Authority caveat. Nobody on this site claims to know Navajo authoritatively. Every Navajo form is copied from a named published source with its example number; none is invented. Formal rules are hypotheses to be tested against Grant's texts.

SOURCED stated in a published source we opened HYPOTHESIS our formalization or inference GRANT'S CONJECTURE Grant's own claim OPEN needs data or reading

One structure for Navajo: the clause, the noun phrase and the verb word on a single spine

Working note (newest). Not yet filed into a section; listed under Working notes.

Verbs on this page → dictionary: yiztał ‘kick’ · dabiiłtsą́ ‘see’ · yi-dloh ‘laugh’ · neezdá ‘sit down; make sit; keep (somewhere)’ · joolyé ‘be called (named)’ · ditą́ʼ ‘be deep; (with areal ho-) the area is deep’ · yishááł ‘I am walking (1sg, progressive)’ · neezdá ‘sat down; with yikáá’ ‘on it’: sat down on it’ · déyá ‘I am going; with noun=góó: I am going (to a place)’

2026-11-15 · new page. Grant asked (2026-11-15) for the re-evaluation of older content to speed up under VPISH and the V→v→T default (“something I would normally assume”) and to “use this as an opportunity to create unified structure for the language”. This page is that structure. It adds no Navajo data: every example is copied from the example index under its NX number with its printed judgment and translation, and every hypothesis number (H3, H23 …) is the first-principles reviews’ own. Where a review and this page differ, the review is the record. Grant’s corrections of the same evening on the verb word (headedness as a result of Merge sites; a possible sublexical V-to-T) are set out in §3. Older pages now carry a dated “Restated on the unified structure” note pointing here (§7).

Navajo is described here with one clause spine, one nominal structure and one account of the verb word, and every phenomenon the site has studied is placed on them: where it sits, what moves (beyond what always moves), what the logical form is, how sure we are, which examples show it, and what still does not fit. Labels: SOURCED what a source prints; HYPOTHESIS our analysis; GRANT'S HYPOTHESIS Grant’s own proposal, tested here; OPEN undecided. NOT UNDERSTOOD marks a fact that no current hypothesis derives.

2026-11-15 · aim of this page, and a headedness fact-check. Grant: “What we’re looking for ultimately is a unified syntax and semantics, not just word order.” The structure here is meant to be read that way: one structure whose c-command relations are the scope relations of the logical form, with every argument getting its theta role where it is Merged (VPISH). Word order is one consequence of that structure, and of how each domain is linearized, not the thing being explained. The table in §4 therefore leads with the LF and says, for each structural choice, what it means for interpretation (scope, binding, theta roles and event structure). The same day Grant asked for a sourced check of every headedness claim; it is on the headedness page (fact-check), and its fixes are in §3 below. Grant’s two-tier linking proposal for the verb word is §3.3; its investigation (model, derivations, rivals, predictions) is §3b.

The goal. HYPOTHESIS A unified syntax–semantics for Navajo: a structure in which what c-commands what is what scopes over what, and in which theta roles come from where arguments are Merged. Linear order follows from that structure plus local linearization; on Grant’s Merge-site view the direction of linearization adds no meaning, while the height of a Merge site does.

On this page: 1 What is assumed · 2 The clause · 2b Nouns and postpositions · 3 The verb word · 3b Two-tier linking: investigation · 4 Every phenomenon on the structure (table) · 5 Notes by phenomenon · 6 What is open · 7 Older pages restated

1. What is assumed

HYPOTHESIS Assumed in every derivation on the site (Grant, 2026-11-13 to 2026-11-15; derivation §S1):

  1. VPISH and theta roles. Every argument starts inside the verb’s own projection, where it gets its theta role: internal arguments in VP, the external argument in Spec,vP. VPISH is a device for lining up syntax and meaning, not an account of word order, and it is never the name of a node.
  2. The working default in tensed clauses. The verb moves V→v→T. The subject moves from Spec,vP to Spec,TP, and on to Spec,CP (or C) when needed. Lower copies are silent. Navajo has no non-finite verb forms (Willie 1991:24, review 16), so every clause has a T⁰ for the verb to reach.
  3. What else moves is where the claims are. Each extra Move is stated with its trigger, or its trigger is marked OPEN.
  4. Logical form. Every derivation ends in an LF whose scope order matches c-command at the position where each expression is interpreted.
  5. Specifiers on the left, real X-bar labels, and sentential negation kept apart from constituent negation.
  6. Withdrawn and open. Light d is withdrawn (Grant, 2026-11-14). Light n, a noun-phrase head parallel to light v (Grant’s DP/TP symmetry point), is an OPEN structural question. Telicity is OPEN wherever there is no minimal pair.
  7. Older analyses (flat S, position-class templates, pronominal-argument and adjunction accounts, rule sets A/B) supply data; their analyses appear only as labelled contrast.

2. The clause: one spine

HYPOTHESIS Every clause, main or embedded, question or statement, direct or inverse, is built on this spine (left specifiers; heads written where HS-1 puts them):

[CP  (Spec,CP: fronted wh H14, topic H49)
     [C′ [NegP [Neg′ [TP  DP_subj
                         [T′ [DistP [Dist′ [vP  (outer Spec,vP: inverse O on its way, H21b)
                                                [vP ⟨DP_subj⟩
                                                    [v′ [VP (DP_int) (PP) ⟨V⟩ ] ⟨v⟩ ]]]
                                           ⟨Dist⁰⟩ ]]
                             V+v+(Dist)+T⁰ ]]
                    (doo adjoined inside, H23) ]
               Neg⁰ = da ]]
         C⁰ = Q, =go, (lágo / laanaa: final heads above an optative TP, H52) ]]

OPEN Head direction of the clause. HS-1: T⁰, Neg⁰ and C⁰ are all final; with the default, the verb is last in TP and nothing else moves for verb-final order (the cheaper hypothesis, used in every derivation). HS-2: T⁰ is initial and the remnant VP moves above it in every clause. HS-3: all heads initial with roll-up (Kayne 1994). Which is right is OPEN (headedness). Under any of them the clause’s functional heads that are visible as separate words (da, =go, lágo, yaʼ) are on the right of what they take, except daʼ SOURCED (headedness §2.1).

The basic clause. łį́į́ʼ dzaanééz yiztał (NX-029; Willie 1991:71 (37a)), translated “The horse kicked the mule.”. Merge V ‘kick’ with the mule (Patient); merge v and the horse in Spec,vP (Agent); merge T; V→v→T; the horse moves to Spec,TP. Nothing else moves under HS-1. LF: ∃e[kick(e) ∧ Agent(e,horse) ∧ Patient(e,mule)].

