16)
(10)*yiztał=gonáhodííłtsʼin
(11a)jiztał=gonáhodííłtsʼin
‘As that personi kicked himj, hej hit that personi.’
(11b)haztał=gonízhdííłtsʼin
‘As hei kicked that personj, that personj hit himi.’
(12)*yiztał=goʼatʼéédnéidííłtsʼin
(13)jizloh=goʼáádóóyiztał
3.4.roped=COMPCONJ then3.3.kicked
‘That personi roped himj, and then hek kicked himj.’
(14)*hazloh=goʼáádóóyiztał
4.3.roped=COMPCONJ3.3.kicked
(16a)*yiztał=go ʼáádóóʼatʼéédnéidííłtsin
(16b)yiztał=go ʼáádóóʼatʼééd léiʼnéidííłtsin
indf.
‘Hei kicked himj and then hei hit a girl.’

(16a, b) print the final verb as néidííłtsin, without the ’ that appears in (7), (9b) and (12). We copy it as printed. Willie's rule (15), p. 139, is quoted verbatim:

“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) Also: “Here fourth person marks a change in thematic roles across these clauses, but cannot introduce a new referent in the second clause, since parallel processing strictly applies.” (p. 137, on (11))

The exception (pp. 139–140): disjoint patients are possible “if the patient of the second clause is less definite than that of the first” ((16b), with léiʼ glossed “indf.”).

4. Rules for [C1=go C2]

Notation as on the schema page and the inverse page. C1 is the clause marked with =go, and C2 is the clause after it. cast(C) is the set of referents of C's argument slots. π(ρ) is the person value of a slot. EXT(C) is the external argument (subject) of C. Each rule restates one of Willie's statements. Where the restatement adds something, the added part is labelled.

GO-INTR   C1, C2 intransitive:  cast(C2) = cast(C1) by default, not obligatorily.
          Disjointness is signalled by a nominal after C1 (4) or by π = 4 in C2 (5).
GO-PP     C1, C2 transitive, every slot π = 3:  cast(C2) = cast(C1).   (Willie (6), (15a))
GO-SWAP   Under GO-PP the role-to-referent mapping may differ between C1 and C2;
          the voice choice yi-/bi- in C2 signals the difference.        (Willie (15a), p. 136)
GO-NO-NEW If every slot of C1 is π = 3, C2 introduces no new referent:
          not by π = 4 (10), and not by a nominal adjunct (12).         (Willie (15b), p. 138)
GO-4-AGT  π(ρA of C1) = 4  ⇒  ρA(C2) may be disjoint from ρA(C1)  (13).
GO-PAT    ρP(C2) = ρP(C1)  in every transitive =go pair (13 vs *14),
          unless ρP(C2) is less definite than ρP(C1) (16b, léi’).        (Willie (15c), pp. 139–140)
GO-4-BOTH π = 4 in both clauses: same two referents; the 4th-person referent may be
          agent in one clause and patient in the other (11a, b).         (Willie p. 137)
RuleStatusBasis
GO-INTRSOURCEDWillie 1991:133–134, (2)–(5).
GO-PPSOURCED(6), (7), (8), (15a).
GO-SWAPSOURCED as Willie's prose; OPEN against her printed indices for (9)p. 136; (15a). See §5.2 and T12a.
GO-NO-NEWSOURCED(10), (12), (15b).
GO-4-AGTSOURCED(13), (15c).
GO-PATSOURCED. Whether it is stated over ρP (a role) or over the object (a position) is OPEN(13), *(14), (16a, b). In (13)/(14) the patient is the object in both clauses, so these examples cannot tell role from position (T12b).
GO-4-BOTHSOURCED(11a, b), p. 137.

5. Referents fixed, roles free

5.1 What GO-PP holds constant

HYPOTHESIS GO-PP and GO-NO-NEW constrain the cast of C2, and GO-SWAP leaves the role assignment free. In the site's terms, the second verb word does not choose arguments. Its Σ = {ρA, ρP} is filled from a cast the first clause has already fixed. What the word does choose, through yi-/bi- or π = 4, is which member of the cast goes into which slot. This is the strongest sourced evidence on the site so far that the units the verb word works with are role-slots, while the participants are supplied from discourse. It agrees with the site's earlier claim that nominals are adjuncts (C3, INV-6), and it is evidence for that claim. It does not prove it.

HYPOTHESIS The rules split the same way the number & scope page found. GO-4-AGT and GO-PAT name roles (Willie says “agents” and “patients”). GO-SWAP is about which referent is EXT. So one construction again carries both role-indexed and position-indexed requirements.

5.2 The indices in (9) OPEN

Update 2026-10-17. SOURCED Willie & Jelinek 2000:270, (49) print (9a)'s string, yiztał=go nábidííłtsʼin, as ‘When hei kicked himj, hei was hit by himj’, which is R2. Their tree marks the agent i as topic and the patient j as focus in C1, and the patient i as topic and the agent j as focus in C2. So the match is by topic/focus, not by role in the action. See clause spine §5.3 (FOC-ID), which also shows that GO-PAT could be focus-indexed rather than role- or position-indexed.

Willie's prose says that in (9) the voice switch marks “a change in thematic roles” (p. 136; (15a)). Read literally, her printed indices say something else:

  • (9a): C1 has i as kicker (agent). C2's translation “hej was hit by himi” keeps i as agent. The roles are unchanged, and only the subject changes from i to j.
  • (9b): C1 has j as kicker. C2's translation “hej hit himi” keeps j as agent. But the word “(back)” suggests the one who was kicked is now hitting, which would be a role switch.

So there are two readings. In R1 (indices taken literally), GO-PP matches referents by role, and yi-/bi- in C2 only changes which referent is the subject. In R2 (prose, (15a) and “back”), GO-PP matches referents by grammatical function (EXT matches EXT), and a voice change therefore reassigns the roles. Our two notes: R2 is the reading Willie herself states, and in R2 (9a) would mean ‘as hei kicked himj, hei was hit by himj’. We do not correct the source. This is a question for a speaker or for one of the works Willie cites (Platero 1976; Speas 1990; Hale, Jelinek & Willie 1991) (T12a).

HYPOTHESIS Bearing on VPISH. R2 does not tell V-P from V-H. Both have an EXT: under V-P it is the raised slot, and under V-H it is a case designation (inverse §5). What R2 does show is that parallel processing reads grammatical function and not theta role. Under V-P, that would make the subject position Spec,M/SP the cross-clause anchor. That is a testable link between the inverse and clause chaining. It is not a result.

6. Derivations G1–G5, step by step

Conventions as on the derivations page. Σ schema step, A word-internal linearization, B clause level. Verb words are not segmented beyond Willie's glosses. Entailments for ‘kick’ are the D3 sets (dictionary). Entailments for ‘fist-hit’ and ‘rope’ are ours, from the translations (HYPOTHESIS). We treat =go as a clause-final C head. That category is our assumption: Willie glosses it COMP in (7) and calls it a “complementizer enclitic” (p. 132).

G1. (7) yiztał=go néidííłtsʼin ‘As hei kicked himj, hei hit himj.’

  1. Σ C1: Σ(√KICK) = {ρA, ρP}, both π = 3. Rank is not given (no nominals), so INV-3 leaves voice free. Direct: EXT = ρA, spelled yi- (INV-2).
  2. Merge V + ρP (Comp,V). Merge v + ρA (Spec,v). Merge M/S. Agree M/S with ρA (π = 3). Move (V-P only): ρA to Spec,M/SP, crossing nothing. A Spell-out of the word with prefixes on the left: yiztał.
  3. Merge C = =go with the clause. B C follows its complement: yiztał=go. OPEN Whether =go is inside the verb word (A) or a clitic at clause level (B) is the same open boundary question as =di (X3 step 4).
  4. Σ C2: Σ(√FIST-HIT) = {ρA, ρP}, both π = 3. GO-PP: cast(C2) = cast(C1) = {i, j}. Direct voice, so EXT = ρA = i, and ρP = j. The steps of step 2 give néidííłtsʼin.
  5. B Adjoin C1 to C2. The adjunct comes first, giving [C1=go] C2. Result: (7), with the same roles in both clauses.

G2. (9a) yiztał=go nábidííłtsʼin, worked under R1 and R2

  1. C1 as in G1: EXT = ρA = i (kicker), ρP = j.
  2. Σ C2: cast = {i, j} (GO-PP). Voice = inverse, so EXT = ρP (INV-2, bi-).
  3. Move (V-P): ρP raises from Comp,V to Spec,M/SP, crossing ⟨ρA⟩ in Spec,v (as in inverse I1). V-H: no movement; EXT is a case designation.
  4. Cast assignment, R1 (indices as printed): match by role. ρA(C2) = i, ρP(C2) = j, so EXT(C2) = j. Translation: ‘hej was hit by himi’. Roles are unchanged and the subject has changed.
  5. Cast assignment, R2 (Willie's prose): match by function. EXT(C2) = EXT(C1) = i, so ρP(C2) = i and ρA(C2) = j. Translation: ‘hei was hit by himj’. The subject is unchanged and the roles have swapped.
  6. A Spell-out: nábidííłtsʼin (as printed in (9a)). B [C1=go] C2. The string is the same under R1 and R2. Only the cast assignment differs (T12a).

G3. (13) jizloh=go ʼáádóó yiztał ‘That personi roped himj, and then hek kicked himj.’

  1. Σ C1: Σ(√ROPE) = {ρA, ρP}. π(ρA) = 4, π(ρP) = 3 (Willie “3.4” = object.subject). 4-NO-INV: no yi-/bi- choice, so EXT = ρA.
  2. Merge, Agree (M/S with ρA, π = 4). Move (V-P) ρA to Spec,M/SP. A jizloh. B jizloh=go.
  3. B The conjunction ʼáádóó ‘and then’ follows C1 (“an optional feature of conditional constructions”, p. 138).
  4. Σ C2: Σ(√KICK) = {ρA, ρP}, both π = 3. GO-4-AGT: since π(ρA(C1)) = 4, ρA(C2) = k may be new. GO-PAT: ρP(C2) = j.
  5. The steps of G1 step 2 give yiztał. Result: three referents across two clauses, with the patient as the constant.

G4. *(14) hazloh=go ʼáádóó yiztał: where it crashes

  1. Σ C1: π(ρA) = 3, π(ρP) = 4 (“4.3.roped”). So the 4th person is on the patient, which marks it as distinct.
  2. C2: yiztał, both π = 3. GO-PAT requires ρP(C2) = ρP(C1).
  3. ✗ ρP(C1) has π = 4 and ρP(C2) has π = 3. Third and fourth “do not match” in person features (p. 137), so they cannot be the same referent. GO-PAT cannot be satisfied, and Willie stars the sentence. The word-level steps (Merge, Agree, spell-out) all go through. The failure is at the level of the cast, after both words are built.

G5. (5) yah ʼííyáa=go dashneezdá ‘When hei came in, hej sat down.’

  1. Σ C1: Σ = {ρ1}, π = 3. C2: Σ = {ρ1}, π = 4 (“4-sat down”).
  2. GO-INTR: the default cast(C2) = cast(C1) is only a presumption. π = 4 in C2 is a person value that does not match C1's π = 3, so the subjects are disjoint (i ≠ j).
  3. A dashneezdá beside dah neezdá (2): the 4th-person form is written as one word. We do not segment it further than Willie's gloss. B [C1=go] C2.
  4. HYPOTHESIS Compare *(10). The transitive C2 náhodííłtsʼin with a π = 4 slot is out after an all-3rd C1, but the intransitive (5) is fine. So GO-NO-NEW is a property of the transitive chain (GO-PP), not of =go as such.

7. Compared with BOUND-3R (T11e)

FeatureWest Greenlandic 3r (Lybech 2020:115, 156)Navajo =go chain (Willie 1991:132–140)
What the special form marksCoreference with the matrix subjectDisjoint reference (π = 4); coreference is the unmarked case
Where it appearsThe dependent clauseTransitive: in C1 (the =go clause), agent only (15c). Intransitive: in C2 (5).
AnchorMatrix subjectThe whole cast (GO-PP), with the patient as the constant (GO-PAT)
Is it a separate person value?No: same person (Lybech p. 115 n. 4)Yes: π = 4 does not match 3 (p. 137)

Qualified 2026-10-10. “Navajo marks other” holds for =go chains. Under ‘say’ two fourth persons mark the same person, and two third persons must be different (reported speech §6, PI-MATCH).

HYPOTHESIS The answer to T11e is no. The =go chain is not BOUND-3R. The two do the opposite jobs: Greenlandic marks same, and Navajo marks other. The Navajo device is not anchored to a subject. It is also an instance of PERSON-4 (compared §5), an unranked person value that, because it cannot match a third person, forces disjointness. This leaves the three-way split of T11 in place. The cross-clause behaviour is a consequence of PERSON-4 plus GO-PP, not a fourth job. Willie does not call this switch-reference on the pages read, and we do not either.

8. Open questions

IDQuestionWhat would settle it
T12aIn (9a, b), does the voice switch in C2 change the roles (R2, Willie's prose) or only the subject (R1, her printed indices)?A speaker judgment, or Platero 1976 / Speas 1990 / Hale, Jelinek & Willie 1991. OPEN 2026-10-10: Willie's summary (p. 149) says the arguments are “coindexed according to theta role”, which favours R1 (reported speech §8). 2026-10-10 (afternoon): in relatives the cast is shared with roles free ((29), (30)), and 4/4 coreference crosses a patient/agent split ((49), (52)) (relatives §11). 2026-10-17: Willie & Jelinek 2000:270 (49) print the same string yiztał=go nábidííłtsʼin with R2 indices (‘When hei kicked himj, hei was hit by himj’) and state that “the focus pronoun of the first clause is coindexed with the focus pronoun of the second” (p. 270). The conflict is now between two printings of one sentence; still OPEN (clause spine §5.3, T31b).
T12bIs GO-PAT stated over the patient role or over the object position? A test needs a pair in which C2 is inverse with a 4th-person agent in C1. 4-NO-INV may make that impossible.New data. Do not pre-fill. OPEN
T12cWhy can π = 4 introduce a new agent only from C1 (13), but a new intransitive subject from C2 (5)?Willie §5.2–5.4 and Ch. 6. Partly answered 2026-10-10: p. 149, a second-clause 4th person needs a first-clause 4th person as its antecedent; intransitive conditionals allow a 4th person or nominal in the consequent (p. 148). Why is not explained. OPEN
T12dDoes the definiteness exception (16b) also hold for a new agent, e.g. an indefinite nominal with léiʼ as the C2 agent?Not in the pages read. OPEN
T12eIs =go inside the verb word (rule set A) or a clause-level enclitic (B)?Same evidence as the =di question (headedness B1–B3). OPEN
T12fDo relative clauses (Ch. 6) and reported speech (§5.3) obey GO-PP, or is it specific to =go? Willie says it “applies … differently across construction type” (p. 135).Willie §5.3, Ch. 6. Answered for reported speech 2026-10-10: no. With all third persons the subjects must be disjoint (22), so GO-PP is specific to =go (reported speech §6). Answered for relatives 2026-10-10 (afternoon): yes. Willie (27): parallel processing “applies to sentences with relative clauses also” (p. 166) (relatives §5). So GO-PP holds in adjunct clauses (=go, relatives), not under ‘say’; ‘believe’ is free (§9).

Not read this run: Willie §5.2 (event vs person focus, from p. 140), §5.3 reported speech, and Ch. 6. §5.2–5.4 were read on 2026-10-10 (reported speech page). Ch. 6 is the next increment for T8b.

Same person or someone else? The fourth person in “he said …” sentences and in person focus

Working note. Kept as written (URL, data and status labels unchanged). Filed under The fourth person; the consolidated argument is the essay fourth-person. Rule IDs: rules index.

Verbs on this page → dictionary: jizloh ‘rope’ · dabiiłtsą́ ‘see’ · yi-sh-cha ‘cry’ · booyeéł ‘carry (an animate); be carried = ride (in the translations)’ · doogááł ‘arrive, will go / come (future)’ · ní ‘say’

Restated on the unified structure (2026-11-15). The =go clause is restated as a CP with =go in C⁰, adjoined at the left of the main clause at the TP level or higher (H47), with participant pairing H45 (core NOT UNDERSTOOD); reported speech as two CPs, H40a against H40b (OPEN). 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, §4, =go and reported-speech 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 finishes Willie's (1991) Chapter 5: §5.2 on event vs person focus (pp. 140–142), §5.3 on reported speech (pp. 143–148), and the chapter summary §5.4 (pp. 148–150). The intransitive quotative pair from §4.2.3 (p. 114) is added as background. Read with this site's rules, the pages give four results. (1) In a “he said that he will come” sentence with only third persons, the two “he”s must be different people (22). That is the reverse of the =go chains, where two third-person clauses must keep the same cast. (2) The fourth person does not mean “same” or “different” by itself. In both constructions a fourth person never picks up a third person, and two fourth persons pick up each other. What flips between the constructions is the plain third-person default. (3) The fourth person on the verb ‘say’ does two separate jobs. It is a respect form for a specific person, or a quotative (“it is said”) with no specific speaker. They share a form and are kept apart here. (4) In §5.2, putting a fourth person on the object of ‘see’ forces that slot to mean a person, never an event (p. 142). Here the person value of a role slot decides what kind of thing can fill it, with no noun doing the work.