2b. Nouns and postpositions: one nominal structure

HYPOTHESIS The nominal side uses the same X-bar shape, with left specifiers:

[DP (Spec: possessor, linked by agreement)  [D′ [ (nP?) [NP … N ] ]  D⁰ ]]      D⁰ overt only: =yę́ę, =ígíí, -í-, léiʼ, łaʼ
[DP [D′ CP D⁰=yę́ę/=ígíí ]]                                                       relative / fact clause (H27b, H34)
[PP DP [P′ P⁰ ]]                                                                  grammatical P: probe, agrees with Spec,PP (H64)
[PP [P′ DP/place P⁰ ]]                                                            lexical P: no probe (H64)

3. The verb word

SOURCED The facts any account must cover (headedness §2.2–2.3). The stem is last, and the prefixes come before it in a fixed order. Hale (2003:14–15, after Young & Morgan 1987:37–38, Rice’s glosses) lists ten positions, leftmost first: I preverb · II iterative/reversionary · III distributive plural · IV object · V deictic subject · VI qualifier · VII/VIII mode/subject (portmanteau) · IX voice (“classifier”) · X stem. Kibrik (2005, ex. (2), citing Young & Morgan 1987) labels one word, leftmost first, as derivational aspect · lexical aspect · inflectional aspect · mode · person · valency · root, and says the inflectional order outward from the root is “nearly mirror-image” to the common root–aspect–tense–mood–person order (that is, Navajo against the Mirror Principle; this is not the same “mirror” as Grant’s mirror between word and clause, see the fact-check, row 14). Rice (2000:4, quoted in Goad & Travis 2021:2): “in Athapaskan languages morphemes of greater scope appear to the right of morphemes within their scope”, said of positions 1–8; subjects are right of objects (Bird 2001). Preverb and stem form one lexical unit yet sit at opposite edges of the word, “much in the manner of Dutch and German verb-particle arrangements in verb-second finite clauses” (Hale 2001:678). Hale treats the stem as root plus a verbal host carrying aspectual inflection, the suffix -ł in the future and progressive (Hale 2003 n. 10).

3.1 The headedness mismatch as a result of Merge sites

GRANT'S HYPOTHESIS Grant’s hypothesis (Grant, 2026-11-15, building on his working draft Merge Sites and Agree Sites: Exploratory Notes, August 2026, §2 and §§3.1–3.3, and on the site’s original head-direction-mismatch rules, headedness §1). Headedness is not a language-wide parameter. It is “a local linearization decision fixed at particular heights of structure (or inside particular morphological domains)” (notes §2). In Navajo the morphological (sublexical) domain and the phrasal domain are linearized in opposite directions, so the order of heads inside the verb word is the mirror image of their order in the clause. Navajo’s ordered prefix slots are “successive Merge sites and Agree sites inside the complex word” (notes §2). German verb-second is the mirror case at the top of the clause: the verbal projections are head-final, and the finite verb is Internally Merged to C in root clauses (notes §3.2). On this view the order of heads in the word is predicted to mirror their order in the clause (a higher head further left in the word, further right in the clause), and that flip, verb-final clause with a prefixing word, is one of the most interesting properties of the language. Precision (fact-check, 2026-11-15): this predicted order is the order Baker’s Mirror Principle predicts for prefixes. Where the sourced Navajo order goes the other way (Rice’s wider scope nearer the stem; Kibrik’s “nearly mirror-image” order), it is a misfit for this prediction, not a confirmation of it (§3.2). Because the order of prefix heads reads off their height, this is a claim about scope, not only about order.

HYPOTHESIS What it predicts. In the clause (HS-1) a higher head is further right: T⁰, then Neg⁰ da, then C⁰. In the word, with linearization reversed, a head Merged higher is further left, and the stem (root plus v) is at the right edge. So:

  1. P1, scope order. Reading the prefixes from the stem leftward should climb the hierarchy: v (classifier) next to the stem, then the heads above it in order (aspect, mode/T, agreement, …), with the highest at the far left.
  2. P2, the same head in both domains. A head that has a clause-level and a word-level Merge site should be final in the clause and initial in the word.
  3. P3, Agree sites. A probe’s slot fixes what it agrees with: the closest argument it c-commands from that height (notes §2, “Agree also has designated sites”).

HYPOTHESIS What counts as the head, and Grant’s two wordings. Grant’s notes §3.3 say the word is built around a rightmost stem, with prefixes “ordered leftward by successive head movement or by templatic Merge”, so that the complex is “right-headed at the word level—the morphological analogue of a head-final projection”. Taken at face value, that matches the clause (the lexical head last in both) rather than flipping it. Grant’s statement of 2026-11-15 (morphology head-initial, phrasal syntax head-final) counts something else as the head: each prefix is the head of its own word-internal layer and comes before the material it combines with. The two are not in conflict about the string; they choose different heads. If the stem counts as the head of the word, Navajo is right-headed in the word and in the clause, and there is no mismatch to explain. If each functional prefix counts as the head of its layer, the word is head-initial and the clause head-final, and that is the mismatch. Only the second reading makes a claim about the order of prefixes among themselves (P1); the first says only that the stem is last. So the data that decide between them are the prefix-to-prefix orders, tested next.

3.2 Test against the sourced prefix data

Sourced factP1 (prefix heads, linearized leftward by height)Verdict
Stem last; voice/“classifier” (IX) right next to it (Hale 2003; Kibrik’s valency next to the root)v is the head closest to V, so it is next to the stemHYPOTHESIS fits
Mode/subject portmanteau (VII/VIII) left of the classifier (Hale 2003; Kibrik’s mode · person · valency)T⁰ and the agreement it carries are above v, so further leftHYPOTHESIS fits
Hale’s conjunct hierarchy Qual > M/S > Voice > V with head-initial selection (headedness §4)exactly P1 for the conjunct sector (IV–IX)HYPOTHESIS fits (a published analysis of the same shape)
Kibrik: inflectional aspect left of mode; derivational and lexical aspect outermostaspect is normally below T, so it should be right of mode; derivational material is normally lowestOPEN misfit unless these aspect prefixes are Merged above T in Navajo (no test yet)
Kibrik: the inflectional order outward from the root is “nearly mirror-image” to the common suffix orderP1 with the usual hierarchy predicts the same outward order as suffixing languages (lowest nearest the root)OPEN misfit on the usual hierarchy; fits only if person is low and aspect high in Navajo
Object prefix (IV) left of subject (VIII); Rice: greater scope to the right (positions I–VIII; subject over object)the subject is higher, so its agreement should be leftOPEN misfit, unless object agreement is spelled on a head above subject agreement (H58a, review 16, which needs a probe that skips the closer subject: OPEN)
Distributive da- (III) left of mode; Dist⁰ is just above vP and below T (H21a)Dist⁰ below T should be right of modeOPEN misfit with H21a’s height; P1 would need Dist⁰ above T, which H21a’s agreement facts do not yet rule out
Iterative and distributive in one order in Navajo, the reverse in some relatives; Rice: no scope relation between them (Hale 2001:689)no scope relation, so P1 predicts nothing either wayHYPOTHESIS neutral
Reflexive da after ʼá in ʼádadiidleesh (NX-493) but before ʼahi in daʼahiidleesh (NX-497) (review 16; the source does not segment)one Dist⁰ should have one placeOPEN an observation for the ordering rule, not yet a test
Preverb and stem: one lexical unit, opposite edges (Hale 2001:678)not predicted by P1 alone: the preverb would be the lowest material and should sit by the stemOPEN misfit for P1 alone; see §3.3–3.4

HYPOTHESIS Result. On the sourced data the Merge-site account fits the conjunct core of the word: stem last, classifier next to it, mode/subject outside that, and Hale’s head-initial conjunct hierarchy is that account’s shape. It does not yet fit four sourced orders: aspect left of mode, object agreement left of subject agreement, the distributive left of mode, and the split preverb. Each misfit can be removed only by a claim about height (where those heads are Merged in Navajo) that has not been tested. Rice’s and Goad & Travis’s “anti-mirror” (the last-attached affix nearest the root) is the rival: it fits the object/subject and person/aspect orders, but it puts the classifier’s place next to the stem down to prosody rather than syntax (Goad & Travis 2021:5, 12) and conflicts with Hale’s conjunct hierarchy; it is kept as labelled contrast. Precision (fact-check, 2026-11-15): P1 is the Mirror Principle’s order for prefixes, and Rice’s generalization for positions I–VIII is its reverse; the four misfits are instances of that one generalization, not four separate exceptions. The fit of the conjunct core rests on Hale’s hierarchy (the Qualifier selects Mode/Subject, Hale 2001:680), which Rice’s scope analysis and Goad & Travis (2021:4, position 6 as v) put the other way. Since prefix order is read as height here, the decisive data are scope data. Fit and misfit are recorded here; nothing is settled. OPEN

3.3 Two-tier linking inside the word

GRANT'S HYPOTHESIS OPEN Grant’s hypothesis, untested (Grant, 2026-11-15). Inside the verb word the structure is not head-to-head plus spec-to-spec V-to-T but some other two-tier structure. Verbs (more or less) select pronouns: the pronominal prefixes are the verb’s arguments. A linking element ties each prefix pronoun to its referent. The two tiers travel by different means: head-to-head movement carries the prefix tier, and spec-to-spec movement carries the tier of overt nouns. This is now the main word-internal candidate; sublexical V-to-T (§3.4) is kept as an earlier, untested alternative.

HYPOTHESIS How it sits on the spine. Each argument position in VP and vP holds a pronoun that is pronounced as a prefix, and it gets its theta role there (VPISH). An overt noun, when there is one, sits in a specifier and is linked to one prefix pronoun. The link is what gives the noun its role. Interpretation: the noun binds the pronoun’s variable. At LF a noun’s scope is fixed by its own specifier (the noun tier), and its role is read from the prefix it is linked to (the prefix tier). So the model splits the two jobs that the clause spine otherwise gives to one position.

HYPOTHESIS What it predicts, tested lightly against facts already on the site (no new Navajo data):

  1. L1. Role by link, not by position. A noun has the role of the prefix it is linked to. Fit: the site already states the one-noun default this way (“role by linking”, H4″; §4 table).
  2. L2. Linking is restricted by the prefix tier. The linking element should limit which prefixes can have a noun. Fit: H31 is exactly such a restriction: “a noun, a relative DP and the enclitic are third person, and can be linked only to a third-person participant” (results §8). First- and second-person prefixes have no noun to link to. Strain: review 15’s respect, report and story ji- pairs a noun with a fourth-person prefix. The model keeps H31 only if that ji- is a respect or report mark on a third-person argument (H57a, favoured by the Japanese comparison), not a different person. OPEN
  3. L3. The two tiers can come apart in order. Because the tiers move by different means, the order of the nouns need not match the order of the prefixes, and the link has to record which noun goes with which prefix. Fit, as a reinterpretation: the inverse. The noun tier reorders (the patient moves above the agent, H3, H21b, and takes scope over it), while the prefix tier marks the same relation as a choice of prefix, yi- against bi- (H9′). On this model yi-/bi- would spell out the link itself: whether a prefix pronoun’s linked noun is the higher one. That restates H9′; it does not yet test it. OPEN
  4. L4. Prefix order need not show argument height. If argument prefixes ride the head-to-head tier, their order is fixed by where they attach in the word, not by the height of the arguments they stand for. Then the object prefix (IV) to the left of the subject prefix (VIII), a misfit for P1 (§3.2), is not evidence against the subject being higher. Partial: this takes the misfit out of P1’s domain, but the model does not yet say why the object pronoun attaches further out. Until it does, the misfit is moved, not derived. (Compare H58a, review 16: object agreement spelled on a head above subject agreement.) OPEN
  5. L5. Nouns stay outside the prefix tier. Only pronouns enter the word; nouns are linked from specifiers. Test: a noun inside the prefix string would count against the model. Possessor raising (“the noun goes into the verb, the person becomes the object”, possessor raising) is the case to check first. It is not yet restated on the spine (§2b). OPEN

HYPOTHESIS Labelled contrast. The model shares one premise with Jelinek’s Pronominal Argument Hypothesis, as Hale discusses it (Hale 2003 ms., “On the significance of Eloise Jelinek’s Pronominal Argument Hypothesis”): the prefixes are the arguments. It differs on the nouns. Here they are in specifiers, move spec-to-spec and enter scope relations (the inverse patient over the agent; quantifier scope, H22), rather than being adjuncts outside the argument structure. What sublexical V-to-T explained (the stem’s shape fixed by mode, and the stranded preverb) is not lost: the head-to-head tier still carries the stem.

GRANT'S HYPOTHESIS Investigated, 2026-11-15: §3b states the model precisely and derives the sourced facts with it. Result for L4: the object-left-of-subject order is derived, not just moved, if the subject pronoun is on the head tier and the object pronoun on the phrasal tier (LK-p, which is Hale’s nuclear/nonnuclear split); it is not derived if both pronouns head-move.

3.4 Sublexical V-to-T (earlier alternative)

GRANT'S HYPOTHESIS Grant’s hypothesis, untested; since Grant’s two-tier proposal (2026-11-15, §3.3) an earlier alternative, not the main candidate (Grant, 2026-11-15, explicitly uncertain: “there may be some kind of sublexical V to T movement”). Inside the word, the stem raises through word-internal heads to a word-level T-type position, the way V raises to T in the clause. Candidates for the word-level T-type material: the mode/aspect material, including Hale’s aspectual verbal host -ł on the stem (Hale 2003 n. 10).

HYPOTHESIS What it predicts, beyond the plain Merge-site account.

  1. Stem shape is fixed by the heads the stem reaches. The stem should change with mode and aspect and not with, say, the object’s person. SOURCED The stem changes with mode while the participants stay the same in yi-sh-kʼá (NX-479; Willie 1991:30 (4a)), translated “I am grinding it.” and wóshkʼááł (NX-483; Willie 1991:30 (4e)), translated “I wish to grind it.” (review 16, (4a–e)); the iterative changes the stem too ((4a) against (4b)). Fit. But a rival in which T⁰ is spelled on the verb by lowering is just as local (review 16, H63), so this does not decide it. OPEN
  2. Stranding. Material Merged with the stem low should be left behind when the stem raises, as German leaves its particle in verb-second clauses. SOURCED Preverb and stem form one lexical unit at opposite edges of the word (Hale 2001:678). Fit, and this is the one fact in §3.2 that the plain Merge-site account misses. Whether the preverb is really Merged with the stem (Rice 2000, Hale 2001: one lexical unit) or at its own high site is OPEN.
  3. Stem at the right edge. If the word-level T-type position is at the right edge (Hale’s suffixal host), the stem ends up last however the prefixes are linearized. This makes stem-last independent of P1, so the prefix-to-prefix misfits of §3.2 remain exactly as they are.
  4. One string, two heads. A single exponent can belong to two heads the stem reaches: naashnish (NX-478; Willie 1991:29 (3)), translated “I am working.”, also read “he worked” in Willie’s prose, needs no movement difference, only first-person subject agreement against the perfective value with the same exponent (review 16). Neutral.