On this page: 1 Question · 2 Source · 3 Event vs person focus · 4 Reported speech · 5 Rules · 6 What the fourth person matches · 7 Derivations S1–S6 · 8 Bearing on T12a, T12c, T12f · 9 Open

1. The question

The =go chains page left T12f open: does parallel processing (two all-third-person transitive clauses share their cast) hold outside =go clauses? It also left T12a open (do the indices in Willie's (9) match roles or grammatical functions?) and T12c (why a new fourth-person agent comes only from the first clause). Willie's remaining Chapter 5 sections bear on all three. For Grant's two phenomena, this page asks:

  • Theta roles as arguments (phenomenon 2). When ‘say’ embeds a clause, are the speaker slot of ‘say’ and the subject slot of the embedded verb linked by role, by person value, or by a nominal? And can a person value on a slot restrict what fills it?
  • Rule set A vs rule set B (phenomenon 1). Each verb word is built by prefixation (A). The embedded clause comes before the verb of speech, and the topic nominal comes before the whole sentence. Both orderings are clause level (B). Where is the changeover in these sentences?

2. Source opened this run (2026-10-10, ~5am PT)

  • Willie, MaryAnn 1991. Navajo Pronouns and Obviation. PhD dissertation, University of Arizona (the scan already on the box). We read page images of printed pp. 141–149 (PDF pp. 147–155; PDF = printed + 6 in this chapter) at 150 dpi. Every Navajo line of (17)–(30) was checked on 300 dpi crops. Printed pp. 113–114 (PDF 119–120) were read the same way, with (9)–(10) on a 300 dpi crop. The OCR text layer was used only to find pages. It garbles ł, the nasal hooks and the tones, so no form here is taken from it.

Transcription and glosses. Navajo is given as printed (glottal stop as ’). Willie's glosses are copied exactly, including one that looks wrong (see (23)). Her gloss order for transitive verbs is object.subject. (17) yiiłtsą́ “3.1.saw” translated ‘I saw (Mary)’ confirms this: the 1 is the seer. The =go chains page had this ordering only as our reading of jizloh/hazloh, and (17)–(18) now support it from the translations. Subscript indices are hers. Willie gives the translations and no separate Leipzig line. Her gloss line is the only interlinear line shown, and we add none.

Willie says she introduced the quotative “earlier in Chapter 3” (p. 143). In the scan, the section is numbered 4.2.3 (p. 114). We cite it as p. 114 §4.2.3.

3. Event vs person focus (§5.2, pp. 140–142) SOURCED

Willie 1991:140–141: “We have seen earlier that the fourth person is so highly referential that it excludes the adjunction of a simple nominal; like first and second person, fourth does not match in phi features with third. However, there is a complex sentence type where this generalization does not apply. This involves focus on a person rather than the event the person participates in.”

Willie 1991:141–142, (17)–(20)
(17)Maryyicha=goyiiłtsą́
3.crying=while3.1.saw
‘I saw Mary crying.’
(18)Maryjicha=gohoołtsą́
4.crying=while4.1.saw
‘I saw Mary when she was crying.’ (Mary, I saw her when she was crying.)
(19)Billłį́į́ʼbooyeéł=goyiiłtsą́
horse3.3.carrying=while3.1.saw
‘I saw Bill riding the horse.’
(20)Billłį́į́ʼhooyeéł=gohoołtsą́
horse4.3.carrying=while4.1.saw
‘I saw Bill when he was riding the horse.’

Willie, p. 141: “In (17) the speaker reports on an event he saw; in (18) he reports on seeing a person in a particular situation. Fourth person pronouns cannot be modified by nominal adjuncts; this means that the nominal Mary cannot be syntactically associated with the subject of the subordinate clause subject. But this nominal can be a topic for the whole complex sentence; this provides the contrast in readings shown.” (p. 141–142, wording as printed.) On (20), p. 142: “a switch from third to fourth person forces a reading on which the focus is on the person involved in the activity rather than on the activity as a whole. Note that in (20), the fourth person patient pronoun in the main verb can refer only to the person and never to an event.”

Print: in (19) the typescript has booyeéł with a damaged first “oo”, and the =go is spaced off from the verb in both (19) and (20). We write it attached, as in (17). The first e of booyeéł/hooyeéł is printed without a tone mark and the second with one. We copy that. Willie p. 113 also notes a respect context (“referring to his mother”) where a nominal can adjoin to a fourth person. That is a second exception, and we have not built on it.

4. Reported speech (§5.3, pp. 143–148) SOURCED

4.1 Background: the quotative (§4.2.3, p. 114)

Willie 1991:114, (9)–(10)
(9)yichajiní
3s.crying4s.said
‘They said he is crying.’
(10)jííchajiní
4s.cried4s.said
‘(They) said that person cried.’

P. 114: “In these two examples the fourth person in the main clause conveys a quotative. The speaker of these utterances is signaling that he does not have direct knowledge of the events, but is reporting on something he has heard about.” Pp. 115–116 contrast this “Reportative” with “a use of the fourth person in the verb jiní to indicate politeness or respect”, where the fourth person “refers to a specific person who is present or nearby, and is used in deference”.

4.2 Person combinations with ‘arrive’ under ‘say’ (pp. 143–147)

Willie 1991:143–147, (21)–(30)
(21)deesháałní
1.arrive3.said
a. ‘He said “I will be there”.’ b. ‘He said that I would be there.’
(22)doogááłní
3.arrive3.said
‘Hei said hej will be there.’
(23)deesháałjiní
1.arrive3.said [sic]
a. ‘That personi said that hei would go.’ b. *‘They say I will go.’ c. *‘That person says that I will go.’
(24)doogááłjiní
3.arrive4.said
a. ‘That personi said hej will come.’ b. ‘It is said that he will come.’
(25)jidoogááłní
4.arrive3.said
‘He said that person will come.’
(26)jidoogááłjiní
4.arrive4.said
a. ‘Hei said that hei would come.’ b. ‘They say that person will come.’
(28)díínááłní
2.arrive3.said
‘He said that you should go.’
(29)díínááłjiní
2.arrive4.said
a. ‘It is said that you should go.’ b. *‘That person said you should go.’
(30)díínááłnijiní
2.arrive2.4.said
‘That person said [of you] that you should go.’

[sic] in (23): the gloss line prints jiní as “3.said”. In (24), (26), (29) and in (9)–(10) on p. 114, the same word is glossed “4.said” / “4s.said”, and the prose on (23) says “the fourth person is used as a respect form”. We copy the printed gloss and mark it; we do not correct it.

Willie's comments, verbatim. On (21): “The two interpretations show quoted speech and indirect speech.” On (22): “The subjects of the main and complement clauses must be disjoint in reference when all third person arguments are employed.” On (23): “On the permitted reading of (23), the fourth person is used as a respect form. The quotative interpretation shown in (23b) is excluded since it is inappropriate to use a reportative concerning one's own actions. The reading in (23c) is excluded, since the possiblility of reference to the speaker of the whole complex utterance is not present when the subject of the verb of speech is fourth person, as in (21).” On (24): “both respect and quotative readings are permitted.” On (25): “Only the respect reading appears when the complement verb has a fourth person pronoun.” On (26): “The preferred reading for (26) is where the fourth person pronouns are coindexed, and used for respect. On the second reading, the fourth person on the main clause verb is a quotative.” On (28): “The use of the second person in the complement clause adds a modal feature (obligation) to the reading.” On (30): “the hearer has not necessarily been informed, directly or recently, that he should go; but a person referred to by a respect fourth person expects the hearer to go.” (pp. 143–147; spellings as printed.)

Willie's summary (27), p. 146, verbatim: “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, coference between the subjects of the main and complement clauses is signaled; d. Fourth person on the verb of speech can signal a quotative.” Then: “It is evident that conditions (27c) and (27d) overlap; this is the source of the ambiguity seen in the two readings of (26).”

P. 149: “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.”

5. Rules

Notation as on the =go chains page. M is the matrix clause headed by the verb of speech, and K is its complement clause. π(ρ) is the person value of a slot. EXT is the subject. “Quotative” and “respect” are Willie's terms. Each rule restates one of her statements. Where the restatement adds something, the added part is labelled.

RS-1      π(EXT K) = 1, π(EXT M) = 3:  EXT K = EXT M ("quoted") or EXT K = speaker of
          the whole utterance ("indirect").                        (Willie (21), (27a))
RS-33     π(EXT K) = π(EXT M) = 3:  EXT K ≠ EXT M, obligatorily.     (Willie (22), (27b))
RS-44     π(EXT K) = π(EXT M) = 4:  EXT K = EXT M is signalled (preferred reading).
                                                                    (Willie (26a), (27c))
RS-Q      π(EXT M) = 4 may be a QUOTATIVE: no specific speaker ("it is said").
                                                                    (Willie (24b), (26b), (29a), (27d); p. 114)
RS-RESP   π(EXT M) = 4 may be a RESPECT form: a specific person, present or nearby.
                                                                    (Willie pp. 115–116, 144)
RS-4×1    π(EXT M) = 4, π(EXT K) = 1:  respect only. No quotative (23b), and the
          "speaker of the utterance" reading of 1 is lost (23c).    (Willie p. 144)
RS-4×3    π = 4 on one verb, 3 on the other:  the two subjects are disjoint.
          With 4 on M: respect or quotative (24). With 4 on K only: respect only (25).
RS-2      π(EXT K) = 2 adds obligation (28). With π(EXT M) = 4 only the quotative
          survives (29a, *29b). The respect reading needs 2 on M as well: nijiní
          "2.4.said" (30).                                           (Willie pp. 146–147)
FOCUS-4   In [K=go] M, if π(EXT K) = 4, a 3rd-person nominal cannot link to EXT K; it is
          a topic of the whole sentence, and the reading is person focus (18), (20).
          A π = 4 patient of M "can refer only to the person and never to an event" (20).
                                                                    (Willie pp. 141–142)
RuleStatusBasis
RS-1, RS-33, RS-44, RS-QSOURCED(21)–(26); Willie's summary (27a–d).
RS-RESPSOURCEDpp. 115–116, 144–145.
RS-4×1SOURCED(23) and the prose on it. The (23) gloss “3.said” is a misprint on our reading (§4.2).
RS-4×3SOURCED for the readings. “The two subjects are disjoint” for (25) is our HYPOTHESIS(24a) has printed indices i ≠ j. (25) has no indices: ‘He said that person will come’ reads as disjoint, but Willie does not say so.
RS-2SOURCED(28)–(30).
FOCUS-4SOURCED(17)–(20), pp. 141–142. The “topic of the whole sentence” attachment is Willie's description. The tree is ours (S1).

6. What the fourth person matches HYPOTHESIS

Put the =go results next to the reported-speech results:

Person values (clause 1, clause 2)=go chain (Willie §5.1)‘say’ + complement (Willie §5.3)
3, 3Same cast, obligatory (GO-PP, (6)–(8))Different subjects, obligatory (RS-33, (22))
4, 4Same two referents (GO-4-BOTH, (11a, b))Same subject, preferred (RS-44, (26a))
3 and 4 mixedDifferent (GO-4-AGT (13); GO-NO-NEW (10))Different (RS-4×3, (24a), (25))

Update 2026-10-10 (evening): part (i) below does not hold for the optative. Under ‘think’, an optative 4 subject is the 3rd-person thinker (Willie (18)). See fourth person uses §11.

PI-MATCH (ours). Across both constructions: (i) a π = 4 slot never takes the referent of a π = 3 slot; (ii) two π = 4 slots take the same referent by default; (iii) what two π = 3 slots do is set by the construction: the same referent in a =go chain, different referents under ‘say’. Parts (i) and (ii) are Willie's “fourth does not match in phi features with third” (pp. 137, 141) applied to both tables. Part (iii) is the new point. The fourth person is not a “disjoint-reference marker” or a “coreference marker” as such. It only flips the result against whatever the plain third person does in that construction. This sharpens =go chains §7, where Navajo was said to mark “other” where Greenlandic marks “same”. That holds for =go. Under ‘say’, the Navajo 4/4 pair marks same, and the Greenlandic contrast is weaker than §7 said.

Not merged: respect and quotative. RS-RESP and RS-Q are two jobs that share one form. The respect 4 refers to a specific person (pp. 115–116). The quotative 4 refers to no one in particular (“It is said …”, (24b), (29a)). In schema terms (ours), the respect 4 is ρA of ‘say’ valued to a definite deferred-to person, as in PERSON-4 (compared §5). The quotative 4 is ρA left without a specific referent. That makes it closer to an impersonal or reportative than to a person. Willie herself lists them separately, together with the Impersonal (§4.2.5; since read, see the fourth person uses page). Whether the two are one morpheme with two uses or two homophonous morphemes is OPEN (T13a).

A slot's person value restricts its filler (FOCUS-4). In (17), the object slot of ‘see’ is π = 3 and the sentence reports “an event he saw”. In (18) and (20) the object slot is π = 4 and “can refer only to the person and never to an event” (p. 142). Our reading: Σ(√SEE)'s ρP can be linked to the =go event or to a person. π = 4 (definite, human: PERSON-4) removes the event option. What decides “I saw the crying” vs “I saw Mary” is the person value on the slot. No noun does it: Mary is in both strings, in the same place. This is a second piece of evidence, after the cast-fixing in =go chains, for Grant's claim that the verb works with role-slots and their features rather than with lexical items.

The extra slot on ‘say’ in (30). With a respect-4 speaker and a 2nd-person complement subject, the plain jiní is out ((29b)). The good form nijiní “2.4.said” puts a 2nd-person value on ‘say’ itself. Our reading: ‘say’ gets a second role slot (Willie's translation: “[of you]”) whose referent is the complement's subject. Which role that slot is (addressee, or the one spoken about) is OPEN (T13b). It is a verb gaining a role-slot to keep person values in step across clauses. English does this with a noun phrase. It is not data about English.

7. Derivations S1–S6, step by step

Conventions as on the derivations page. Σ schema step, A word-internal linearization (prefixes to the left of the stem), B clause level (complements and adjuncts to the left of their head). Verb words are not segmented beyond Willie's glosses. 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). The complement of ‘say’ is treated as the clausal argument of ‘say’ (ρTh, as in the Greenlandic X-derivations). That is our assumption: Willie says only that these clauses have no determiner or nominalizer enclitic (p. 149).

S1. (17) vs (18): one string shape, two attachments

  1. Σ K: Σ(√CRY) = {ρ1}. In (17) π = 3 (yicha). In (18) π = 4 (jicha).
  2. Merge v + ρ1; Merge M/S; Agree M/S with ρ1. Move (V-P) ρ1 to Spec,M/SP. A Spell-out: yicha / jicha.
  3. Merge C = =go. B yicha=go / jicha=go.
  4. Σ M: Σ(√SEE) = {ρA, ρP}, π(ρA) = 1. In (17) π(ρP) = 3 (yiiłtsą́ “3.1”); in (18) π(ρP) = 4 (hoołtsą́ “4.1”). Same Merge/Agree steps; A yiiłtsą́ / hoołtsą́.
  5. B Adjoin [K=go] to M, to its left.
  6. Link Mary. (17): Mary is a linked adjunct of ρ1 of K (π = 3 matches a 3rd-person nominal). It is attached inside the =go clause, at its left edge (B). (18): ρ1 of K is π = 4, which a 3rd-person nominal cannot match (FOCUS-4). Mary is adjoined to the whole sentence as a topic (B, higher), and is read as the referent that both π = 4 slots pick up (PI-MATCH ii).
  7. HYPOTHESIS Result: the same left-to-right string Mary + [K=go] + M, with Mary attached at two heights. This is a pure rule-set-B difference. Nothing inside the words differs except the person values. Under V-P and V-H alike, nothing word-internal moves differently. So (17)/(18) do not decide VPISH. They show that the changeover point A → B is the same in both: the edge of each verb word.

S2. (20) Bill łį́į́ʼ hooyeéł=go hoołtsą́: person focus with a transitive K

  1. Σ K: Σ(√CARRY) = {ρA, ρP}. Gloss “4.3” = object 4, subject 3. Our reading (HYPOTHESIS): ρA = the horse (carrier, π = 3), ρP = Bill (carried, π = 4). So English ‘ride’ corresponds to a Navajo verb on which the rider is the patient.
  2. Merge V + ρP, v + ρA, M/S; Agree; Move (V-P) ρA. 4-NO-INV: no yi-/bi- choice. A hooyeéł. B łį́į́ʼ (linked to π = 3 ρA) precedes it inside K: łį́į́ʼ hooyeéł=go.
  3. Σ M: Σ(√SEE), π(ρA) = 1, π(ρP) = 4: hoołtsą́.
  4. PI-MATCH (ii): the two π = 4 slots (carried, seen) take one referent. GO-PAT (patients corefer across =go) is also satisfied: both π = 4 slots are patients. Bill cannot link to either π = 4 slot (FOCUS-4) and is the sentence topic. B Bill [łį́į́ʼ hooyeéł=go] hoołtsą́.
  5. FOCUS-4: ρP of ‘see’ is a person, not the riding event.
  6. OPEN (19) booyeéł “3.3” begins with b-. If that is the inverse bi-, then (19) is Bill (ρP) before łį́į́ʼ (ρA) with the inverse, which fits the order rule on the inverse page. But Willie's gloss has no INV, unlike (8) “INV.3.3.kicked” (T13c).

S3. (22) doogááł ní ‘Hei said hej will be there.’