OPEN Comparison. The plain Merge-site account derives stem-last and the conjunct order directly, but not the split preverb. Sublexical V-to-T adds the split preverb (by stranding) and makes stem/mode interactions local, at the cost of a word-internal Move whose landing site (which head, on which side) is not yet stated. Neither removes the aspect, object/subject and distributive misfits. A test would need a stem alternation conditioned by a head that the stem should not reach (for sublexical V-to-T, a disjunct head), or one preverb–stem pair whose preverb provably merges high.

SOURCED What this replaces. Until 2026-11-15 the site said that head movement “predicts the wrong order” inside the word and that the word “needs its own rule”. That framing is withdrawn on Grant’s correction: the mirror order is the prediction of the headedness mismatch, and what is open is whether the details of the prefix order fit it (§3.2).

3b. The two-tier linking model (GRANT'S HYPOTHESIS): investigation

2026-11-15 · investigation of Grant’s two-tier linking hypothesis. Grant (2026-11-15): inside the verb there is “some other two-tiered structure”, not head-to-head plus spec-to-spec V-to-T; verbs (more or less) select pronouns, and some element links each pronoun to its referent, working through head-to-head and spec-to-spec movement. Grant asked for an investigation “from that angle”. This section makes the hypothesis precise (as model LK), derives the site’s sourced facts with it, compares it with the alternatives in the literature and lists what would decide it. It adds no Navajo data: every form is copied from the example index under its NX number. §3.3 (light tests L1–L5) stands; this section builds on it. Follow-up 2026-11-15/16: §3b.7 tests P-LK1–P-LK7, adds Grant’s theta-linking strand (which replaces LK-4 and derives H9′ and the trigger of H3 from one LF condition, given an imported ranking), revises H21a to H21a′ and treats deictic ji- as a referent.

Verdict in plain words. GRANT'S HYPOTHESIS HYPOTHESIS The model works for the facts about which noun goes with which prefix: H31 and its gap, the fourth person, possessors, postposition agreement, relatives and the personless passive agent all fall out of one condition (a noun must agree in person with the pronoun it is linked to). It works for the word only in one version: the subject pronoun is on the head tier (it travels with the verb into T⁰ and comes out as the mode/subject portmanteau), while the object pronoun is on the phrasal tier, with the nouns and the adjunct prefixes. That version derives the object-left-of-subject order, which no head-movement account on this page derived, and it turns out to be Hale’s own split into a “nuclear” and a “nonnuclear” plane (Hale 2003 ms.:33, 38 (32)), restated with movement instead of a phonological filling procedure. The version in which both pronouns ride the head tier (as Baker & Kramer’s clitics do) brings the misfit back. It fails, or needs extra, on three facts: the split preverb, the respect ji- unless it is third person (H57a), and *NX-345 against NX-346. Nothing in it is tested by data it was built to fit; seven predictions below say what would test it. OPEN

3b.1 The model, stated so it can be checked

GRANT'S HYPOTHESIS as formalized here (our formalization, HYPOTHESIS; the labels LK-0 to LK-4 are this page’s):

  1. LK-0 Argument shell. Each argument is Merged in its theta position (VPISH) as [DP DPref [D′ D⁰i,φ]]. D⁰ is the pronoun pronounced as a prefix (Hale 2003 ms.:10–11: the pronouns are D and “(trivially) maximal projections”). The left specifier holds the referent: an overt noun, or a silent pronoun with its own person. The linking element is the specifier–head relation inside the shell; it is a relation, not a new node or category.
  2. LK-1 Link condition. The referent agrees with D⁰ in person and number and shares its index. Interpretation: ⟦D⁰i⟧ = xi, defined only if φ(xi); the referent identifies xi. A quantified or indefinite referent raises and binds xi by λ-abstraction, as any quantifier does.
  3. LK-2 Tier 1 (heads). A D⁰ can head-move into the verb word. This is head movement out of a specifier, so it needs prior Agree with the host head (the precondition argued for clitic doubling by Preminger 2019, abstract).
  4. LK-3 Tier 2 (phrases). The referent moves spec-to-spec (outer Spec,vP, Spec,TP, topic and focus positions). Where it lands fixes its scope; theta roles stay where the shell was Merged.
  5. LK-4 Cross-tier spell-out. The exponent of D⁰ may depend on where its referent landed: bi- when the referent c-commands the other third-person referent, yi- otherwise (this restates H3 with H9′; it is not a derivation of them).

HYPOTHESIS Which pronoun is on which tier: LK-h against LK-p. Because a complement-less D⁰ is a head and a phrase at once, the model has two versions. LK-h: every argument D⁰ head-moves (object D⁰ to v⁰, subject D⁰ to T⁰). This is Baker & Kramer’s analysis of object markers, a D adjoined to v (Baker & Kramer 2018 ms.:5 (7)). LK-p: only the subject D⁰ head-moves (to T⁰, spelled with mode as the portmanteau, Hale 2003 ms.:10, 68 n. 11); the object D⁰ behaves as a phrase and is pronounced on the phrasal tier, to the left of the head complex.

Sourced factLK-h (both pronouns on the head tier)LK-p (subject on the head tier, object on the phrasal tier)
Object prefix (IV) left of mode/subject (VII/VIII) (Hale 2003 ms.:8 (7), 14–15 (10))Object D⁰ adjoins to v⁰, which is right of M/S in the head complex: object predicted right of subject. The §3.2 misfit returns. contradictedPhrasal tier precedes the head complex: object left of M/S, with no probe skipping the subject. fits
One prefix shape, two places: outer for the Patient, inner for the Agent (NX-499, NX-020; H58, review 16)Both inner (head tier); the difference must be stipulatedPatient pronoun on the phrasal tier (outer), Agent pronoun on the head tier (inner). Derives H58. fits
Hale’s planes: Qual, M/S, v, V “nuclear”; object, preverb, iterative/reversionary and plural da “nonnuclear”, prefixed last, “working leftward (upwards in the structure)” (Hale 2003 ms.:33, 38 (32), 40)Object on the wrong planeSame split. LK-p is Hale’s two planes with movement for his filling procedure. fits
Distributive da- (III) left of mode; Dist⁰ just above vP (H21a)Dist⁰ is a head below T: right of mode. Misfit staysFits only if da- is phrasal (as Hale’s plane puts it) and not the Dist⁰ head of H21a: H21a would have to become a phrase agreeing with the counted argument. needs extra
Preverb and stem one lexical unit at opposite edges; low preverbs (Hale 2003 ms.:19–21 (14), (16), (21))Not derivedA phrase Merged low should be pronounced inside the phrasal tier, not leftmost. Needs a spec-to-spec move of the preverb to the word’s left edge, trigger unknown. NOT UNDERSTOOD
Deictic subject (V) right of the object (IV), left of the qualifier (Hale 2003 ms.:14–15 (10); left out of (32), p. 38)Fits if ji- is a headFits if ji- is the highest head of the head complex, not fused with M/S. needs extra (P-LK7)