  1. Σ K: Σ(√ARRIVE) = {ρ1}, π = 3. Merge, Agree, Move (V-P). A doogááł.
  2. Σ M: Σ(√SAY) = {ρA, ρTh}. π(ρA) = 3. Merge V + K as ρTh. No C enclitic is spelled (p. 149). Merge v + ρA; Agree; Move (V-P). A ní.
  3. B The complement precedes the verb of speech: doogááł ní.
  4. Reference: RS-33. ρA(M) ≠ ρ1(K). Compare G1 on the =go page, where the same person values force sameness. The derivations up to this step are parallel. The difference is the construction (adjunct =go vs complement of ‘say’), not the words (PI-MATCH iii).

S4. (26) jidoogááł jiní: two readings from one structure

  1. Σ K: Σ(√ARRIVE) = {ρ1}, π = 4. A jidoogááł. Willie writes ji- before doogááł, the 3rd-person form. We do not segment further.
  2. Σ M: Σ(√SAY) = {ρA, ρTh = K}, π(ρA) = 4. A jiní. B jidoogááł jiní.
  3. Reading a (RS-RESP + RS-44, preferred): ρA(M) is a specific respected person, and PI-MATCH (ii) links ρ1(K) to it. ‘Hei said that hei would come.’
  4. Reading b (RS-Q): ρA(M) has no specific referent (quotative), so there is nothing for ρ1(K) to match. ρ1(K) is a respect 4 on its own: ‘They say that person will come.’
  5. HYPOTHESIS The ambiguity is in how ρA of ‘say’ is valued, not in Merge, Move or word order. Willie's own account: (27c) and (27d) “overlap” (p. 146).

S5. (23) deesháał jiní and *(23b, c)

  1. Σ K: Σ(√ARRIVE) = {ρ1}, π = 1: deesháał. M: π(ρA) = 4: jiní. B deesháał jiní.
  2. RS-1 would offer two values for the 1: the reported speaker or the utterance speaker. With 3 on ‘say’ (21), both are available.
  3. ✗ (23c) The utterance-speaker reading is out when ‘say’ is π = 4 (Willie p. 144).
  4. ✗ (23b) The quotative is out: “it is inappropriate to use a reportative concerning one's own actions” (p. 144). This is a reason from use, not from the rules above.
  5. ✓ (23a) Respect: the 1 is the respected person's own “I”, translated ‘hei’. HYPOTHESIS So a π = 1 slot in K can take the referent of a π = 4 slot in M, but a π = 3 slot cannot ((24a), PI-MATCH i). The first person under ‘say’ belongs to the reported speaker, not to the ordinary person scale.

S6. (29b) * vs (30) díínááł nijiní

  1. Σ K: Σ(√ARRIVE) = {ρ1}, π = 2: díínááł (with obligation, (28)).
  2. M in (29): Σ(√SAY) = {ρA, ρTh}, π(ρA) = 4: jiní. Respect reading ✗ (29b). Only the quotative survives (29a).
  3. M in (30): Σ(√SAY) = {ρA, ρx, ρTh}. The slot ρx is π = 2 (gloss “2.4” = object 2, subject 4). A the 2nd-person prefix goes left of jiní: nijiní.
  4. HYPOTHESIS The respect reading needs the hearer on ‘say’ as well as in K. So one way to state *(29b) is that a respect-4 speaker cannot reach into K to the 2nd person; ‘say’ needs its own π = 2 slot. ρx's role is OPEN (T13b). B díínááł nijiní.

8. Bearing on T12a, T12c, T12f

  • T12f (does parallel processing hold outside =go?): answered for reported speech. SOURCED No. With all third persons, ‘say’ + complement forces disjoint subjects (22), (27b). GO-PP is specific to the =go chain, as Willie's “differently across construction type” (p. 135) suggested. Update 2026-10-10 (afternoon): relative clauses (Ch. 6) are now read, and there parallel processing holds (relatives §5).
  • T12a (roles or functions in (9)?): moved, not settled. Willie's own summary, p. 148–149, says that in transitive conditionals “there is a presupposition that the arguments across the two clauses are coreferent, coindexed according to theta role.” That wording favours R1 (match by role, as her printed indices for (9) have it) over R2 (match by function, her prose at (15a)). So her text is split: (9)'s indices and p. 149 against (15a) and “(back)”. Still OPEN. It needs a speaker.
  • T12c (why a new π = 4 agent only from clause 1?): Willie's reason found. P. 149: “Therefore, it is impossible to use a fourth person pronoun in the second clause unless there is a fourth person pronoun in the first clause to serve as its antecedent.” In intransitive conditionals, by contrast, “a fourth person or a nominal may be added in the consequent clause to mark disjoint reference” (p. 148). SOURCED as her statement. That a C2 fourth person needs an antecedent while a C1 fourth person does not is not explained further on these pages OPEN.

9. Open questions

IDQuestionWhat would settle it
T13aAre the respect 4 and the quotative 4 on jiní one morpheme with two uses, or two morphemes with one form?Willie §4.2.2–4.2.5 in full; Y&M 1987; a speaker. Narrowed 2026-10-10 (evening): Willie §4.2.1–4.2.6 read. She treats them as uses of one fourth person (pp. 113, 121–122), distinguished by distribution (4 on jiní only vs on every verb; nouns with léiʼ), with no form difference in pp. 109–122 (fourth person uses §10). A morphological test is still OPEN
T13bIn (30) nijiní “2.4.said”, which role does the 2nd-person slot of ‘say’ carry: addressee, or the one spoken about (“[of you]”)?A source that prints ‘say’ with an object outside reported speech; a speaker. OPEN
T13cIs the b- of (19) booyeéł “3.3.carrying” the inverse bi- (horse = carrier, Bill = carried)? Willie prints no INV.A paradigm of the ‘carry (animate)’ verb; a speaker. OPEN
T13dIs PI-MATCH (iii) right: is the 3/3 default set by the construction? What does it do in relative clauses (Ch. 6) and in other complement-taking verbs (thinking, wanting)?Willie Ch. 6 (on the box); new data for other verbs. Answered 2026-10-10 (afternoon): yes; relatives share the cast, ‘believe’ is free (relatives §9). Other complement verbs: T14a.
T13eDoes FOCUS-4 hold for other perception verbs, and can a π = 3 object of ‘see’ in (19) be read as the person, or only as the event?Willie gives only (17)–(20). A speaker. OPEN
T13fVPISH: is the complement of ‘say’ in a complement position of ‘say’ (as S3 assumes), or adjoined like =go? Willie: these are the only subordinate clauses without a determiner/nominalizer enclitic (p. 149).Extraction or wh-question data out of the complement (cf. the multiple-wh test). OPEN

Not read this run: Willie Ch. 6 (relative and nominalized clauses, from p. 151; since read: relatives page), §4.2.4 (narratives) and §4.2.5 (the Impersonal) beyond p. 117 (since read: fourth person uses page, which also revises PI-MATCH (i) for the optative).

When 'that person' (ji-) is the doer: the verb's object goes silent and a postposition's object takes bi- (T67b)

Working note (newest). Not yet filed into a section; listed under Working notes.

Verbs on this page → dictionary: yiztał ‘kick’ · jizloh ‘rope’ · yishxash ‘bite’ · dabiiłtsą́ ‘see’ · yi-sh-cha ‘cry’ · ní ‘say’

Restated on the unified structure (2026-11-15). The fourth person is restated as a person value on an argument merged in Spec,vP or VP, with nothing extra moving: the optative fourth bound by the wisher (H51), the respect/report/story ji- as an Agree mark (H57a, OPEN against H57b), the areal ho- as a personless area argument (H54a, H60). 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. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §4, fourth-person 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.

With a fourth-person subject (ji- 'that person, one'), Willie's data show the same split that p-object-b found for 'I' and 'you'. A postposition's 'him/her' is bi-, never yi-. In the one sourced minimal pair it is bił 'with her' under 'that person', against yił under a plain 3rd person SOURCED. The verb's own 3rd-person object never shows yi- under ji-: 11 tokens, with 4 exact counterparts that have yi- under a plain 3rd person SOURCED. Hale's footnotes predict both results: the 4th person "does not behave like a third person in obviation" SOURCED. So for this rule the 4th person sorts with 'I' and 'you', not with 'he' HYPOTHESIS. With only one postposition token, "always" stays OPEN. There is also a tension with Hale 2001, where ji- carries no person feature at all OPEN.

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 Minimal pairs · 6 Hale 1997 vs Hale 2001 · 7 Rulesets A and B · 8 Derivations · 9 What it means for the role schema · 10 Fidelity · 11 Open questions

Notation as on p-object-b. A = the word level (prefixal, left-headed); B = the clause level (right-headed). Σ = the role schema (Dowty's proto-roles as the baseline). H = Hale's condition for yi-: a true 3rd-person subject c-commands the 3rd-person object. Merge builds structure: it puts two pieces together into a phrase. Agree copies feature values such as person and number from one piece to a probe that needs them. The person labels are 1, 2 and 3, plus 4 = fourth person ('that person', ji- as subject, ha-/ho- as object). S = subject and O = object. Willie writes the object first in her glosses: "3.4.kicked" = 3rd-person object, 4th-person subject (p. 153).

1. Question

T67b (from p-object-b §11): "4th-person subject (ji-) with a 3rd-person P object: PB-6 predicts b-, and the verb's object Ø." PB-6 says the spell-out of a 3rd person depends on the head that agreed with it. On a postposition (P) it is yi- if H holds and bi- otherwise, never Ø. On E's object slot (AGRo) it is yi- if H holds, Ø under a 1st, 2nd or 4th-person subject, and bi- otherwise. The prediction therefore rests on whether a 4th-person subject counts as a "true" 3rd person for H.

2. Sources

SOURCED Willie, MaryAnn 1991. Navajo Pronouns and Obviation. PhD dissertation, University of Arizona. Box scan; for chs. 4–6 the PDF page is the printed page + 6 (checked against the printed numbers). Page images read this run: pp. 112, 113, 115, 122, 124, 127, 128, 129, 138, 141, 142, 171, 172 and 173. Text layer only: p. 109 (§4.2.1).

SOURCED Hale, Ken 1997. "Remarks on the syntax of the Navajo verb" (hale023), Parts I–III. Read from the text layer: PDF p. 16 (fn. 9) and PDF pp. 26–27 (the fourth-person paragraph, fn. 16, fn. 17). Page image read: p. 27, trees (17)/(18).

SOURCED Hale, Ken 2001. "Navajo verb stem position and the bipartite structure of the Navajo conjunct sector." Linguistic Inquiry 32: 678–693; §4 "The Deictic Subject Prefixes", pp. 689–690. Box PDF; page images of PDF pp. 12–13 read this run.

Not opened: Young & Morgan 1987; Kari 1976; Perkins 1978. The box file labelled perkins1978 is the wrong document (see sources).

3. Rules first

FS-1  "If a third person object appears in the c-command domain of a first,           SOURCED (Hale 1997 fn. 9, PDF p. 16)
      second, or fourth person subject, ... If the third person object
      morphology appears in E, then it is realized as Ø; otherwise, it is
      realized as b-."
FS-2  "The fourth person shares zero-AGRs with the third person. It does not           SOURCED (Hale 1997 fn. 17)
      behave like a third person in obviation, however—only the 'true' third
      person (that realized pronominally as bí, for example) requires that a
      clausemate object in its c-command domain show obviative" marking.
FS-3  j- is "in reality a DP occupying syntactic subject position at                   SOURCED (Hale 1997 p. 26; trees
      d-structure"; it agrees with M (AGRs, zero) and ends up as a prefix to E.        (17)/(18), p. 27 image)
FS-4  The position-V prefixes "are not invested with a person feature. They are        SOURCED (Hale 2001 p. 690, image)
      third person only in the sense that a full DP or NP is third person",
      i.e. "adjunct nominal expressions linked to true arguments".
FS-5  "Fourth person, like first and second, is definite, human, and not               SOURCED (Willie p. 122, image)
      involved in the animacy hierarchy."
FS-6  Prediction (PB-6 + FS-1/FS-2): under a ji- subject, a 3rd-person P object        HYPOTHESIS (tested in §4–5)
      is b- and the verb's 3rd-person object is Ø.
FS-7  Timing: H mentions the subject, so a P's agreement cannot be spelled out          HYPOTHESIS (revises PB-7's
      before the subject is merged. Under VPISH the first point at which both are       "spelled out in the PP")
      present is the vP phase. P agreement is spelled out there (B). E's Ø is
      spelled out in the word (A).

Answer to T67b: yes, in all the sourced data. FS-6 holds for 1 postposition token and 11 verb tokens, with no counterexample. The postposition side still rests on a single token, and that token comes from a special construction, the quotative (§10). So the general claim stays OPEN.

4. The data

Forms are given as printed. Lines marked "source" are Willie's own glosses. The Leipzig lines are ours, built only from those glosses and the translations. Verb words are not segmented beyond what Willie's glosses give.

Willie 1991 p. 115 (11), (12) (page image): the postposition token
(1)ʼashkii léiʼʼatʼéédbiłjooʼash=gonáhodiiʼnilłájiní
source: boy a girl s3.with 4s.walking=when 4s.picked up (PL) EMP 4s.said
boy INDFgirl3-with4S.walk=when4S.pick.up(PL)EMP4S.say
'(I heard) some boy was walking with a girl when they were picked up (presumably by the police).' Willie: "the fourth person appears in each of the inflected verbs" (p. 115).
(2)ʼashkiiʼatʼéédyiłyiʼash=gonábiʼdiiʼnilłájiní
source: boy girl 3s.with 3DUAL.walk=when 3.picked up(PL) EMP j4s.said
boygirl3-with3.DU.walk=when3.pick.up(PL)EMP4S.say
'That person told me that a boy and girl were picked up while walking.' Here "fourth person appears only on the final verb, the verb of speech" (p. 115).

Same postposition -ił 'with', same comitative object ʼatʼééd 'girl', same walking event. The only change is the person of the walking clause's subject: 4 in (1), 3 in (2). The P's object is b- in (1) and y- in (2).

Willie 1991: 3rd-person object of the verb under a 4th-person subject (page images)
(3)jiztał
3O.4S.kick (Willie's gloss on (50), (52): "3.4.kicked")
'That person kicked it.' (30b, p. 128). Also (7b) John jiztał "3.4.kicked" 'She kicked him, John.' (p. 112)
(4)jishxash
source: 3.4.bit
'That person bit it.' (23b, p. 124). Also (31e) mósí jishxash 'That person bit the cat.' (p. 129)
(5)jizloh=goʼáádóó yiztał
source: 3.4.roped=COMP CONJ then 3.3.kicked
3O.4S.rope=COMPthen 3O.3S.kick
'That personi roped himj, and then hek kicked himj.' (13, p. 138). One clause has a 4th-person subject and the other a 3rd-person subject, and only the second shows yi-.
(6)haztał=yę́ęjiiłtsą́
source: 4.3.kicked=REL 3.4.saw
4O.3S.kick=REL3O.4S.see
'That personi saw the one who kicked himi.' (49, p. 173)
(7)haztał=yę́ęyiyiiłtsą́
source: 4.3.kicked=REL 3.3.saw
4O.3S.kick=REL3O.3S.see
'He saw the one who kicked that person.' (45, p. 172)

Count of verb tokens with a 4th-person subject and a 3rd-person object, taken from the page images: (7b) jiztał, (13) jizloh, (23b) jishxash, (30b) jiztał, (31e) jishxash, (46) jiztał, (47) jiiłtsą́, (49) jiiłtsą́, (50) jiztał and jiiłtsą́, and (52) jiztał. That makes 11 tokens of 4 verbs, and none shows yi- or bi-. That the object is a silent piece and not a yi- that was absorbed into the prefix is Hale's analysis (FS-1), not something visible on the page HYPOTHESIS (T68d).

5. Minimal pairs: same verb, 3rd- vs 4th-person subject

Host of the 3rd-person object3rd-person subject4th-person subjectFits
P -ił 'with'(12) ʼatʼééd yił yiʼash=go (p. 115)(11) ʼatʼééd bił jooʼash=go (p. 115)FS-1 "otherwise b-"; PB-6
Verb 'kick'ʼashkii yiztał 'The boy kicked it.' (p. 129)(30b) jiztał 'That person kicked it.' (p. 128)FS-1 Ø in E
Verb 'bite'ʼashkii yishxash 'The boy bit it.' (p. 129)(23b) jishxash 'That person bit it.' (p. 124)FS-1 Ø in E
Verb 'see'(45) yiyiiłtsą́ "3.3.saw" (p. 172)(49) jiiłtsą́ "3.4.saw" (p. 173)FS-1 Ø in E
Verb 'kick', one sentence(13) second clause yiztał (p. 138)(13) first clause jizloh 'roped' (p. 138)FS-1 (different verbs, same frame)

The pairs for 'kick' and 'bite' come from the same discussion. Willie gives the ʼashkii yiztał/yishxash replies on p. 129 as 3rd-person alternatives to (30b) and (28b), to explain "Why are fourth person forms employed" SOURCED. The pattern across the table: a 4th-person subject leaves the verb's object without a marker and gives the postposition's object b-, exactly as 'I' and 'you' do on p-object-b §6 SOURCED (forms); the grouping of 4 with 1/2 is HYPOTHESIS (it is what FS-1 states).

6. Hale 1997 vs Hale 2001: is ji- a third person?

Hale 1997 lists 4 with 1 and 2 in fn. 9 (FS-1). He adds that the 4th person shares zero AGRs with the 3rd person but is not a 3rd person "in obviation" (FS-2) SOURCED. Hale 2001 instead treats the position-V prefixes as lacking any person feature: they are "adjunct nominal expressions linked to true arguments", third person "only in the sense that a full DP or NP is third person, like ashkii" (FS-4) SOURCED. Hale 2001 §4 does not discuss obviation or yi- SOURCED (pp. 689–690 read).

Read literally, FS-4 makes ji- work like the noun ʼashkii: a pronoun-like adjunct linked to an ordinary 3rd-person subject argument. That subject would then be a true 3rd person, and H would predict yiztał, as in ʼashkii yiztał. The data say jiztał, and bił rather than yił. So whatever triggers yi- must see something that ji- adds to the subject, beyond a [3] value in Mode/Subj (M) HYPOTHESIS. Two ways to state it, both HYPOTHESIS:

  • FS-8a (Hale 1997 style). The subject DP carries a feature that "true 3rd" excludes (call it [4]). H is stated over the subject DP's features, not over M's agreement. This is consistent with FS-2: same zero AGRs, different obviation.
  • FS-8b (Hale 2001 style, repaired). ji- carries no person, but linking ji- to the subject turns it into a definite human that is outside the animacy ranking (FS-5). H only applies to subjects that are on that ranking. This would also explain why Willie's 4th person "can be used instead of Inverse voice forms" (p. 122).

Sentence (1) separates these from a plain reference account. It has an overt noun, ʼashkii léiʼ 'some boy', linked to the 4th-person subject, even though Willie says elsewhere that "nouns are syntactically third person, so a nominal cannot be adjoined to a fourth person pronoun" (p. 109, text layer) SOURCED. Even with that noun the postposition keeps b-. Hale fn. 16 names exactly this kind of DP: "in third or fourth person, [DP NP léiʼ] introduces a discourse referent". He calls the overt-nominal 4th person "a form of agreement" that "must be analyzed differently—e.g., perhaps with j- base-generated as an adjunct to E" SOURCED. In that use "the referential properties of the fourth person are identical to those of a normal third person" (fn. 16) SOURCED, yet (1) still has bił. So H is not about reference. So the result does not depend on which of Hale's two analyses of j- is chosen. In both, the 4th-marked subject fails H HYPOTHESIS (T68b).

7. Rulesets A and B

StepWhereRulesetWhat happens
1PPB (head-final: [DP P])Merge P -ił with its object (pro linked to ʼatʼééd). Agree: P's probe is valued [3].
2VP / vPB (head-final; Hale's VP is head-right)Merge the subject in vP (VPISH): in (1) a DP with [4] (or linked to j-), in (2) a true 3rd person.
3vP phaseBSpell-out of P's [3] (FS-7): H is checked. (2) H holds → yił; (1) H fails → bił.
4M, EA (prefixal)Agree M ← subject (zero AGRs for both 3 and 4, FS-2). E's AGRo is valued [3] when the verb has an object.
5WordASpell-out of E's [3]: H holds → yi-; subject 4 → Ø (FS-1). Morphological Move of j- to its prefix position before E (Hale p. 27).
6Word edgeA→B changeoverThe finished verb word (jooʼash, jiztał) is ordered last in its clause (head-final B). The PP comes before it.

The word is still the changeover point. One thing changes from p-object-b: P agreement cannot be decided "inside the PP" alone, because H needs the subject. The PP is built in B, but its spell-out waits for the vP phase, which is also in B. The Ø rule belongs only to the A cycle of the word. Both cycles check the same subject property. The difference is that only the word cycle has a Ø in its list of spell-outs HYPOTHESIS.

8. Derivations

D1. (1) ʼatʼééd bił jooʼash=go 'when that person was walking with a girl' (clause 1 of Willie (11); vP/TP spine; the structure of the verb word as on derivations.)

1  Merge  [PP pro(3)  -ił ]                              B; head-final PP
2  Agree  P ← [3] from pro                               P's person probe valued
3  Merge  [VP [PP ...] walk ]                            B; head-right VP
4  Merge  [vP DP[4]  [VP ...] v ]                        subject starts low (VPISH); DP linked to j-
                                                          (and here to 'ashkii léi', Hale fn. 16 type)
5  Spell  P's [3] at the vP phase (FS-7): H? subject is   PB-6 + FS-1/FS-2
          [4], not a true 3rd → H fails → b-  ⇒ bił
6  Merge  M (AGRs); Agree M ← subject: zero (FS-2)         A; no AGRo: 'walk' has no object
7  Move   j- prefixed in the word (morphological Move)    A  ⇒ joo'ash (not segmented further)
8  Adjoin 'at'ééd linked to pro in PP; 'ashkii léi'      B
          linked to the subject
9  Order  'ashkii léi'  'at'ééd bił  joo'ash  =go         B; head-final; =go at the clause edge

D2. (2) ʼatʼééd yił yiʼash=go: the same as D1 except at steps 4–5 and 7. Step 4 has a true 3rd-person subject (pro linked to 'ashkii', DU). At step 5, H holds, so the P's object is y- ⇒ yił. There is no j-, so step 7 does not apply ⇒ yiʼash. The person of the matrix 'say' (4) has no effect on the P inside the =go clause. This fits Hale fn. 17's "clausemate object" SOURCED (wording). Whether it holds for P objects in general rests on one token HYPOTHESIS (T68c).

D3. (30b) jiztał 'That person kicked it.' vs ʼashkii yiztał 'The boy kicked it.'

1  Merge  [VP pro(3)  kick ]                            B; head-right VP
2  Merge  [vP DP[4] [VP ...] v ]                         VPISH; for 'ashkii yiztał: DP = true 3rd pro
3  Merge  M; Agree M ← subject: zero AGRs                A (FS-2; same for 3 and 4)
4  Merge  E; Agree E (AGRo) ← [3] from the object        A; this is the verb's own object slot
5  Move   subject DP to Spec,EP                          Hale p. 27 trees (17)/(18)
6  Spell  E's [3]: subject [4] → Ø (FS-1)                 A; for 'ashkii yiztał: H holds → yi-
7  Move   j- prefixed before E (morphological Move)       A
8  Output jiztał  vs  yiztał                              the rest of the word as printed, not segmented

Across D1–D3 the subject property is the same. The output differs by host: b- on P (B cycle) and Ø on E (A cycle). This is the split of p-object-b, now also under the 4th person HYPOTHESIS.

9. What it means for the role schema (Focus 2)

In (1) and (2) the role schema Σ is the same. The walkers are the proto-agent-like movers and the girl is the comitative brought in by -ił. In jiztał and ʼashkii yiztał the kicker is the proto-agent and the thing kicked is the proto-patient HYPOTHESIS (role assignment is ours). Even so, the object marker changes: yi-, bi- or Ø. It changes only with the subject's category, true 3rd person vs 4th person. Role content alone does not fix how a slot is spelled out. What fixes it is a property of what fills the subject slot: [4] or the link to ji- HYPOTHESIS. This does not refute "theta roles, not words, as arguments". It does narrow it. Even the verb's core slots are spelled out in a way that depends on a feature of their filler, beyond its role, so a pure role schema is not enough to predict the verb word. This is one more point against the strong thesis, which Grant already sees as not established; see the verb guide and the proto-role schema.

10. Fidelity

  • Willie forms were read on page images this run. The PDF offset is +6 for chs. 4–6 (the p-object-b note used +5 for earlier chapters). The text layer garbles ł, so no form was copied from it.
  • (1) is a quotative. Willie treats it as a special case where nouns occur with fourth-person pronouns (p. 115). It is the only postposition token found, so it does not show a plain 4th-person subject with a P object.
  • Willie's gloss order is inconsistent. (46) jiztał=yée is glossed "4.3.kicked=REL" and (47) jiiłtsą́ "4.3.saw", although the translations have the 4th person as subject. (49), (50) and (52) gloss the same words as "3.4". We follow the translations and p. 153 ([sic] on the 4.3 glosses). (46) prints =yée where the others have =yę́ę; copied as printed.
  • (27b) yiztal and (28b) haztal are printed with a plain l on p. 127 and are not used here. (30b), (13) and the p. 129 replies print ł.
  • The role assignments in §9, FS-6 to FS-8 and the derivations are ours. FS-1 to FS-5 are quotations.
  • Pages also read but not used: p. 141 (18) Mary jicha=go hoołtsą́ 'I saw Mary when she was crying.' (no object); p. 142 (20) Bill łį́į́ʼ hooyééł=go hoołtsą́ "horse 4.3.carrying=while 4.1.saw" (4th-person object, see T68e); p. 171 (43) (1st-person P object).

11. Open questions

IDQuestionWhat would settle it
T68aA second 4th-person subject with a 3rd-person postposition object, outside the quotative: polite, story or impersonal use. Is it always b-?Willie chs. 4–6 further examples; Young & Morgan 1987 paradigms. OPEN
T68bHale 1997 (FS-2: zero AGRs, not 3rd person in obviation) vs Hale 2001 (FS-4: no person feature). Which one derives the absence of yi-, FS-8a or FS-8b? Does any later Hale text restate H?Hale & Platero 1996; Hale 2003 (if openable). OPEN Update 2026-10-28: answered for FS-8a over a "position V filled" rule on indefinite-agent-b (agentive passive biʼdistał vs jiztał; HYPOTHESIS); FS-8b still open (T69e).
T68cIs H checked against the subject of the P's own clause (D2) or against a higher one? The reverse case is needed: a 4th-person subject inside a =go or relative clause, a 3rd-person matrix subject, and a P in the inner clause.Willie ch. 5 (=go chains); clause-chains. OPEN
T68dIs the verb's object under ji- really a silent piece (FS-1), or a yi- absorbed by the prefix in the sound system?Kari 1976; Young & Morgan 1987 on yi- + ji-. OPEN
T68eWith a 4th-person object (Willie (20) hooyééł "4.3.carrying"), no inverse is needed. Does a 4th-person P object (ha-, Willie table (27)) behave the same way?A sourced ha-+P form with a 3rd-person subject. Willie (19) print is damaged. OPEN
T68fAlgonquian: do the unspecified-actor ("X") forms pattern with speech-act persons, not with 3rd persons, for obviation, as the Navajo 4th person does here?Oxford's Algonquian chapter §4 onward. OPEN

Sources: Willie 1991 pp. 112, 113, 115, 122, 124, 127–129, 138, 141–142, 171–173 (page images) and p. 109 (text layer); Hale 1997 (hale023) PDF p. 16 fn. 9, pp. 26–27 incl. fn. 16, fn. 17 and trees (17)/(18) (image); Hale 2001 LI 32 pp. 689–690 (images). See also p-object-b, hale-general-b, fourth person, 4th: five uses, subject on the head, open questions, sources.

2026-11-14 evening · First-principles review 14: Willie on the impersonal and optative fourth person, and the fourth person instead of the inverse (ch. 4, §§4.2.5–4.3)

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.

HYPOTHESIS Plain-words verdict. When a fourth person stands in a wish, it means the one doing the wishing: the speaker in “I want to sleep”, the father in “My father wishes he could buy the horse”, the hearer in the yes/no question “Do you wish to go to sleep?”. With the prohibitive lágo there is no wish, and the same verb means “him”. That the wisher is the only possible reading fits best if the wish clause sits inside the “think” clause, so this counts for the complement analysis (H40b) and leaves H40a and H40b even. Using a fourth person to avoid the inverse follows from one ranking rule that compares only third persons (H53). But Willie’s own question-and-answer pairs show that a lone noun can be the doer “in context”, so the one-noun rule (H4″) is a default, not grammar. The spine already predicts that ambiguity, with nothing extra moving. Two fourth persons in one clause are two different people (“That person bit that person”), so H32b holds only across clauses. Telicity: changing laanaa to lágo on the same verb changes the reading, so that change is not about telicity; otherwise OPEN. The grid gains a yes/no question and five wh-questions, all active; there is still no negative or passive question.

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–H49 from reviews 1–13 (here mainly H3, H4″, H6′, H23, H31, H32a/b, H40a/b, H41, H45, H48), head direction (HS-1, T⁰ final, against HS-2, T⁰ initial with remnant VP movement; the derivations are written for HS-1), categories other than the spine’s, Agree, binding and any discourse rule. Treating the optative mode as a value of T⁰, so that the working default applies to optative clauses, is HYPOTHESIS too. The live question for word order is what else moves. Every derivation ends in a logical form (LF) whose scope order matches c-command. Silent pronouns are written with their person: x₃, y₃ (third), z₄, w₄ (fourth). The role labels are ours. Willie’s analysis (nominals adjoined to pronouns, an internal “ergative” agent in the inverse, the animacy hierarchy as a principle of interpretation, “indirect speech acts”) appears only as labeled contrast.

The source’s claims, as its own

  • SOURCED Impersonal (§4.2.5). “Another use of the fourth person is the Impersonal construction” (p. 117); such sentences “are used to instruct or advise the hearer about acceptable behavior” (p. 118) and “often contain a value judgment or a negative. That is, some modal feature” (p. 119).
  • SOURCED Optative (§4.2.6). “Since it is not considered polite to speak openly of one’s own desires or wishes, a speaker will use fourth to express that wish” (p. 120). “with lágo and the verb in an optative fourth person, it gives a prohibitive reading” (p. 120). “in the actual meaning the fourth person in optative mode always refers to the speaker of the clause indicating that reference outside the clause is not possible” (p. 121); “it is linked to the speaker in (17) by default because there is no subordinate clause; it is linked to third person human in (18), and in (19) linked to second person” (p. 121).
  • SOURCED Animacy (§4.2.7). “when a lower-ranked agent acts upon a higher-ranked patient, and all arguments are third person, then the Inverse voice must be used” (p. 122). “Fourth person, like first and second, is definite, human, and not involved in the animacy hierarchy” (p. 122). “All third person pronouns, if they do not have a nominal adjunct, must have a discourse antecedent … The fourth person … does not require a discourse antecedent -- it is simply deictic” (p. 124). With a 1, 2 or 4 argument, “the single nominal … can be linked to either an external or an internal argument; this is not the case with third person constructions, where a single nominal is always linked to the internal argument” (p. 126). “external arguments in Inverse constructions must be definite. This limitation does not apply to any Direct construction” (p. 126).
  • SOURCED Question–answer pairs. The reply to a question that presupposes neither referent “must include either two nominals, two fourth person pronouns, or a combination of the two” (p. 129). The third-person replies ʼashkii yishxash and ʼashkii yiztał “show incorrect word order … Nonetheless, this reading … is permitted in context” (pp. 129–130). The fourth-person forms “cannot be Inverse” (p. 130).

The checked data

SOURCED Exact forms, source glosses and translations, with example numbers and printed pages. Glosses list object before subject (jishxash “3.4.bit” is “That person bit it”). Printing notes and crops are at the linked index entries.

  • Impersonal (p. 119), cited. NX-236 (15): Doo hatis njigháa da “NEG 4.over 4.walks NEG”, “One does not step over people.” · NX-237 (16): ʼabíní=go hojitaał=go yáʼátʼééh “Morning=while 4.sings=while 3.good”, “It is good for one to sing in the morning.”
  • Optative (pp. 120–121). NX-415 (17a): ʼajoołháásh laanaa “4.OPT.sleep wishfully”, “Lit. That certain person wants to sleep.” “(I want to sleep.)” · NX-416 (17b): ʼajoołháásh lágo “4.OPT .sleep NEG”, “Don’t let him go to sleep.” · NX-417 (18): Shizhéʼé łį́į́ʼ nahijólniih nizin “1.father horse 3.4.OPT.buy 3.thinks”, “Lit.: My father thinks that that person should buy a horse.” “My father wishes he could buy the horse.” · NX-418 (19): ʼajoołháásh laanaa nínízin=ísh “4.OPT.sleep wishfully 2.thinks=Q”, “Do you wish to go to sleep?”
  • Animacy (pp. 122–126). NX-419 (20a): ʼashkii łééʼchąąʼí yishxash “boy dog 3.3.bit”, “The boy bit the dog.” · NX-420 *(20b): łééʼchąąʼí ʼashkii yishxash, “(The dog bit the boy.)” · NX-421 (20c): ʼashkii łééʼchąąʼí bishxash “INV3.3.bit”, “The boy was bitten by the dog” · NX-422 (21): łééʼchąąʼí shishxash “dog 1.3.bit”, “The dog bit me.” · NX-423 (22): łééʼchąąʼí hashxash “dog 4.3.bit”, “The dog bit that person.” · NX-424 (23a): hashxash “4.3.bit”, “It bit that person.” · NX-425 (23b): jishxash “3.4.bit”, “That person bit it.” · NX-426 (24): Yíh, haztał “4.3.kicked”, “Oops, It kicked him [that person].” (a rodeo rider kicked by the horse, as a conversation opener) · NX-427 *(25): Yíh, biztał, “It kicked him [third person].” (starred as an opener: it “would not suffice as an initial remark”) · NX-428 (26): séígoʼ hashshish “scorpion 4-3-sting”, “A scorpion stung her.”
  • Questions and replies (pp. 127–130); no gloss lines. NX-429 (27a): haash yítʼiid, “What did he do?” → NX-430 (27b): mósí yiztal, “He kicked the cat.” · NX-431 (28a): haash yítʼiid, “What did it do?” → NX-432 (28b): haztal, “It kicked him [that person].” (plain l as printed; haztał in the p. 130 repeat) · NX-433 (29a): haash yidzaa, “What happened to him?” → NX-434 (29b): mósí bishxash, “He was bitten by the cat.” · NX-435 (30a): haash yidzaa, “What happened to it?” → NX-436 (30b): jiztał, “That person kicked it.” · NX-437 (31a): haash hóótʼiid, “What happened?” → NX-438 (31b): ʼashkii mósí yishxash, “The boy bit the cat.”; NX-439 (31c): ʼashkii mósí bishxash, “The boy was bitten by the cat.”; NX-440 (31d): mósí hashxash, “The cat bit that person.”; NX-441 (31e): mosi jishxash, “That person bit the cat.”; NX-442 (31f): hojishxash, “That person bit that person.” · Third-person replies, p. 129: NX-443 (second (28)): ʼashkii yishxash, “The boy bit it.”; NX-444 (second (30)): ʼashkii yiztał, “The boy kicked it.”