HYPOTHESIS The word on LK-p. Phrasal tier, specifiers on the left, ordered as Hale’s plane orders it: preverb (I) · iterative/reversionary (II) · distributive da- (III) · object pronoun (IV) · then the head complex, head-initial by selection (Hale 2003 ms.:32): deictic subject (V) · qualifier (VI) · mode/subject (VII/VIII) · voice v⁰ (IX) · stem (X). So Grant’s headedness flip (§3.1) becomes a difference between tiers: the head complex is head-initial, the phrasal tier is specifier-left. Whether every prefix lands on the tier this assigns it is P-LK1.

Analysis A: Direct: łį́į́ʼ dzaanééz yiztał (NX-029) “The horse kicked the mule.” Two shells; subject D⁰ to T⁰ (head tier), the horse to Spec,TP (phrase tier); working default Hypothesis

Analysis A: Direct: łį́į́ʼ dzaanééz yiztał (NX-029) “The horse kicked the mule.” Two shells; subject D⁰ to T⁰ (head tier), the horse to Spec,TP (phrase tier); working defaultEach argument is Merged in its theta position as a shell [DP referent [D′ D⁰]] (GRANT'S HYPOTHESIS as formalized here, LK-0): D⁰ is the pronoun pronounced as a prefix, with person/number and an index; its left specifier holds the referent (an overt noun, or a silent pronoun with its own person). The link L is the specifier–head relation inside the shell: the referent must agree with D⁰ in person (LK-1, which gives H31, since a noun is third person) and binds D⁰’s variable at LF. Tier 1 (heads): the subject’s D⁰ head-moves to T⁰ and is spelled with mode as the mode/subject portmanteau (Hale 2003:10, 68 n. 11). Tier 2 (phrases): the referent moves spec-to-spec. The object’s D⁰ has no complement, so it is a minimal and a maximal projection at once (Hale 2003:10–11: pronouns are “(trivially) maximal projections”); on LK-p it is pronounced on the phrasal tier, to the left of the head complex, and is not drawn with a head-movement arrow. Working default: V⁰ head-moves through v⁰ to a final T⁰ (V→v→T), and the subject’s referent raises from the left specifier of its shell in Spec,vP to left Spec,TP; lower copies are silent. With T⁰ final (HS-1, HYPOTHESIS) the verb word is pronounced last, so nothing else has to move for verb-final order; under HS-2 the material left in vP would have to move above T (OPEN). L is a relation, not a node; the link head is labelled D⁰ (light d stays withdrawn: D⁰ here is the overt prefix pronoun, not a silent determiner on the noun). Panel A. Merge V ‘kick’ with the object shell (Patient; the mule in its specifier); merge v and the subject shell in Spec,vP (Agent; the horse in its specifier). The subject D⁰ moves to T⁰; this is head movement out of a specifier, licensed by T’s agreement with it (the precondition argued for clitic doubling in Preminger 2019, abstract). The horse raises to Spec,TP. Nothing else moves. Order: horse, mule, verb word; the object prefix precedes the portmanteau because it is on the phrasal tier (LK-p). LF: ∃e[kick(e) ∧ Agent(e,x) ∧ Patient(e,y)], with x = the horse and y = the mule (each noun identifies the variable of the D⁰ it is linked to) Full structure in the list below the figure.TPDPłį́į́ʼT′vPT⁰DPv′DPłį́į́ʼD′VPv⁰D⁰x: subject pronoun, 3rd personV′DPV⁰yiztałDPdzaanéézD′D⁰y: object pronoun yi-, 3rd personverb word (head tier; yi- on the phrasal tier, left)head-movehead-movehead-moveAgree φ: mode/subject portmanteauMove

Analysis B: Inverse: dzaanééz łį́į́ʼ biztał (NX-030) “The mule was kicked by the horse.” Same shells and roles; the mule’s referent moves spec-to-spec above the raised agent; its D⁰ is spelled bi-; working default Hypothesis

Analysis B: Inverse: dzaanééz łį́į́ʼ biztał (NX-030) “The mule was kicked by the horse.” Same shells and roles; the mule’s referent moves spec-to-spec above the raised agent; its D⁰ is spelled bi-; working defaultEach argument is Merged in its theta position as a shell [DP referent [D′ D⁰]] (GRANT'S HYPOTHESIS as formalized here, LK-0): D⁰ is the pronoun pronounced as a prefix, with person/number and an index; its left specifier holds the referent (an overt noun, or a silent pronoun with its own person). The link L is the specifier–head relation inside the shell: the referent must agree with D⁰ in person (LK-1, which gives H31, since a noun is third person) and binds D⁰’s variable at LF. Tier 1 (heads): the subject’s D⁰ head-moves to T⁰ and is spelled with mode as the mode/subject portmanteau (Hale 2003:10, 68 n. 11). Tier 2 (phrases): the referent moves spec-to-spec. The object’s D⁰ has no complement, so it is a minimal and a maximal projection at once (Hale 2003:10–11: pronouns are “(trivially) maximal projections”); on LK-p it is pronounced on the phrasal tier, to the left of the head complex, and is not drawn with a head-movement arrow. Working default: V⁰ head-moves through v⁰ to a final T⁰ (V→v→T), and the subject’s referent raises from the left specifier of its shell in Spec,vP to left Spec,TP; lower copies are silent. With T⁰ final (HS-1, HYPOTHESIS) the verb word is pronounced last, so nothing else has to move for verb-final order; under HS-2 the material left in vP would have to move above T (OPEN). L is a relation, not a node; the link head is labelled D⁰ (light d stays withdrawn: D⁰ here is the overt prefix pronoun, not a silent determiner on the noun). Panel B. The same Merge steps and theta roles as panel A. Beyond the default, the mule (the patient’s referent, Tier 2) moves spec-to-spec to a position above the raised agent (H3, H21b: via outer Spec,vP, an intermediate step not drawn; landing site and trigger OPEN). The object’s D⁰ stays the patient’s pronoun; only its spell-out changes: bi- when the referent it is linked to c-commands the other third-person referent (LK-4, a restatement of H3 with H9′, not a derivation of it). Order: mule, horse, verb word. Scope = c-command: the mule scopes over the horse; the event description is unchanged. LF: [the mule] λy [[the horse] λx ∃e[kick(e) ∧ Agent(e,x) ∧ Patient(e,y)]] Full structure in the list below the figure.TPDPdzaanéézTPDPłį́į́ʼT′vPT⁰DPv′DPłį́į́ʼD′VPv⁰D⁰x: subject pronoun, 3rd personV′DPV⁰biztałDPdzaanéézD′D⁰y: object pronoun bi-, 3rd personverb word (head tier; bi- on the phrasal tier, left)head-movehead-movehead-moveAgree LK-4: referent above the agent’s → bi-MoveMove