SOURCED What the stars are on. *(20b) is starred with a parenthesized translation, which names the reading it lacks. *(25) is starred as a conversation opener, a discourse judgment, not a claim that the string is ill-formed (Willie compares it to an English pronoun used without an antecedent). The same strings haash yítʼiid and haash yidzaa are printed twice, with “he/him” and with “it”.

HYPOTHESIS Our editorial glosses (not the source’s). Agent>patient, from the source glosses and translations only, stems not segmented: NX-417 my.father horse buy.4>3.OPT think.3; NX-418 sleep.4.OPT wish think.2=Q; NX-423 dog bite.3>4; NX-425 bite.4>3; NX-442 bite.4>4. We do not segment haash, hóótʼiid or hojishxash. Whether haash contains the question enclitic =sh is not shown by the source (no gloss), so it is not used as evidence for H11 or H12.

Hypotheses tested, revised or added

  • Carried over: H3 (inverse: the patient moves above the agent, which has raised to Spec,TP), H4″ (a single nominal goes with the lowest third-person participant), H6′ (only a definite or presuppositional patient makes that Move), H23 (sentential negation is a final Neg⁰ above TP), H31 (a noun is linked 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), H40a/H40b (a clause before “think/say” is a separate clause / the Theme complement), H41a (an adjoined say clause fixes the context of the root clause), H45 (=go pairs), H48 (an event is third person).
  • H50 (new) · a silent inverse-moved patient needs a linguistic antecedent. A silent third-person pronoun that has made the inverse Move (H3) must have been mentioned before; a silent third-person agent in Spec,TP can be resolved from the situation. This sharpens H6′ for silent pronouns. It replaces Willie’s broader “all third person pronouns … must have a discourse antecedent”, which her own (24) contradicts (its agent “it”, the horse, has not been mentioned).
  • H51 (new) · the wisher binds the fourth person of a wish. In an optative clause that expresses a wish (with laanaa, or as the complement of nízin “think”), a fourth-person subject is bound by the holder of the wish: the closest c-commanding attitude subject, and in a root clause the speaker. The bound fourth person is exempt from H32a. Without a wish (lágo) it is free.
  • H52 (new) · laanaa and lágo are final heads above the optative TP. H52a laanaa “wishfully” is the C⁰ of a wish clause. H52b lágo is a Neg⁰ that takes the whole optative TP, under a silent directive C⁰. The rival (H52c) has lágo spell out ¬ and the directive together in one head; nothing here separates them.
  • H53 (new) · the ranking condition. Among third-person arguments of one verb, a higher-ranked one (here: human over animal) must c-command a lower-ranked one at spell-out. First, second and fourth persons are not ranked. With a lower-ranked agent and a higher-ranked third-person patient, only the inverse Move (H3) satisfies it. This is the first statement on the site of the animacy ranking as a condition that forces the inverse; it is still a stipulation (why ranking is computed on c-command is not derived).
  • H32b′ (revision) · fourth persons in different clauses. Two fourth-person pronouns in different clauses of one sentence refer to the same person. Two fourth persons in one clause are different people (from H45c for a verb’s two arguments; for a verb and its postposition object, NX-236).
  • H4″-pref (revision) · the one-noun rule is a default. H4″ gives the preferred reading of a lone noun with two third-person participants; context can override it.

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 (HS-1); under HS-2 each clause would also move its remnant VP above T⁰, as on the spine page.

  1. The shared clause-internal steps (all of NX-415–444). Transitive clause: merge V with its internal argument (a noun or a silent pronoun), which gets the Patient role; merge v; merge the agent in Spec,vP, which gets the Agent role (VPISH). V→v→T; the agent moves Spec,vP→Spec,TP; its lower copy is silent (working default). Person prefixes and voice 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, as in earlier reviews. A noun is linked to a third-person participant only (H31). In a bi- clause the patient also moves above the raised agent (H3). Spell-out: the verb word is last. With one or two nouns, direct order (agent, patient, verb) needs nothing else to move.
  2. Impersonal, cited: NX-236 (15), NX-237 (16). (15): merge P “over” with a silent fourth person w₄ (“people”); the PP merges with V “walk” (the path is a role assigned by V; label ours); merge v and z₄ in Spec,vP (Agent); V→v→T; z₄ to Spec,TP (default). doo adjoins inside TP (H23; its exact site is OPEN), and Neg⁰ da takes the TP. A generic operator over the person is merged above NegP (as in review 6). LF: GEN z[person(z)] ¬∃e[walk(e) ∧ Agent(e,z) ∧ over(e,w₄)], z ≠ w₄. The two fourth persons are in one clause and are not the same individual (“one” and “people”), which already limits H32b (derivation 6). (16): yáʼátʼééh is glossed “3.good”: its argument is the singing, an event, and it is third person, while the singer is fourth person (ji-). H48 passes out of sample. Whether the =go clause is the argument of “good” or an adjoined clause with a silent third-person subject resolved to its event is the same open choice as H48a/H48b in review 13. LF (H48b form): WHEN(GEN z ∃e2[sing(e2) ∧ Agent(e2,z) ∧ in.morning(e2)], ∃s[good(s) ∧ Theme(s,it₃)]), it₃ = e2. Sentential versus constituent negation: (15) is sentential (doo … da around the clause); nothing here is constituent negation. Grid: neither is a question. The impersonal is an active clause with a fourth-person agent, not a passive. No question with an impersonal agent is on these pages.
  3. Optative with laanaa: NX-415 (17a). Merge V “sleep”; merge v and z₄ in Spec,vP (Agent); merge T (optative, HYPOTHESIS); z₄ raises to Spec,TP (default); merge C⁰ laanaa (H52a). Spell-out: verb word, then laanaa, as printed. The holder of a root wish is the speaker, so z₄ is bound by the speaker (H51). LF: WISH_speaker(∃e[sleep(e) ∧ Agent(e,z₄)]), z₄ = speaker: “(I want to sleep.)”. The “Lit.” translation, “That certain person wants to sleep”, is the unbound reading, which Willie gives as literal only.
  4. Optative with lágo: NX-416 (17b), sentential negation. The same verb word, the same Merge, roles and default. Neg⁰ lágo takes the whole optative TP (H52b), and a silent directive C⁰ is merged above NegP. No wish operator, so H51 does not apply and z₄ is free: “him”. LF: DIR_hearer(¬∃e[sleep(e) ∧ Agent(e,z₄)]), z₄ ≠ speaker by the printed “him”. ¬ c-commands the event and is c-commanded by the directive, as the order of the LF shows. lágo negates the clause, not a phrase in it: it is sentential negation, and like da (H23) it is clause-final with no doo. Result: (17a)/(17b) are a minimal pair, one verb word with laanaa or lágo, and the reference of the fourth person changes from the speaker to someone else. This is what H51 predicts. It also contradicts Willie’s “linked to the speaker … by default because there is no subordinate clause” (p. 121) as a statement about optative fourths in general: (17b) has no subordinate clause and is not linked to the speaker. The tree, panel A, draws the H52c form (one final head for ¬ and the directive); H52b gives the same scope order.
  5. Optative under “think”: NX-417 (18), NX-418 (19), and H40a vs H40b. (18): inner clause: merge V “buy” with łį́į́ʼ (Patient; linked to the only third-person participant, H31); merge v and z₄ (Agent, ji-); default. Under H40b that clause is the Theme complement of V “think”; merge v and the third-person thinker x₃ in Spec,vP (Experiencer), linked to Shizhéʼé (H31), which raises to Spec,TP and c-commands z₄. H51: z₄ is bound by the thinker. LF: ∃e1[think(e1) ∧ Exp(e1,my.father) ∧ Theme(e1, WISH ∃e2[buy(e2) ∧ Agent(e2,z₄) ∧ Patient(e2,horse)])], z₄ = my.father. (19): the same, with laanaa in the inner C⁰ (H52a), a second-person thinker, and C⁰[Q] =ísh merged last, written on the final verb word. LF: ASK_POL(∃e1[think(e1) ∧ Exp(e1,you) ∧ Theme(e1, WISH ∃e2[sleep(e2) ∧ Agent(e2,z₄)])]), z₄ = you. Q c-commands the think clause and the wish inside it, so the question is about the wish, as translated. H32a is contradicted by (18) (fourth = third person) and (19) (fourth = second person): it must exempt the bound fourth person of a wish (H51). H40a vs H40b. Under H40b the binding is c-command by the thinker in Spec,TP, so it is obligatory, as Willie says (“reference outside the clause is not possible”). Under H40a the wish clause is a separate root clause; to get the thinker it needs review 12’s context fixing (H41a) extended from “say” to “think”, and since H41 makes the shift optional, it also predicts the speaker reading (“My father thinks: I wish I could buy the horse”) and the unbound “Lit.” reading, which the source says the sentence does not actually have. So H40a needs an extra stipulation here and H40b does not. Result: a second point for H40b (after review 12’s (30)), against two for H40a (review 11’s (31)/(32), review 12’s missing enclitic). Even; not settled. In (19) laanaa sits in the inner clause and the thinker, not the speaker, binds z₄: the closest attitude subject wins, as H51 states. See the tree, panel B.
  6. Two fourth persons in one clause: NX-442 (31f), and H32b. hojishxash “That person bit that person”: merge V “bite” with w₄ (Patient, ho-), merge v and z₄ (Agent, ji-), default. The reply must introduce two people who are not presupposed (p. 129), and H45c (a plain transitive verb cannot have the same individual as agent and patient) gives z₄ ≠ w₄. LF: ∃e[bite(e) ∧ Agent(e,z₄) ∧ Patient(e,w₄)], z₄ ≠ w₄. H32b as stated (“two fourth persons in one sentence are the same person”) is contradicted by NX-442 and by NX-236 (“one” ≠ “people”). Every case that supported it (reviews 10, 12 and 13) had the two fourth persons in different clauses. Result: H32b′ (different clauses only) replaces it; the earlier derivations are unaffected.
  7. Ranking: NX-419–421 (20), NX-422/423 (21)/(22), NX-428 (26). (20a): boy in Spec,vP (Agent), dog in VP (Patient); the boy raises to Spec,TP and c-commands the dog; H53 is met. LF: ∃e[bite(e) ∧ Agent(e,boy) ∧ Patient(e,dog)]. *(20b) on the reading “The dog bit the boy”: dog in Spec,vP, raised to Spec,TP, c-commands the boy in VP; the dog ranks lower, so H53 fails: starred, as printed. (20c): the same roles, and the patient ʼashkii (definite, H6′) moves above the raised agent (H3), so the boy c-commands the dog: H53 met, and the verb is spelled with bi-. LF: ∃e[bite(e) ∧ Agent(e,dog) ∧ Patient(e,boy)], the same as the English “was bitten by the dog”. (21) and (22): the patient is first or fourth person, not ranked, so a direct clause with a dog as agent is fine; the noun is linked to the only third-person participant, the agent (H31). LF (22): ∃e[bite(e) ∧ Agent(e,dog) ∧ Patient(e,z₄)]. (26): the same, with an indefinite agent noun in Spec,TP. LF: ∃x[scorpion(x) ∧ ∃e[sting(e) ∧ Agent(e,x) ∧ Patient(e,z₄)]]. Willie’s “external arguments in Inverse constructions must be definite. This limitation does not apply to any Direct construction” matches H6′, which restricts only the moved patient. Result: H53 derives (20)–(22) and (26), H3 and H6′ pass out of sample (NX-421), and H31 passes out of sample (NX-422, 423, 428). What else moves: only the inverse patient in (20c), the same single Move as before.
  8. Conversation openers: NX-426 (24), NX-427 *(25). Rodeo: the horse kicked the rider. As a third-person clause, H53 forces the inverse (rider over horse), so the only third-person form is biztał, whose patient is a silent pronoun that has made the inverse Move. With nothing said yet, H50 rules it out as an opener: starred, as printed. (24) haztał has a fourth-person patient: H53 does not apply, no inverse Move is made, and the silent third-person agent (the horse) can be resolved from the scene (H50 does not restrict it). LF (24): ∃e[kick(e) ∧ Agent(e,it₃) ∧ Patient(e,z₄)], it₃ = the horse in view. Result: derived with H53 + H50. Willie’s own account (every third-person pronoun without a noun needs a discourse antecedent) wrongly excludes (24) too, because the horse in (24) has not been mentioned either. Ours restricts the requirement to the inverse-moved patient, a stipulation to be tested (OPEN prediction below).
  9. Wh-questions: NX-429, 431, 433, 435, 437. (27a)/(28a) haash yítʼiid: merge V “do” with haash (Theme, label ours; a wh-word merged where a noun with that role would be, as in review 3); merge v and a silent x₃ (Agent); default. C⁰[Q] is merged at the top and linked to haash by Agree. Because the subject is silent, nothing shows whether haash stays in VP or moves to Spec,CP; neither order needs an extra Move. LF: ASK_WH(λx ∃e[do(e) ∧ Agent(e,x₃) ∧ Theme(e,x)]). The string carries no person or animacy difference between “he” and “it”: x₃ is resolved from the context, as H50 allows for a silent agent. (29a)/(30a) haash yidzaa: the same, with “happen to” (Theme = what; the one it happens to is a silent y₃, a role we label Patient). LF: ASK_WH(λx ∃e[happen(e) ∧ Theme(e,x) ∧ Patient(e,y₃)]). (31a) haash hóótʼiid “What happened?”: no participant is named. LF: ASK_WH(λx ∃e[happen(e) ∧ Theme(e,x)]), where the answers (31b–f) show that x ranges over events. OPEN H48 at risk: if the ho- of hóótʼiid (and of review 11’s ʼáhóótʼiid, NX-385) is a fourth-person pronoun referring to the event, H48 is contradicted. Willie’s summary lists “pleonastic or expletive subjects, with the prefix ho-” (p. 131), treated in §4.1, which we have not reviewed. That is the next push.
  10. Replies and the one-noun rule: NX-430, 434, 438–444. (27b) mósí yiztal “He kicked the cat”: the noun in VP as Patient, the silent human agent raised to Spec,TP; H4″’s preferred reading. (29b) mósí bishxash: the silent patient y₃ (given by the question, as H50 requires) moves above the raised agent (H3), and the noun is linked to the agent: “He was bitten by the cat”. (31b–e): as in derivation 7; in (31d)/(31e) the noun is linked to the only third-person participant (H31). Now the third-person replies of p. 129. (Second (30)) ʼashkii yiztał “The boy kicked it”, a reply to “What happened to it?”: the noun is the agent, which Willie says is “permitted in context”. On the spine this costs nothing: the noun is merged in Spec,vP and raised to Spec,TP, the given “it” is a silent pronoun in VP, and the string is the same as with the noun in VP. LF: ∃e[kick(e) ∧ Agent(e,boy) ∧ Patient(e,it₃)], it₃ = the thing asked about. Result: H4″ is contradicted as a rule of grammar by NX-444, and is kept only as a default (H4″-pref). The working default predicts the ambiguity of a lone noun with two third persons, with nothing else moving in either parse. Willie’s “a single nominal is always linked to the internal argument” (p. 126) is contradicted by her own NX-444. Earlier results that used H4″ to exclude a reading, rather than to predict the preferred one, need rechecking (OPEN). OPEN A source conflict, (second (28)), NX-443. ʼashkii yishxash is translated “The boy bit it”, but it is offered as a reply to “What did it do?”, where “it” is the biter. On the printed translation it is like NX-444 (the boy as agent; a second counterexample to H4″ as a rule). On the reading the question calls for, “It bit the boy”, the boy is the patient in its usual place, and the clause is a direct form with a lower-ranked agent acting on a higher-ranked third person, a counterexample to H53 “in context”. Willie’s prose (“both … show incorrect word order”) fits the first; her question fits the second. We do not choose.
  11. “Fourth-person forms cannot be inverse” (p. 130). haztał “It kicked him [that person]” cannot mean (32) “It was kicked by him”. In our terms ha- marks the fourth person as the patient, and the inverse Move (H3) never changes roles, so the meaning “that person kicked it” needs a different verb word, jiztał (30b). This is consistent with review 10’s result that the ha- patient need not move (H29 was dropped). Nothing new moves.