Analysis C: Fourth-person subject: John jiztał (NX-468) “She kicked him, John.” John can be linked only to the third-person shell; working default Hypothesis

Analysis C: Fourth-person subject: John jiztał (NX-468) “She kicked him, John.” John can be linked only to the third-person shell; working defaultEach argument is Merged in its theta position as a shell [DP referent [D′ D⁰]] (GRANT'S HYPOTHESIS as formalized here, LK-0): D⁰ is the pronoun pronounced as a prefix, with person/number and an index; its left specifier holds the referent (an overt noun, or a silent pronoun with its own person). The link L is the specifier–head relation inside the shell: the referent must agree with D⁰ in person (LK-1, which gives H31, since a noun is third person) and binds D⁰’s variable at LF. Tier 1 (heads): the subject’s D⁰ head-moves to T⁰ and is spelled with mode as the mode/subject portmanteau (Hale 2003:10, 68 n. 11). Tier 2 (phrases): the referent moves spec-to-spec. The object’s D⁰ has no complement, so it is a minimal and a maximal projection at once (Hale 2003:10–11: pronouns are “(trivially) maximal projections”); on LK-p it is pronounced on the phrasal tier, to the left of the head complex, and is not drawn with a head-movement arrow. Working default: V⁰ head-moves through v⁰ to a final T⁰ (V→v→T), and the subject’s referent raises from the left specifier of its shell in Spec,vP to left Spec,TP; lower copies are silent. With T⁰ final (HS-1, HYPOTHESIS) the verb word is pronounced last, so nothing else has to move for verb-final order; under HS-2 the material left in vP would have to move above T (OPEN). L is a relation, not a node; the link head is labelled D⁰ (light d stays withdrawn: D⁰ here is the overt prefix pronoun, not a silent determiner on the noun). Panel C. The subject shell’s D⁰ is fourth person (spelled ji-); its specifier can hold only a silent pronoun of the same person (z₄), never a noun, because a noun is third person (LK-1 = H31). So John can only be the referent in the object shell’s specifier: the printed reading “She kicked him, John” follows without a one-noun default. The derivation with John in the subject shell crashes on person agreement inside the shell. Nothing moves beyond the default. Sourced fact: Willie 1991:112 (7b) glosses jiztał “3.4.kicked”. LF: ∃e[kick(e) ∧ Agent(e,z₄) ∧ Patient(e,y)], with y = John Full structure in the list below the figure.TPDPsilent z₄ (fourth person)T′vPT⁰DPv′DPsilent z₄ (fourth person)D′VPv⁰D⁰z₄: subject pronoun ji-, 4th personV′DPV⁰jiztałDPJohnD′D⁰y: object pronoun, 3rd personverb wordhead-movehead-movehead-moveMove
(1) The two-tier linking model (GRANT'S HYPOTHESIS, formalized as LK): argument shells on the clause spine Hypothesis NX-029, NX-030: Willie 1991:71 (37a, b): łį́į́ʼ dzaanééz yiztał “The horse kicked the mule.”, dzaanééz łį́į́ʼ biztał “The mule was kicked by the horse.”; NX-468: Willie 1991:112 (7b): John jiztał, glossed “3.4.kicked”, “She kicked him, John.” Leaves are printed words. The shells, D⁰, T⁰, v⁰, the silent pronouns, the landing sites and the role labels are our assignments.
Bracket notation (copy)

A: [TP [DP łį́į́ʼ] [T′ [vP [DP ⟨DP⟩ [D′ [D⁰ x: subject pronoun, 3rd person]]] [v′ [VP [V′ [DP [DP dzaanééz] [D′ [D⁰ y: object pronoun yi-, 3rd person]]] [V⁰ yiztał]]] v⁰]] T⁰]] B: [TP [DP dzaanééz] [TP [DP łį́į́ʼ] [T′ [vP [DP ⟨DP⟩ [D′ [D⁰ x: subject pronoun, 3rd person]]] [v′ [VP [V′ [DP ⟨DP⟩ [D′ [D⁰ y: object pronoun bi-, 3rd person]]] [V⁰ biztał]]] v⁰]] T⁰]]] C: [TP [DP silent z₄ (fourth person)] [T′ [vP [DP ⟨DP⟩ [D′ [D⁰ z₄: subject pronoun ji-, 4th person]]] [v′ [VP [V′ [DP [DP John] [D′ [D⁰ y: object pronoun, 3rd person]]] [V⁰ jiztał]]] v⁰]] T⁰]]

Structure as a list

Analysis A

  • TP
    • DP (id S-high; landing site of Move from Rs): łį́į́ʼ
    • T′
      • vP
        • DP (id shellS)
          • DP (id Rs; struck copy; moved to S-high): łį́į́ʼ
          • D′
            • D⁰ (head; id Ds; head-moves to T; goal of Agree φ: mode/subject portmanteau): x: subject pronoun, 3rd person
        • v′
          • VP
            • V′
              • DP (id shellO)
                • DP (id Ro): dzaanééz
                • D′
                  • D⁰ (head; id Do): y: object pronoun yi-, 3rd person
              • V⁰ (head; id V; head-moves to v): yiztał
          • v⁰ (head; id v; head-moves to T)
      • T⁰ (head; id T; probe for Agree φ: mode/subject portmanteau with Ds; hosts the “verb word (head tier; yi- on the phrasal tier, left)” (V + v + T + Ds))

Analysis B

  • TP
    • DP (id O-high; landing site of Move from Ro; goal of Agree LK-4: referent above the agent’s → bi-): dzaanééz
    • TP
      • DP (id S-high; landing site of Move from Rs): łį́į́ʼ
      • T′
        • vP
          • DP (id shellS)
            • DP (id Rs; struck copy; moved to S-high): łį́į́ʼ
            • D′
              • D⁰ (head; id Ds; head-moves to T): x: subject pronoun, 3rd person
          • v′
            • VP
              • V′
                • DP (id shellO)
                  • DP (id Ro; struck copy; moved to O-high): dzaanééz
                  • D′
                    • D⁰ (head; id Do; probe for Agree LK-4: referent above the agent’s → bi- with O-high): y: object pronoun bi-, 3rd person
                • V⁰ (head; id V; head-moves to v): biztał
            • v⁰ (head; id v; head-moves to T)
        • T⁰ (head; id T; hosts the “verb word (head tier; bi- on the phrasal tier, left)” (V + v + T + Ds))

Analysis C

  • TP
    • DP (id S-high; landing site of Move from Rs): silent z₄ (fourth person)
    • T′
      • vP
        • DP (id shellS)
          • DP (id Rs; struck copy; moved to S-high): silent z₄ (fourth person)
          • D′
            • D⁰ (head; id Ds; head-moves to T): z₄: subject pronoun ji-, 4th person
        • v′
          • VP
            • V′
              • DP (id shellO)
                • DP (id Ro): John
                • D′
                  • D⁰ (head; id Do): y: object pronoun, 3rd person
              • V⁰ (head; id V; head-moves to v): jiztał
          • v⁰ (head; id v; head-moves to T)
      • T⁰ (head; id T; hosts the “verb word” (V + v + T + Ds))

3b.2 The sourced facts, derived

Each row uses only forms already on the site. “Fits” = derived by LK-0 to LK-4 with LK-p; “needs extra” = derived only with an added assumption, named; “contradicted” = the model predicts the opposite. LFs: scope = c-command.