Analysis A: (17b) NX-416, drawn in the H52c form: lágo is one final head over the whole optative TP, spelling both ¬ and the directive; the fourth person is not bound by the speaker Hypothesis

Analysis A: (17b) NX-416, drawn in the H52c form: lágo is one final head over the whole optative TP, spelling both ¬ and the directive; the fourth person is not bound by the speakerPanel A. Merge V “sleep” (no internal argument); merge v and a silent fourth-person pronoun z₄ in Spec,vP, which gets the Agent role. Merge T (optative). Merge lágo as the final head over the whole TP, spelling the prohibitive: ¬ plus directive force (H52c). It takes the whole clause, so this is sentential, not constituent, negation. The rival H52b splits it into a Neg⁰ lágo (a NegP over TP) under a silent directive C⁰; the scope order is the same, and nothing on these pages separates the two (OPEN). No wish operator is present, so nothing binds z₄ (H51), and it is read as “him”. The clause uses the working default: V⁰→v⁰→T⁰, and the subject first merges in left Spec,vP, where it gets its role (VPISH), and raises to left Spec,TP; its lower copy is silent. With T⁰ final (HS-1, HYPOTHESIS) the verb word is pronounced last in TP, so nothing else has to move. Treating the optative mode as a value of T⁰ is itself HYPOTHESIS. LF: DIR_hearer(¬∃e[sleep(e) ∧ Agent(e,z₄)]) Full structure in the list below the figure.CPC′TPC⁰lágoDPAgent: silent fourth person z₄ (the sleeper; not the speaker here)T′vPT⁰DPAgent: silent fourth person z₄ (the sleeper; not the speaker here)v′VPv⁰V′V⁰ʼajoołhááshverb wordhead-movehead-moveMove

Analysis B: (19) NX-418 under H40b: the laanaa clause is the Theme of “think”; =ísh is the matrix C⁰[Q]; the thinker binds the fourth person Hypothesis

Analysis B: (19) NX-418 under H40b: the laanaa clause is the Theme of “think”; =ísh is the matrix C⁰[Q]; the thinker binds the fourth personPanel B. Inner clause: merge V “sleep”; merge v and z₄ in Spec,vP (Agent); merge T (optative); z₄ raises to Spec,TP. Merge C⁰ laanaa (“wishfully”, H52a) on that TP. The CP merges as the complement of V “think” and gets the Theme role; V is final, so the wish clause precedes nínízin. Merge v and the silent second-person thinker in Spec,vP (Experiencer); it raises to the matrix Spec,TP, from where it c-commands z₄ and binds it (H51: the holder of the wish binds the wish clause’s fourth person). C⁰[Q] =ísh is merged last and written on the final verb word. Under H40a (not drawn) the laanaa clause would be a separate root clause with the think clause adjoined to it, and z₄ would be linked to the thinker without c-command, which makes the source’s “reference outside the clause is not possible” a stipulation. The clause uses the working default: V⁰→v⁰→T⁰, and the subject first merges in left Spec,vP, where it gets its role (VPISH), and raises to left Spec,TP; its lower copy is silent. With T⁰ final (HS-1, HYPOTHESIS) the verb word is pronounced last in TP, so nothing else has to move. Treating the optative mode as a value of T⁰ is itself HYPOTHESIS. LF: ASK_POL(∃e1[think(e1) ∧ Exp(e1,you) ∧ Theme(e1, WISH ∃e2[sleep(e2) ∧ Agent(e2,z₄)])]), z₄ = you Full structure in the list below the figure.CPC′TPC⁰=íshDPExperiencer: silent second person “you” (the thinker)T′vPT⁰DPExperiencer: silent second person “you” (the thinker)v′VPv⁰V′CPV⁰nínízinC′TPC⁰laanaaDPAgent: silent fourth person z₄, bound by the thinker (H51)T′vPT⁰DPAgent: silent fourth person z₄, bound by the thinker (H51)v′VPv⁰V′V⁰ʼajoołhááshverb wordverb wordhead-movehead-movehead-movehead-moveAgree Q: yes/no over the whole clauseMoveMove
(1) Optative fourth person: ʼajoołháásh lágo (NX-416) and ʼajoołháásh laanaa nínízin=ísh (NX-418) Hypothesis NX-416/418, Willie 1991 ch. 4 (17b)/(19), printed pp. 120–121: ʼajoołháásh lágo, glossed “4.OPT .sleep NEG”, “Don’t let him go to sleep.”; ʼajoołháásh laanaa nínízin=ísh, glossed “4.OPT.sleep wishfully 2.thinks=Q”, “Do you wish to go to sleep?” Leaves are printed tokens; splitting =ísh off as C⁰ follows the source’s “=Q”. C⁰, Neg⁰, T⁰, v⁰, the silent pronouns and the role labels are our assignments.
Bracket notation (copy)

A: [CP [C′ [TP [DP Agent: silent fourth person z₄ (the sleeper; not the speaker here)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [V⁰ ʼajoołháásh]]] v⁰]] T⁰]] [C⁰ lágo]]] B: [CP [C′ [TP [DP Experiencer: silent second person “you” (the thinker)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [CP [C′ [TP [DP Agent: silent fourth person z₄, bound by the thinker (H51)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [V⁰ ʼajoołháásh]]] v⁰]] T⁰]] [C⁰ laanaa]]] [V⁰ nínízin]]] v⁰]] T⁰]] [C⁰ =ísh]]]

Structure as a list

Analysis A

  • CP
    • C′
      • TP
        • DP (id S1-high; landing site of Move from S1): Agent: silent fourth person z₄ (the sleeper; not the speaker here)
        • T′
          • vP
            • DP (id S1; struck copy; moved to S1-high): Agent: silent fourth person z₄ (the sleeper; not the speaker here)
            • v′
              • VP
                • V′
                  • V⁰ (head; id V1; head-moves to v1): ʼajoołháásh
              • v⁰ (head; id v1; head-moves to T1)
          • T⁰ (head; id T1; hosts the “verb word” (V1 + v1 + T1))
      • C⁰ (head; id C1): lágo

Analysis B

  • CP
    • C′
      • TP
        • DP (id S3-high; landing site of Move from S3): Experiencer: silent second person “you” (the thinker)
        • T′
          • vP
            • DP (id S3; struck copy; moved to S3-high): Experiencer: silent second person “you” (the thinker)
            • v′
              • VP
                • V′
                  • CP (id wishCP)
                    • C′
                      • TP
                        • DP (id S2-high; landing site of Move from S2): Agent: silent fourth person z₄, bound by the thinker (H51)
                        • T′
                          • vP
                            • DP (id S2; struck copy; moved to S2-high): Agent: silent fourth person z₄, bound by the thinker (H51)
                            • v′
                              • VP
                                • V′
                                  • V⁰ (head; id V2; head-moves to v2): ʼajoołháásh
                              • v⁰ (head; id v2; head-moves to T2)
                          • T⁰ (head; id T2; hosts the “verb word” (V2 + v2 + T2))
                      • C⁰ (head; id Cw): laanaa
                  • V⁰ (head; id V3; head-moves to v3): nínízin
              • v⁰ (head; id v3; head-moves to T3)
          • T⁰ (head; id T3; goal of Agree Q: yes/no over the whole clause; hosts the “verb word” (V3 + v3 + T3))
      • C⁰ (head; id CQ; probe for Agree Q: yes/no over the whole clause with T3): =ísh

H53, H50 and H51: what the pages show

ExamplePersons (agent, patient)PredictionPrinted
NX-419 (20a)boy₃, dog₃good (higher-ranked agent c-commands)good
NX-420 (20b)dog₃, boy₃, directbad (H53)*
NX-421 (20c)dog₃, boy₃, inversegood (H3 Move meets H53)good
NX-422/423, 428animal₃, first / fourthgood (not ranked)good
NX-426 (24)horse₃, fourthgood as openergood
NX-427 (25)horse₃, rider₃, inverse, silentbad as opener (H53 + H50)* (as opener)
NX-444 (second 30)boy₃, it₃, directgood on both parses (H4″ only a default)good “in context”
NX-443 (second 28)printed: boy, it; question: it, boyprinted: as NX-444; question reading: bad by H53good “in context” (OPEN)
NX-415 / 416fourth, laanaa / lágobound by speaker / free (H51)“I” / “him”
NX-417 / 418fourth under “think”bound by thinker (H51, c-command under H40b)father / you

Light n

OPEN Nothing on these pages bears on a light n in the noun phrase. The nouns are bare (ʼashkii, mósí, łééʼchąąʼí, séígoʼ) or possessed (Shizhéʼé), and none needs structure inside the noun phrase to get its role or reference. The ranking in H53 is stated over whole arguments.

Telicity test

  • SOURCED The laanaa/lágo contrast is not driven by telicity. NX-415 and NX-416 have the same verb word (and so the same telicity), and the reading changes (wish with the speaker as sleeper, against a prohibition about “him”). Within the pair telicity is held fixed; the conclusion concerns only this contrast.
  • OPEN H51 across verb types. The wisher reading holds with “sleep” (NX-415, 418) and “buy” (NX-417), but these differ in other ways too (embedding, laanaa), so this is not a controlled test.
  • OPEN H53 and the fourth-person substitute. Tested only with “bite”, “kick” and “sting”, all punctual (our classification). Whether the ranking condition holds with an activity or a state verb is untested. The source makes no telicity claim.

Negation

SOURCED Two negatives: the cited Doo hatis njigháa da (NX-236) and the prohibitive ʼajoołháásh lágo (NX-416, glossed NEG). HYPOTHESIS Both are sentential: the negative head takes the whole clause and is pronounced last (H23, H52b). lágo needs no doo. No constituent negation (hanii) is on these pages, so Grant’s sentential/constituent split is not tested further here. A text-layer search of pp. 117–131, checked against the renders, finds no other doo, da or hanii.

Question × polarity × voice grid (cumulative after review 14)

HYPOTHESIS New: polar-affirmative × active gains NX-418, a yes/no question over “think” with an optative fourth person inside (Q at the top, written on the last verb word). Wh-affirmative × active gains NX-429, 431, 433, 435 and 437, all with silent participants. The replies NX-434 and NX-439 are inverse declaratives, and their English passives fill no passive cell. The impersonal NX-236/237 are declaratives with a fourth-person agent, not passives and not questions.

Question cellActive / directInverse / bi-Passive / impersonal (source gloss)
Polar, affirmativeNX-119–134, 142, 203, 204, 291, 307, 418NX-053; *NX-052OPEN (NX-217 candidate only)
Polar, negativeNX-153–161, 184, 205–209, 308 (answer NX-309)OPENOPEN (declaratives NX-216, NX-236 only)
Wh, affirmativeReview 3 set; NX-113, 212, 220, 221, 374–376, 378–385, 429, 431, 433, 435, 437Review 3 setNX-213, 215
Wh, negativeNX-138, 210, 211; NX-247 (by reading)OPENOPEN

OPEN Still empty: polar-affirmative × passive; polar-negative × inverse and × passive; wh-negative × inverse and × passive; any negative passive question; any question with an impersonal agent. Next places to look: Willie ch. 4 §§4.0–4.2.4 (pleonastic ho-, quotative and narrative fourth) and ch. 2 (passives); a question with lágo or with a negated wish.

Where we part from the source’s analyses (labeled contrast)

  • HYPOTHESIS Not adopted: “All third person pronouns … must have a discourse antecedent” (p. 124), contradicted by (24), whose agent “it” has none; H50 limits the requirement to the inverse-moved patient.
  • HYPOTHESIS Not adopted: “a single nominal is always linked to the internal argument” (p. 126), contradicted by her NX-444.
  • HYPOTHESIS Not adopted: the optative fourth is “linked to the speaker … by default because there is no subordinate clause” (p. 121), contradicted by (17b) “him”; H51 ties the binding to the wish, not to the absence of a clause.
  • HYPOTHESIS Restated: the inverse agent is an “internal (Ergative)” argument (p. 123). On the spine the agent raises to Spec,TP in both voices and the patient moves above it; the data here do not separate the two accounts.
  • HYPOTHESIS Restated: the animacy hierarchy as a principle of interpretation is H53, a condition on c-command at spell-out; fourth person “not involved in the animacy hierarchy” is H53’s “not ranked”.

What this changes

  • HYPOTHESIS H50–H53 added. H53 is the first statement of the animacy ranking as what forces the inverse Move (still a stipulation). H51 derives the wisher reading and its absence with lágo.
  • HYPOTHESIS Revised: H32a (exempts the bound fourth person of a wish), H32b → H32b′ (different clauses only), H4″ → H4″-pref (a default, overridable in context), H6′ sharpened by H50 for silent patients.
  • HYPOTHESIS H40a vs H40b: (18)/(19) count for H40b. The tally is now even, not settled.
  • HYPOTHESIS Pass out of sample: H3, H6′ (NX-421, 434, 439), H31 (NX-417, 422, 423, 428, 440, 441), H48 (NX-237). H45 cannot be tested here (no transitive =go pair).
  • Working default (V→v→T, subject to Spec,TP): used in every clause, including the optative ones (treating the optative as a T⁰ value is HYPOTHESIS). Nothing resists it. What else moves: only the inverse patient (NX-421, 427, 434, 439). Objects stay in VP; laanaa, lágo and =ísh are final heads above TP. The one-noun ambiguity of NX-444 is predicted by the default itself.
  • HYPOTHESIS Telicity: the laanaa/lágo change holds with telicity fixed; the rest is OPEN. Grid: +NX-418 (polar), +NX-429/431/433/435/437 (wh), all active.

OPEN predictions

  • OPEN H48: is the ho- of hóótʼiid “happened” a pleonastic subject or a fourth-person pronoun? Willie ch. 4 §4.1 should say.
  • OPEN H50: an inverse clause with a silent patient that was not mentioned but is visible in the scene should be as bad as *(25); a direct clause with a silent agent that was not mentioned should be as good as (24).
  • OPEN H51: a fourth person in a wish clause under a fourth-person thinker, and a wish with an object fourth person: is the object also bound?
  • OPEN H52: lágo with a non-fourth subject, and lágo together with laanaa, to tell H52b from H52c.
  • OPEN H53: a state or activity verb with a lower-ranked agent and a third-person human patient; and the source conflict in NX-443.
  • OPEN H4″-pref: recheck earlier results that used H4″ to exclude a reading.
  • OPEN Grid: still no polar passive question, no negative inverse or negative passive question, and no question with an impersonal agent.

2026-11-15 · First-principles review 15: Willie on pleonastic ho-, nouns with fourth-person verbs, and the polite, quotative and narrative fourth person (ch. 4, §§4.0–4.2.4)

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.

HYPOTHESIS Plain-words verdict. Willie shows that Navajo has a separate ho- for “the area” or “it” in general, a subject that never means a person; so the ho- in “What happened?” is that one, not a fourth person, and H48 is safe. Our rule that nouns go only with third persons (H31) explains most of the pages, including why “water” cannot go with the area ho- verb and why “boy” cannot go with a fourth-person “cried”. But Willie’s own respect, report and storytelling examples put a noun with a fourth-person verb (“my mother”, “some boy”, a coyote), so H31 is wrong as an exceptionless rule. The best repair is that in those uses the ji- is not a person value at all but a mark of respect or of reporting, since a coyote is not a person and the fourth person is said to refer only to humans (H57a, favored but not settled). Two fourth persons in different clauses are not always the same person (“They said that person cried”), so H32b′ holds only when both refer to someone specific. Every clause fits the V-to-T default, and nothing else has to move except, perhaps, the “that Kee ran away” clause in (4). Telicity: the area ho- contrasts hold among states only, and the quotative vs storytelling contrast holds with the same verb “walk”, so neither is about telicity; the rest is OPEN. There are no questions, no negation and no passive on these pages; Willie says (3) is not a passive. The grid does not change.

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–H53 from reviews 1–14 (here mainly H3, H4″-pref, H7, H31, H32a, H32b′, H34, H40a/b, H43, H47, H48), head direction (HS-1, T⁰ final; the derivations are written for HS-1), Agree, and any discourse rule. The live question for word order is what else moves. Every derivation ends in a logical form (LF) whose scope order matches c-command. Silent pronouns are written with their person: x₃, y₃ (third), z₄, w₄ (fourth), u₄ (the unspecified sayer of H43), and a for an areal pronoun. The role labels (Theme, Experiencer, Agent, Patient) are ours. Willie’s analysis (nominals adjoined to pronouns, Condition B, phi-feature matching for adjunction, the fourth person as a deictic) appears only as labeled contrast.

The source’s claims, as its own

  • SOURCED Pleonastic ho- (§4.0–4.1). “the fourth person has both subject and object forms, ji- and ho-; in contrast, there is only one pleonastic argument, the subject pronoun ho-, which has the same shape as the fourth person object pronoun” (pp. 102–103). “while the fourth person refers exclusively to human beings, pleonastic ho- refers to an abstract space, area, or entity” (p. 103). “Pleonastic subjects, like other items that occupy argument positions in Navajo, are pronominal affixes, not nominals” (p. 103). “a bare nominal adjunct, as in (1d), can not be adjoined when pleonastic ho- is present, and a locative postpositional phrase cannot be adjoined when ho- is not present” (p. 105). “The generalization is that a bare nominal can not appear in a sentence without a personal pronoun to which it is adjoined” (p. 107). Of (3): “This example is not a Passive; an enlightening gloss is not easy to find” (p. 107). Of (4): “I analyze the adjoined subordinate clause as an adjunct to the object of the postpositional phrase bee” (p. 108).
  • SOURCED Nouns and the fourth person (§4.2.1). “Third and fourth person pronouns do not match in phi (person) features, so coreference is blocked. Furthermore, nouns are syntactically third person, so a nominal cannot be adjoined to a fourth person pronoun” (p. 109). Against Uyechi (1990), who claims nominals can be linked to both third and fourth pronominals: “This does not occur in my speech, or other speakers I have consulted” (p. 110). Free pronouns that agree in person “give a contrastive or emphatic reading” (p. 110).
  • SOURCED Polite fourth (§4.2.2). “In the presence of a third person referent, ji-/ho- are used to identify the subject and object respectively” (p. 111). “where the speaker wants to express respect, it is possible to adjoin a nominal to a fourth person pronoun. One frequently hears this when the speaker is referring to his mother” (p. 113); “This usage does not seem to occur with other kin terms” (p. 113).
  • SOURCED Quotative and narrative (§§4.2.3–4.2.4). In (9)/(10) “the fourth person in the main clause conveys a quotative. The speaker of these utterances is signaling that he does not have direct knowledge of the events” (p. 114). (11), with the fourth person on every verb, “is a Reportative”; in (12) “the fourth person pronoun in the final verb is not a reportative, but refers to a specific person who is present or nearby, and is used in deference” (pp. 115–116). The léiʼ of (11) “is needed … its omission would produce an ungrammatical sentence, since the referents are not presupposed” (p. 116; the version without it is not printed). “Aside from these special contexts, the normal course of events is that only third person will permit nominal adjunction” (p. 116). “There are certain conventions employed in narrative texts where the fourth person permits nominal adjunction” (pp. 116–117); (14) “introduces a traditional story” (p. 117).

The checked data

SOURCED Exact forms, source glosses and translations, with example numbers and printed pages. Glosses list object before subject (jiztał “3.4.kicked” is “She kicked him”). Printing notes and crops are at the linked index entries.

  • Pleonastic ho- with “deep” (pp. 103–104). NX-445 (1a): ditą́ “3-is deep”, “It is deep.” · NX-446 (1b): tó ditą́ “water, 3-is deep”, “The water is deep.” · NX-447 (1c): hoditą́ “ho-is deep”, “It (the area) is deep.” · NX-448 *(1d): tó hoditą́ “water ho-is deep” · NX-449 (1e): táyiʼ hoditą́ “water:in ho-is deep”, “It is deep there in the water.” “The water is deep there.” · NX-450 *(1f): táyiʼ ditą́ “water:in 3-is deep”.
  • Psych verbs (pp. 105–107). NX-451 (2a): shił nizhóní “1sg.with 3.nice”, “With me, it is nice” “I like it.” · NX-452 (2b): shił hózhóní “1sg.with ho.nice”, “I like it(the area).” · NX-453 *(2c): kintah shił nizhóní “town 1sg.with 3.is nice”, “I like the town.” · NX-454 (2d): kintah-di shił hózhóní “town -at ho.is nice”, “I like it in town.” · NX-455 *(2e): Kintah-di shił nizhóní “town-at 1sg.with 3 .is nice”, “I like it in town [being in town].” (starred on that reading; the prose gives a good reading “I like him/it [when I’m] ]in town, [not here where I am now]”) · NX-456 (3): Shił béé hózin “1.with 3.about ho.known”, “It is known to me by means of it.)” “I know it.” · NX-457 (4): Kee yóóʼeelwod=ígíí bił béé hózin “3.ran away=COMP 3.with 3.means of ho.known”, “He/she knows that Kee ran away.”
  • Nouns and free pronouns (pp. 109–111); no gloss lines. NX-458 *(5a): ʼatʼééd, yicha “[I, as a girl, I cried]” · NX-459 *(5b): ʼatʼééd, yínícha “[You, as a girl, you cried].” · NX-460 *(5c): ʼatʼééd, jíícha “[That person, as a girl, cried]” · NX-461 (6a): Shí yícha “I, I cried.” · NX-462 (6b): Ní yínícha “You, you cried.” · NX-463 (6c): Bí, yícha “Him, he cried.” · NX-464 (6d): ʼashkii yícha “The boy, he cried.” · NX-465 (6e): Hó, jíícha “That certain person, he cried.” · NX-466 *(6f): ʼashkii jíícha (no translation).
  • Polite and respect fourth (pp. 112–113). NX-467 (7a): Jíícha “4. cried”, “That certain person cried.” · NX-468 (7b): John jiztał “3.4.kicked”, “She kicked him, John.” · NX-469 (7c): John haztał “4.3.kicked”, “John kicked that certain person.” · NX-470 (8): shimá ʼéí Baaʼ joolyé “1.mother 4.called”, “My mother is called Baaʼ.”
  • Quotative, reportative and narrative (pp. 114–117). NX-471 (9): yicha jiní “3s.crying 4s.said”, “They said he is crying.” · NX-472 (10): jíícha jiní “4s.cried 4s.said”, “(They) said that person cried.” · NX-473 (11): ʼashkii léiʼ ʼatʼééd bił jooʼash=go náhodiiʼnil łá jiní “boy a girl s3.with 4s.walking=when 4s.picked up(PL) EMP 4s.said”, “(I heard) some boy was walking with a girl when they were picked up (presumably by the police).” · NX-474 (12): ʼashkii ʼatʼééd yił yiʼash =go nábiʼdiiʼnil łá jiní (printed with a space before =go; gloss words do not line up one-to-one) “boy girl 3s.with 3DUAL.walk=when 3.picked up(PL) EMP j4s.said”, “That person told me that a boy and girl were picked up while walking.” · NX-475 (13): mąʼii yiildlosh jiní “coyote 3.walking 4.said”, “(I hear) that there is a coyote around here.” · NX-476 (14): mąʼii jooldloosh jiní “4.walking 4.said”, “Once upon a time there was this coyote walking along...”

Hypotheses tested, revised or added

  • HYPOTHESIS H54 · areal ho- (new). H54a: the subject ho- of (1c), (1e), (2b), (2d) spells a silent areal pronoun a, a real argument with a role (the Theme of “deep”, “nice”), which is neither third nor fourth person. H54b: the ho- of “know” (3), (4) has no referent and no role; since VPISH gives roles in Spec,vP, a role-less subject would merge directly in Spec,TP. Rival H54′: one item throughout, always an areal argument with a role (in (3) “the situation”). H54c: a place phrase (táyiʼ, kintah-di) can restrict the areal argument; with “deep” it is licensed only that way.
  • HYPOTHESIS H55 · place nouns (new, a stipulation). kintah “town” names a place and can be linked only to an areal argument or stand inside a place phrase. Proposed only to derive *(2c) (below); nothing else supports it.
  • HYPOTHESIS H56 · the “that” clause of “know” (new). In (4) the =ígíí clause stands first, before bił, though it is understood as the object of béé. H56a: it is merged high (adjoined to TP) and linked by Agree to the third-person object of -ee, with no Move. H56b: it is merged as the object of -ee and moves leftward out of the place phrase over bił (one extra Move, like review 3’s H3-loc extraction from a postposition phrase).
  • HYPOTHESIS H57 · nouns with ji- in respect, report and story (new). H57a: in (8), (11) and (14) the ji-/ho- on a verb whose argument is a noun does not spell a fourth-person value of that argument; it spells, by Agree, a feature of a higher head (respect, or report/narrative mode), and the argument itself is third person, so H31 is kept. H57b: H31 simply has listed exceptions (a respected mother; reportative and narrative contexts).
  • HYPOTHESIS H32b″ (revision of H32b′). Two fourth persons in different clauses of one sentence refer to the same person when both refer to someone specific. H43’s unspecified sayer (nobody in particular) and H57a’s respect/report marks have no referent of their own and are exempt.
  • HYPOTHESIS Tested again, unchanged: H31 (nouns and enclitics are third person and link only to third persons), H43 (the fourth-person subject of jiní is a specific respected person or an unspecified sayer), H34 (a nominalized clause names a fact), H47 (=go clause above the main Spec,TP), H48 (events are third person; fourth person only persons).

Derivations, step by step

  1. “Deep”: NX-445–450 (1a–f). (1a) ditą́: merge V “be deep”; merge v and a silent x₃ in Spec,vP (Theme); merge T; V→v→T, x₃ raises to Spec,TP. LF: ∃s[deep(s) ∧ Theme(s,x₃)]. (1b): the same, with the noun tó as the Theme (third person, so H31 allows it); it raises to Spec,TP and is pronounced first. LF: ∃s[deep(s) ∧ Theme(s,water)]. (1c) hoditą́: the Theme is the areal pronoun a, spelled ho- (H54a). LF: ∃s[deep(s) ∧ Theme(s,a)], a = the area in view. *(1d) tó hoditą́: the only argument is a, which is not third person; the noun has no third-person participant to be linked to (H31), so it is starred, as printed. (1e) táyiʼ hoditą́: as (1c), and the place phrase is adjoined to TP and restricts a (H54c). LF: ∃s[deep(s) ∧ Theme(s,a) ∧ in(a,water)]; the two printed translations (“deep there in the water”, “the water is deep there”) are two English renderings of this one LF. *(1f) táyiʼ ditą́: with a third-person Theme there is no areal argument for the place phrase to restrict; starred, as printed, if H54c holds for “deep”. Result: H31 + H54a derive *(1d) without any extra rule; H54c derives *(1f). Willie’s “a bare nominal can not appear … without a personal pronoun to which it is adjoined” has the same effect for (1d) once ho- is not a “personal” pronoun. The tree, panel A, draws (1e).
  2. “Nice”: NX-451–455 (2a–e). (2a) shił nizhóní: merge V “be nice” with the place phrase shił “1sg.with” as its complement (Experiencer, the postposition’s object, as Willie says); merge v and a silent x₃ in Spec,vP (Theme); default. The verb word is last; the Experiencer phrase stays in VP. LF: ∃s[nice(s) ∧ Theme(s,x₃) ∧ Exp(s,me)]. (2b): Theme = a (H54a). (2d) kintah-di shił hózhóní: as (2b), with kintah-di adjoined to TP, restricting a (H54c). LF: ∃s[nice(s) ∧ Theme(s,a) ∧ Exp(s,me) ∧ at(a,town)]. *(2e) on the printed reading: no areal argument to restrict, as *(1f). Its good reading in the prose (“I like him/it [when I’m] in town”) has the place phrase as an adjunct that locates the liking, not the liked thing: ∃s[nice(s) ∧ Theme(s,x₃) ∧ Exp(s,me) ∧ at(s,town)]. So with “nice” a place phrase can be a plain adjunct; H54c’s “only that way” cannot be general. OPEN Why *(1f) has no such adjunct reading (“it is deep, in the water”) is NOT UNDERSTOOD. *(2c) kintah shił nizhóní “I like the town” is NOT UNDERSTOOD: H31 lets the third-person noun kintah be the Theme, raised to Spec,TP, exactly as tó is in (1b); the order (noun, Experiencer phrase, verb) is the default order, and nothing else would have to move. So H31 + the default predict it good. Willie’s generalization does not exclude it either: (2c) has a third-person pronoun (the subject of nizhóní) for kintah to be adjoined to. Only the stipulation H55 (place nouns go only with the areal argument or inside a place phrase) derives it; one might suspect that kintah is itself built on a postposition (our guess about its make-up, not the source’s gloss, which is “town”), but nothing on these pages shows it.
  3. “Know”: NX-456 (3), NX-457 (4), H54b and H56. (3) Shił béé hózin: merge V “be known” with two place phrases, béé “3.about” (Theme, the postposition’s third-person object y₃) and shił “1.with” (Experiencer), the Experiencer phrase above the Theme phrase, as the order shows. Under H54b nothing merges in Spec,vP (no argument with an external role) and the role-less ho- pronoun merges directly in Spec,TP; V→v→T as usual. Under H54′ the areal a merges in Spec,vP with a role and raises by the default. The order is the same either way. LF: ∃s[known(s) ∧ Exp(s,me) ∧ Theme(s,y₃)] (the expletive adds nothing; under H54′ add a as the “situation”). Willie says (3) is not a passive; we agree: no agent is demoted and no voice head is needed. (4) Kee yóóʼeelwod=ígíí bił béé hózin: the same, with a third-person Experiencer (bił “3.with”, x₃) and the =ígíí clause as what is known. Inside that clause: merge V “run away”, merge v and Kee in Spec,vP (Agent; label ours), default; =ígíí (glossed COMP) closes it. Both bił and béé have third-person objects, so H31 allows the clause to link to either; only the role decides (a proposition cannot be the one who knows), as H4′’s role-compatibility override did. The clause stands before bił, so it is not inside the béé phrase at spell-out: H56a (adjoined to TP, linked by Agree, no Move) or H56b (one leftward Move out of the phrase). Nothing on these pages separates them (OPEN). LF: ∃s[known(s) ∧ Exp(s,x₃) ∧ Theme(s,p)], p = the fact that ∃e[run.away(e) ∧ Agent(e,Kee)] (H34; “know” takes a fact). The clause has no quantifier or operator that could take scope over “know”, so its high position under H56a creates no scope mismatch. Result: H34 and H31 pass out of sample; the only possible extra Move on these pages is H56b.
  4. Nouns and free pronouns: NX-458–466 (5), (6). *(5a)–*(5c): ʼatʼééd “girl” with a first-, second- or fourth-person “cried”: the noun is third person and the only participant is not (H31), so all three are starred, as printed. (6d) ʼashkii yícha: the noun is the third-person Agent (our label) in Spec,vP, raised to Spec,TP. LF: ∃e[cry(e) ∧ Agent(e,boy)]. *(6f) ʼashkii jíícha: as *(5c). (6a–c, e): the free pronouns Shí, Ní, Bí, Hó carry person, so each matches the verb’s subject prefix: merge the overt pronoun in Spec,vP (Agent), raise it to Spec,TP; the prefix is its agreement. LF (6e): ∃e[cry(e) ∧ Agent(e,z₄)], with contrast on z₄. So the restriction is on nouns, not on overt subjects: Hó goes with jíícha, ʼashkii does not. Whether the contrastive pronoun moves on to Spec,CP (review 2’s H7, focus fronting) cannot be seen, since the subject is first either way (OPEN). Result: H31 passes out of sample on nine examples, and Willie’s “nouns are syntactically third person” is the same idea in her terms.
  5. Polite fourth: NX-467–469 (7). (7a) Jíícha: z₄ (Agent) in Spec,vP, raised to Spec,TP; z₄ = the person within hearing. LF: ∃e[cry(e) ∧ Agent(e,z₄)]. (7b) John jiztał “She kicked him, John”: John is merged as the complement of V (Patient), the only third-person participant (H31); the agent is a silent z₄ in Spec,vP, raised to Spec,TP. LF: ∃e[kick(e) ∧ Agent(e,z₄) ∧ Patient(e,John)]. (7c) John haztał “John kicked that certain person”: now the patient is the fourth person w₄ (ha-) and John is the agent, in Spec,vP and then Spec,TP. LF: ∃e[kick(e) ∧ Agent(e,John) ∧ Patient(e,w₄)]. Result: the same string order, noun + verb, with opposite roles, decided by which participant is third person (tree, panels B and C). The working default yields both with nothing else moving. H53 (ranking among third persons) does not apply: there is only one third person. This fits Willie’s p. 126 statement (review 14) that with a 1, 2 or 4 argument a single noun can be either argument.
  6. Respect: NX-470 (8), against H31. shimá ʼéí Baaʼ joolyé “My mother is called Baaʼ”: the noun phrase shimá ʼéí is the subject of a verb glossed “4.called”. As stated, H31 forbids this: H31 is contradicted as an exceptionless rule, as Willie herself says (“it is possible to adjoin a nominal to a fourth person pronoun” for respect). Under H57a: merge V “be called” with the name Baaʼ as its complement (a name, not a participant; our analysis); merge v and the third-person subject shimá ʼéí in Spec,vP (Theme; label ours), raised to Spec,TP. The ji- spells a respect feature that the subject carries by Agree with a higher head, not a fourth-person value. LF: ∃s[called(s) ∧ Theme(s,my.mother)] with the name Baaʼ, plus respect for the mother (not part of what is asserted). Under H57b it is a listed exception (only with “mother”, p. 113).
  7. Quotative: NX-471 (9), NX-472 (10), NX-474 (12) and H43. (9) yicha jiní: inner clause: V “cry”, x₃ in Spec,vP (Agent), default. Outer: V “say” with that clause as what is said (H40b; under H40a the clause is a separate root and the object of “say” is a pronoun resolved to it; the order is the same), and the unspecified sayer u₄ in Spec,vP (H43), raised to Spec,TP. LF: ∃e1[say(e1) ∧ Agent(e1,u₄) ∧ Theme(e1, ∃e2[cry(e2) ∧ Agent(e2,x₃)])], u₄ = nobody in particular (“They said”). (12) has the same person pattern (third person in the reported clauses, fourth only on jiní), but the sayer is a specific person nearby, spoken of with deference: z₄ in place of u₄, “That person told me”. Result: H43 passes out of sample, and the person pattern alone does not mark deference: the same pattern is quotative in (9) and deferential in (12), so Willie’s contrast of (11) with (12) is really fourth person on every verb vs on “say” only, with “say” only open to both readings. (10) jíícha jiní “(They) said that person cried”: inner z₄ (Agent of “cry”), outer u₄. The crier and the sayers differ, so H32b′ as stated is contradicted (two fourth persons in different clauses, different people). H32b″ exempts u₄, which refers to no one in particular. LF: ∃e1[say(e1) ∧ Agent(e1,u₄) ∧ Theme(e1, ∃e2[cry(e2) ∧ Agent(e2,z₄)])], z₄ ≠ u₄. The cases that supported H32b′ in reviews 10, 12 and 13 had specific fourth persons in both clauses, so they are unaffected.
  8. Reportative and narrative: NX-473 (11), NX-475 (13), NX-476 (14), and H57. (13) mąʼii yiildlosh jiní: as (9), with mąʼii as the third-person Agent (our label) of “walk”. (14) mąʼii jooldloosh jiní “Once upon a time there was this coyote walking along…”: the same, but “walk” carries the fourth person while its subject is the noun mąʼii: H31 is contradicted again as exceptionless. And the fourth person “refers exclusively to human beings” (p. 103), yet here it goes with a coyote. That favors H57a: the ji- of jooldloosh spells the narrative mode, by Agree, on a third-person subject; it is not the human fourth-person pronoun. LF (14): ∃e1[say(e1) ∧ Agent(e1,u₄) ∧ Theme(e1, ∃x[coyote(x) ∧ ∃e2[walk(e2) ∧ Agent(e2,x)]])], as a story opener. (A coyote in a traditional story may count as a person, so the argument is suggestive, not decisive.) (11): the =go clause ʼashkii léiʼ ʼatʼééd bił jooʼash=go is merged above the main Spec,TP (H47): inside it the noun phrase ʼashkii léiʼ is the Agent of “walk” (a verb glossed “4s”) and ʼatʼééd is the third-person object of bił “s3.with”. Main clause náhodiiʼnil “4s.picked up(PL)”: the picked-up pair is a silent pronoun linked to the boy and girl; the one who picked them up is silent and unspecified (“presumably by the police”). Which argument the “4s” of náhodiiʼnil marks, the gloss does not say (OPEN). Then łá (glossed EMP) and jiní with u₄. LF: ∃e0[say(e0) ∧ Agent(e0,u₄) ∧ Theme(e0, ∃x[boy(x) ∧ ∃e1[walk(e1) ∧ Agent(e1,x) ∧ with(e1,girl)] ∧ ∃e2[pick.up(e2) ∧ Agent(e2,v) ∧ Patient(e2,x⊕girl) ∧ e2 during e1]])], v unspecified. The postposition object of bił stays third person even here, so under H57a the report mark reaches only the verb word’s own arguments (OPEN). Willie’s claim that (11) needs léiʼ because the referents are new fits review 13’s H45 (an indefinite may introduce someone new), but the version without léiʼ is not printed, so we do not use it as data.