Fact (sourced)Derivation on LKLFVerdict
H31: a noun, a relative DP and the enclitic are third person and can be linked only to a third-person participant. *ʼatʼééd joodloh=yę́ę yóóʼajoolwod (*NX-359; Willie 1991:174 (53)); strong pronouns link by their own person: Ní, ʼayóí ʼáshííníʼníʼ=nę́ę shaa yisínínah (NX-349; Willie 1991:171 (43)), translated “You, who loved me, forgot about me.”LK-1: a noun in Spec,DP must agree in person with D⁰; nouns are third person, so they fit only third-person shells. In *NX-359 the noun ʼatʼééd sits in a fourth-person shell: person clash. Ní is second person and fits a second-person shell.NX-349: [you]x₂ λx₂ ∃e[forget(e) ∧ Agent(e,x₂) ∧ Theme(e,x₁)] with x₂ = the one who loved x₁fits: H31 becomes a theorem of LK-1, not a separate rule
Respect, report and story ji- with a noun: shimá ʼéí Baaʼ joolyé (NX-470; Willie 1991:113 (8)), translated “My mother is called Baaʼ.”; NX-473, NX-476LK-1 lets a noun into the ji- shell only if that ji- is third person with a respect/report feature (H57a). If it is a fourth person (H57b), NX-470 is predicted bad.∃s[called(s) ∧ Theme(s,x₃) ∧ Name(s,Baaʼ)], x₃ = my mother, [+respect] on x₃’s D⁰needs extra: LK forces H57a over H57b (P-LK4)
Fourth person: Jíícha (NX-467; Willie 1991:112 (7a)), “That certain person cried.”; John jiztał (NX-468; 112 (7b)), “She kicked him, John.”; John haztał (NX-469; 112 (7c)), “John kicked that certain person.”The fourth-person shell takes only a silent same-person referent, so John can be linked only to the third-person shell, whichever role that is (tree, panel C). No one-noun default is needed.NX-468: ∃e[kick(e) ∧ Agent(e,z₄) ∧ Patient(e,y)], y = John; NX-469: ∃e[kick(e) ∧ Agent(e,x) ∧ Patient(e,z₄)], x = Johnfits
Inverse: łį́į́ʼ dzaanééz yiztał (NX-029; Willie 1991:71 (37a)), “The horse kicked the mule.” against dzaanééz łį́į́ʼ biztał (NX-030; 71 (37b)), “The mule was kicked by the horse.”Tier 2: the patient’s referent moves above the raised agent (H3, H21b). Tier 1: the object D⁰ stays and is spelled bi- by LK-4 (tree, panels A, B). Aissen (2000:6 (8)) treats BI and YI as pronouns that incorporate (proximate and obviative object), which is LK’s D⁰; Willie & Jelinek (2000:270) note only one bi- per clause, which LK-4 gives (only one referent can be the higher of two).[the mule] λy [[the horse] λx ∃e[kick(e) ∧ Agent(e,x) ∧ Patient(e,y)]]needs extra: LK-4 is a spell-out rule added for the inverse; the trigger of the move and the definiteness restriction (H6′) are not derived; landing site OPEN
The object-left-of-subject misfit (§3.2)LK-p: object on the phrasal tier, subject on the head tier (3b.1).— (order only; no scope claim, since tiers, not height, order the prefixes)fits on LK-p; contradicted on LK-h
Possessor prefixes: háísh baghan=di Mary yiztsʼǫs? (NX-380; Willie 1991:215 (26b)), John háísh yighan=di yiztsʼǫs? (NX-382; 216 (28b)), John háísh haghan=di yiztsʼǫs (NX-383; 217 (29)); dibé baghan bitsʼą́ą́jįʼ yishááł (NX-508; Willie 1991:44 (36))The possessor is a shell in the left specifier of the possessed noun; its D⁰ (b-, yi-, ha-) is the prefix and an overt possessor (dibé) is its referent. A possessor prefix looks like an object prefix because both are phrasal-tier pronouns.NX-508: ∃e[walk(e) ∧ Agent(e,speaker) ∧ Source(e,ιz.home(z) ∧ Poss(z,s))], s = the sheepfits for the link; which of b-/yi-/ha- appears (NX-380 against NX-382) is LK-4 again and OPEN
Postposition agreement (H64): ʼawééʼ bichʼįʼ yishááł (NX-505; Willie 1991:43 (33)), “I am walking to the baby.”, against *ʼawééʼ-góó yishááł (*NX-506; 43 (34)); Kinłání=góó déyá (NX-503; 42 (31)), “I will go to Flagstaff.”A grammatical P takes a shell (D⁰ bi-, referent ʼawééʼ). A lexical P takes a place DP with no shell, so no prefix. *NX-506: a lexical P with a noun that needs a shell; ruled out by P’s selection, not by LK itself.NX-505: ∃e[walk(e) ∧ Agent(e,speaker) ∧ Goal(e,y)], y = the babyfits; *NX-506 needs the selection statement (H64), which LK does not derive
Relatives =yę́ę: yiztał=yę́ę (NX-312; Willie 1991:152 (1c)), “the one he kicked”; biztał=yę́ę (NX-313; 152 (1d)), “the one who kicked him”; ʼatʼééd yidloh=ę́ę biztał (NX-326; 157 (11))The enclitic is a D⁰ over CP whose index must match one participant’s D⁰ inside the clause; LK-1 applies to that link too (third person only, hence H31 for relatives). The head noun is its referent.NX-312: ιx ∃e[kick(e) ∧ Agent(e,y) ∧ Patient(e,x)]needs extra: the enclitic’s link reaches into CP, which is Agree plus λ-abstraction, not a spec–head relation. *NX-345 against NX-346 NOT UNDERSTOOD
Passive ʼadi- and reflexive: niʼdilzhééh (NX-020; Willie 1991:37 (2.21)), “You are being shaved.”; shiʼdidleesh (NX-021), “I am being painted.”; ʼádílzhééh (NX-499; 37 (20)), “You are shaving yourself.”The agent’s D⁰ is personless (H60), so by LK-1 its shell can hold no noun and no referent: this is why the passive agent can never be named. Slot order: the patient pronoun (ni-, shi-; reflexive ʼádi-) is phrasal-tier (outer), the agent pronoun head-tier (inner) — H58 derived.NX-020: ∃u∃e[shave(e) ∧ Agent(e,u) ∧ Patient(e,y₂)]fits; one change: with no referent there is no phrase to raise to Spec,TP, so “u raises to Spec,TP” becomes “u’s D⁰ reaches T⁰ by head movement”; whether Spec,TP is then empty is OPEN
One noun, two participants (H4″): ʼashkii yiyííbį́į́ʼ (NX-027; Willie 1991:61 (22)), “The boy picked them.”; ʼashkii yiztał (NX-444; 129 (30)), “The boy kicked it.”LK-1 allows the noun in either third-person shell; nothing in the grammar chooses, so the choice is a default or context. That is what H4″ says.NX-444: ∃e[kick(e) ∧ Agent(e,x) ∧ Patient(e,y)], x = the boyfits
Indefinite and quantified referents: ʼashkii léiʼ ʼatʼééd yiztsʼǫs (NX-038; Willie 1991:76 (43)), “A boy kissed the girl.”; NX-343, “… hit some boy”; tʼáá ʼałtso ʼatʼééké ʼashkii łaʼ deiztsʼǫs (NX-267; 80 (60)), both scopesA prefix pronoun can be linked to an indefinite or quantified noun, so D⁰ must be a variable with φ-features only (Déchaine & Wiltschko 2002:410 (2): pro-φP, “no inherent semantics”, a variable), not a definite pro-DP. Willie & Jelinek (2000:259) note the contrast: Spanish and Egyptian Arabic object clitics never double indefinites; Navajo prefixes do.NX-267, surface: ∀x[girl(x) → ∃y[boy(y) ∧ ∃e[kiss(e) ∧ Agent(e,x) ∧ Patient(e,y)]]]; other scope by H22a or H22bfits with D⁰ = φ-variable; Déchaine & Wiltschko’s “obviative = D” (2002:438 (79)) is contradicted for the Navajo obviative object, which is linked to an indefinite in NX-343 (narrow claim)