Analysis A: (1e) NX-449 táyiʼ hoditą́: the areal ho- argument is the Theme of “deep”; the place phrase restricts which area (H54a, H54c) Hypothesis

Analysis A: (1e) NX-449 táyiʼ hoditą́: the areal ho- argument is the Theme of “deep”; the place phrase restricts which area (H54a, H54c)Panel A. Merge V “be deep”; merge v and the silent areal pronoun a in Spec,vP, which gets the Theme role (VPISH; the role label is ours) and is spelled by ho- on the verb word. Merge T; a raises to Spec,TP. The place phrase táyiʼ “water:in” is adjoined to TP on the left and restricts a to the area in the water (H54c). Without ho- (NX-450, *táyiʼ ditą́) there is no areal argument for the place phrase to restrict, and without the place phrase a bare noun has nothing to link to (NX-448, *tó hoditą́): a noun is third person (H31) and the areal pronoun is not (H54a). This is the working default: V⁰→v⁰→T⁰, and the subject first merges in left Spec,vP, where it gets its role (VPISH), and raises to left Spec,TP; its lower copy is silent. With T⁰ final (HS-1, HYPOTHESIS) the verb word is pronounced last in TP. Nothing else moves. LF: ∃s[deep(s) ∧ Theme(s,a) ∧ in(a,water)] Full structure in the list below the figure.TPPPtáyiʼ “water:in” (one printed word; adjunct linked to a, H54c)TPDPTheme: silent areal pronoun a (the area), spelled ho- (H54a)T′vPT⁰DPTheme: silent areal pronoun a (the area), spelled ho- (H54a)v′VPv⁰V′V⁰hoditą́verb wordhead-movehead-moveMove

Analysis B: (7b) NX-468 John jiztał “She kicked him, John”: the fourth-person agent is silent; the noun is the only third-person participant, the patient Hypothesis

Analysis B: (7b) NX-468 John jiztał “She kicked him, John”: the fourth-person agent is silent; the noun is the only third-person participant, the patientPanel B. Merge V “kick” with John as its complement (Patient); merge v and the silent fourth-person z₄ in Spec,vP (Agent), spelled ji-. Agree pairs the object prefix with John (third person) and the subject prefix with z₄. Merge T; z₄ raises to Spec,TP. Because z₄ is silent, John is pronounced first even though it stays in VP. This is the working default: V⁰→v⁰→T⁰, and the subject first merges in left Spec,vP, where it gets its role (VPISH), and raises to left Spec,TP; its lower copy is silent. With T⁰ final (HS-1, HYPOTHESIS) the verb word is pronounced last in TP. Nothing else moves. LF: ∃e[kick(e) ∧ Agent(e,z₄) ∧ Patient(e,John)] Full structure in the list below the figure.TPDPAgent: silent fourth person z₄ (ji-; the person in hearing distance)T′vPT⁰DPAgent: silent fourth person z₄ (ji-; the person in hearing distance)v′VPv⁰V′DPPatient: John (third person, so it can be linked here, H31)V⁰jiztałverb wordhead-movehead-moveMove

Analysis C: (7c) NX-469 John haztał “John kicked that certain person”: the same string order with the roles reversed, because now the agent is the third person Hypothesis

Analysis C: (7c) NX-469 John haztał “John kicked that certain person”: the same string order with the roles reversed, because now the agent is the third personPanel C. Merge V “kick” with the silent fourth-person w₄ as its complement (Patient), spelled ha-; merge v and John in Spec,vP (Agent). Merge T; John raises to Spec,TP. The surface string is noun + verb, as in panel B, but John is now in Spec,TP rather than in VP; which parse holds is decided by the person prefixes, not by order. This is the working default: V⁰→v⁰→T⁰, and the subject first merges in left Spec,vP, where it gets its role (VPISH), and raises to left Spec,TP; its lower copy is silent. With T⁰ final (HS-1, HYPOTHESIS) the verb word is pronounced last in TP. Nothing else moves. LF: ∃e[kick(e) ∧ Agent(e,John) ∧ Patient(e,w₄)] Full structure in the list below the figure.TPDPAgent: John (the only third-person participant, H31)T′vPT⁰DPAgent: John (the only third-person participant, H31)v′VPv⁰V′DPPatient: silent fourth person w₄ (ha-)V⁰haztałverb wordhead-movehead-moveMove
(1) Areal ho- and the polite fourth person: táyiʼ hoditą́ (NX-449), John jiztał (NX-468), John haztał (NX-469) Hypothesis NX-449, NX-468, NX-469: Willie 1991 ch. 4 (1e), (7b), (7c), printed pp. 104 and 112: táyiʼ hoditą́, glossed “water:in ho-is deep”, “It is deep there in the water. / The water is deep there.”; John jiztał, glossed “3.4.kicked”, “She kicked him, John.”; John haztał, glossed “4.3.kicked”, “John kicked that certain person.” Leaves are printed tokens. T⁰, v⁰, the silent pronouns, the adjunction site of the place phrase and the role labels are our assignments.
Bracket notation (copy)

A: [TP [PP táyiʼ “water:in” (one printed word; adjunct linked to a, H54c)] [TP [DP Theme: silent areal pronoun a (the area), spelled ho- (H54a)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [V⁰ hoditą́]]] v⁰]] T⁰]]] B: [TP [DP Agent: silent fourth person z₄ (ji-; the person in hearing distance)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [DP Patient: John (third person, so it can be linked here, H31)] [V⁰ jiztał]]] v⁰]] T⁰]] C: [TP [DP Agent: John (the only third-person participant, H31)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [DP Patient: silent fourth person w₄ (ha-)] [V⁰ haztał]]] v⁰]] T⁰]]

Structure as a list

Analysis A

  • TP
    • PP: táyiʼ “water:in” (one printed word; adjunct linked to a, H54c)
    • TP
      • DP (id S1-high; landing site of Move from S1): Theme: silent areal pronoun a (the area), spelled ho- (H54a)
      • T′
        • vP
          • DP (id S1; struck copy; moved to S1-high): Theme: silent areal pronoun a (the area), spelled ho- (H54a)
          • v′
            • VP
              • V′
                • V⁰ (head; id V1; head-moves to v1): hoditą́
            • v⁰ (head; id v1; head-moves to T1)
        • T⁰ (head; id T1; hosts the “verb word” (V1 + v1 + T1))

Analysis B

  • TP
    • DP (id S2-high; landing site of Move from S2): Agent: silent fourth person z₄ (ji-; the person in hearing distance)
    • T′
      • vP
        • DP (id S2; struck copy; moved to S2-high): Agent: silent fourth person z₄ (ji-; the person in hearing distance)
        • v′
          • VP
            • V′
              • DP: Patient: John (third person, so it can be linked here, H31)
              • V⁰ (head; id V2; head-moves to v2): jiztał
          • v⁰ (head; id v2; head-moves to T2)
      • T⁰ (head; id T2; hosts the “verb word” (V2 + v2 + T2))

Analysis C

  • TP
    • DP (id S3-high; landing site of Move from S3): Agent: John (the only third-person participant, H31)
    • T′
      • vP
        • DP (id S3; struck copy; moved to S3-high): Agent: John (the only third-person participant, H31)
        • v′
          • VP
            • V′
              • DP: Patient: silent fourth person w₄ (ha-)
              • V⁰ (head; id V3; head-moves to v3): haztał
          • v⁰ (head; id v3; head-moves to T3)
      • T⁰ (head; id T3; hosts the “verb word” (V3 + v3 + T3))

H31, H54 and H57: what the pages show

ExampleNoun or phrase · argument it would link toPredictionPrinted
NX-446 (1b)tó · third-person Themegood (H31)good
NX-448 (1d)tó · areal abad (H31 + H54a)*
NX-449 (1e)place phrase · areal agood (H54c)good
NX-450 (1f)place phrase · no areal argumentbad (H54c with “deep”)*
NX-453 (2c)kintah · third-person Themegood by H31 (bad only with H55)* (NOT UNDERSTOOD)
NX-455 (2e)place phrase · none / the likingbad on the area reading; good as an adjunct* on the area reading; good reading in prose
NX-458–460, 466noun · 1st, 2nd or 4th personbad (H31)*
NX-461–465free pronoun or noun · matching persongoodgood
NX-468 / 469John · the only third person (patient / agent)good, opposite rolesgood, as translated
NX-470 (8)shimá ʼéí · subject of a “4.called” verbbad by H31; good by H57agood (respect)
NX-473 (11), NX-476 (14)ʼashkii léiʼ, mąʼii · subject of a fourth-person verbbad by H31; good by H57agood (report, story)

H48 and the ho- of hóótʼiid

HYPOTHESIS Review 14 left H48 (events are third person; the fourth person is only persons) at risk if the ho- of NX-437 haash hóótʼiid “What happened?” (and of review 11’s NX-385 ʼáhóótʼiid) were a fourth person. On Willie’s own classification it cannot be: the fourth person subject is ji-, the fourth person refers “exclusively to human beings”, and the only subject ho- is the pleonastic one, which “refers to an abstract space, area, or entity” (pp. 102–103). “What happened?” has no human participant, and its ho- would be the subject, so it is the areal/pleonastic ho- (H54a or H54b). Result: H48 is no longer at risk. These pages do not discuss hóótʼiid, so the assignment is our inference (OPEN until a source segments it).

Light n

OPEN Nothing here bears on a light n. Two facts may need structure inside the noun phrase, and we say so plainly: léiʼ in ʼashkii léiʼ (glossed “a”, needed for a new referent, p. 116) and ʼéí in shimá ʼéí (unglossed). Whether they sit inside the noun phrase or outside it is OPEN. The possessor of shimá “1.mother” would be the natural place to test a light n (possessor in a specifier, as the subject is in Spec,vP), but these pages give no contrast that needs it.

Telicity test

  • SOURCED The areal ho- contrasts are not driven by telicity. All of (1)–(4) are states (“deep”, “nice”, “known”; our classification), and the contrasts between ho- and third person, and between a bare noun and a place phrase, hold among them. Telicity is held fixed; the conclusion concerns only these contrasts.
  • SOURCED The quotative vs story-opener contrast is not driven by telicity. (13) and (14) have the same verb “walk” (glossed “walking” in both; an activity, our classification) and differ only in the person of its prefix; the reading changes from a report (“I hear …”) to a story opener.
  • OPEN (9) yicha “3s.crying” vs (10) jíícha “4s.cried” differ in person and in aspect (imperfective vs perfective, by the glosses). Perfective viewpoint is not telicity, so this is no test. H31 is tested only with “cry”, “kick”, “walk” and “be called” (OPEN across verb types).

Negation

SOURCED None. A text-layer search of pp. 102–117, checked against the renders, finds no doo, da, lágo or hanii. The sentential vs constituent split is not tested here.

Question × polarity × voice grid (cumulative after review 15)

HYPOTHESIS No change. These pages have no question (no -ísh/-sh, daʼ, háí, haʼátʼíí, haash) and no negation. No passive either: Willie says (3) “is not a Passive”, and the English “were picked up” of (11) and (12) fills no Navajo passive cell (the source glosses are “4s.picked up(PL)” and “3.picked up(PL)”, with no passive label).

Question cellActive / directInverse / bi-Passive / impersonal (source gloss)
Polar, affirmativeNX-119–134, 142, 203, 204, 291, 307, 418NX-053; *NX-052OPEN (NX-217 candidate only)
Polar, negativeNX-153–161, 184, 205–209, 308 (answer NX-309)OPENOPEN (declaratives NX-216, NX-236 only)
Wh, affirmativeReview 3 set; NX-113, 212, 220, 221, 374–376, 378–385, 429, 431, 433, 435, 437Review 3 setNX-213, 215
Wh, negativeNX-138, 210, 211; NX-247 (by reading)OPENOPEN

OPEN Still empty: polar-affirmative × passive; polar-negative × inverse and × passive; wh-negative × inverse and × passive; any negative passive question; any question with an impersonal agent. Next places to look: Willie ch. 2 (passives, §§2.1–2.5) and ch. 7.

Where we part from the source’s analyses (labeled contrast)

  • HYPOTHESIS Restated: “nouns are syntactically third person, so a nominal cannot be adjoined to a fourth person pronoun” (p. 109) is H31; we treat the noun as the argument itself, linked to the verb’s prefix by Agree, not as an adjunct to a pronoun.
  • HYPOTHESIS Not adopted as stated: “a bare nominal can not appear in a sentence without a personal pronoun to which it is adjoined” (p. 107). It derives *(1d) but not *(2c), which has such a pronoun.
  • HYPOTHESIS Restated: the respect, reportative and narrative “exceptions” (pp. 113–117) are not exceptions to H31 under H57a, because the ji- there is not a person value. Willie treats them as nominals adjoined to fourth-person pronouns in special contexts; that is our H57b.
  • HYPOTHESIS Refined: Willie contrasts (11) (reportative) with (12) (deference) by whether the earlier verbs carry the fourth person. Her own (9) has (12)’s pattern with the quotative reading, so only “fourth person on every verb” is diagnostic.
  • HYPOTHESIS Kept open: Willie analyses the clause of (4) as an adjunct to the object of bee; that is our H56a, against H56b (Move out of the phrase).

What this changes

  • HYPOTHESIS H54–H57 added (areal ho-; the place-noun stipulation for *(2c); the position of the “know” clause; nouns with ji- in respect, report and story). Next new number: H58.
  • HYPOTHESIS H31 contradicted as exceptionless by NX-470, NX-473, NX-476; kept under H57a (favored by the coyote of (14)), or with listed exceptions (H57b). It passes on NX-446, 448, 458–466, 468 and 469; NX-453 *(2c) is NOT UNDERSTOOD.
  • HYPOTHESIS H32b′ → H32b″ (specific fourth persons only), forced by NX-472.
  • HYPOTHESIS H48 no longer at risk (the ho- of hóótʼiid is the areal one, on Willie’s classification). H43, H34 and H47 pass out of sample.
  • HYPOTHESIS H40a vs H40b: nothing here decides. łá (glossed EMP) stands between the reported material and jiní in (11) and (12), but whether it belongs to the reported clause or to the “say” clause is not said, so it is not counted. The tally stays even.
  • Working default (V→v→T, subject to Spec,TP): used in every clause; nothing resists it. What else moves: nothing, except possibly the =ígíí clause of (4) under H56b. One case the default does not cover: under H54b a role-less ho- merges directly in Spec,TP with no raising (OPEN, not a counterexample; under H54′ the default applies as usual).
  • HYPOTHESIS Telicity: the areal contrasts (states only) and the quotative/story contrast (same verb) hold with telicity fixed; the rest is OPEN. Grid: unchanged.

OPEN predictions

  • OPEN H55: kintah as a bare argument should be bad with any verb, including the ho- verb of (2d) without -di. If it is good anywhere as a plain third-person argument, H55 fails and *(2c) needs another account.
  • OPEN H54b vs H54′: can the ho- of “know” be restricted by a place phrase, as the areal one of “deep” can? If yes, it has a role (H54′).
  • OPEN H54c: why “deep” rejects an adjunct place phrase (*(1f)) while “nice” takes one on the “when I’m in town” reading (*(2e)).
  • OPEN H56a vs H56b: a “know” clause with the =ígíí clause after bił, right before béé; and an island inside the clause.
  • OPEN H57a vs H57b: a reportative or story sentence with a noun linked to a fourth-person object; and the respect use with a non-human or with a kin term other than “mother” (Willie says it does not occur with other kin terms, which H57a must explain as a lexical fact about respect).
  • OPEN H32b″: a sentence with a specific respected sayer and a fourth person in the reported clause (should be the same person); compare review 12’s H42.
  • OPEN Carried over: H50, H51, H52b/c, H53 tests and the NX-443 conflict from review 14; recheck of results that used H4″ to exclude readings; the grid cells above.