3b.3 Alternatives compared

Analysis (source opened)What it shares with LKWhere it differsWhat would decide
Plain Merge-site linearization, P1 (§3.1–3.2); sublexical V-to-T (§3.4)Prefixes are heads of the spineOne tier only: the object/subject and distributive misfits stay. LK-p is the Merge-site idea applied tier by tier.P-LK1, P-LK3
Pronominal Argument Hypothesis: Jelinek 1984:50, 52 (31) (Linking Rule: “A clitic pronoun may be coindexed with a nominal, providing the L-case of the nominal and the G-case of the clitic pronoun are compatible”); Willie & Jelinek 2000:252, 258, 260; Hale 2003 ms.:5–7The prefixes are the arguments; a matching condition links each noun to one prefix (W&J 2000:260: nominal adjuncts “must match this pronoun in phi-features”)Nouns are adjuncts and nothing moves (W&J 2000:258: “We do not assume that the weak pronouns undergo movement”). On LK the noun is Merged in the argument shell and moves, so it should reconstruct.P-LK2 (reconstruction), P-LK5
Agreement only with dislocated NPs: Baker ms. “Agreement, Dislocation, and Partial Configurationality”:5 (3), 35 (49); Navajo left open (p. 47)Agreement plus overt nounsOn LK the noun sits inside the argument, not in a dislocated adjunct positionP-LK2, P-LK5
Clitic doubling as big DP / D adjoined to v: Baker & Kramer 2018 ms.:2, 5 (7), 12 (22) (“A D adjoined to head H is referentially dependent on the DP it agrees with via H.”); Preminger 2019 and Harizanov 2014 (abstracts only)The shell is a big DP; the prefix is a D that head-moves after AgreeB&K’s object D adjoins to v (= LK-h, which reproduces the misfit). Navajo doubling is obligatory and allows indefinites (W&J 2000:259), so D⁰ cannot be B&K’s definite pronoun. B&K derive weak crossover and Condition B effects from (22).P-LK5
Grammatical against anaphoric agreement: Bresnan & Mchombo 1987:741, 745 (Chicheŵa subject marker ambiguous, object marker “unambiguously anaphoric”)A subject/object asymmetry in the prefixesLK-p puts the asymmetry in the tiers: the subject pronoun fuses with T (agreement-like), the object pronoun stays a phrase (pronoun-like). B&M’s criteria (independent nouns, word order freedom with the marker present) have not been applied to Navajo.P-LK6, P-LK7
Hale’s two planes: Hale 2003 ms. §3, 29–40Nuclear/nonnuclear = head/phrasal tier, object on the nonnuclear planeHale’s procedure is phonological and nothing moves; the object stays in its thematic position (pp. 10–12, with the surface “as if it were in the Spec position projected by” M/S, pp. 11–12). LK-p says the same order comes from tiers of movement.Whether any order fact needs movement rather than Hale’s filling procedure: OPEN

3b.4 Light d, light n and the DP/TP symmetry

HYPOTHESIS Light d stays withdrawn, and L is not light d. Light d was a silent D on bare nouns that did no work. The link head in LK is the overt, obligatory prefix pronoun, which takes the noun as its specifier; LK adds no silent element. What the two share is only that a D-type head sits above a noun’s position; L does a different job (it ties a noun to a theta role), and nothing on this page revives light d.

OPEN Light n. On LK the possessor shell sits in a left specifier inside the noun phrase, the position light n would project (Grant, 2026-11-14). LK removes one argument against light n: the possessor prefix having the shape of an object prefix is explained by the tier, so it is not evidence about n. It gives no argument for light n either. DP/TP symmetry: the shell gives the same relation in both domains (a pronoun head with its referent in its specifier: subject and object shells in the clause, possessor shells in the noun phrase, P-object shells in PP), which is the symmetry the site has been looking for; whether there is a TP-like layer in the noun phrase that the possessor’s D⁰ head-moves to, as the subject’s D⁰ does to T⁰, is P-LK6. OPEN

3b.5 Predictions (all OPEN) and what would decide them

  1. P-LK1 Tier membership. Prefixes that can be separated from the verb or host a clitic should all be phrasal-tier, left of the head complex. Hale reports the preverb separable by -ísh (Hale 2003 ms.:69 n. 12, citing Platero p.c.). Deciding data: whether =ísh or another clitic can follow the object or da- but never M/S. OPEN
  2. P-LK2 Reconstruction. A noun is Merged in the argument shell, so a moved noun should reconstruct for scope and binding into its theta position; on the pronominal-argument view nouns are adjuncts and should not. Deciding data: the inverse reading of NX-267 (“many boys”) with the object noun fronted, and binding of a possessor pronoun inside a fronted object by the subject. OPEN
  3. P-LK3 Tier against height. If da- and the preverb are both phrasal and ordered by height, a preverb Merged low with a high da- should give da- left of the preverb, against the fixed template. Deciding data: a low preverb (Hale’s (14), (16), (21)) with distributive da-; the template order would make the preverb’s leftward move obligatory. OPEN
  4. P-LK4 Respect ji- is third person. It should behave as a third person in the inverse and in H31 contexts, and should not fall under H32b (two fourth-person pronouns in one sentence refer to the same person). Deciding data: a respect ji- clause with a second, disjoint fourth-person pronoun, and one with a second third person in the inverse. OPEN
  5. P-LK5 No Baker & Kramer effects. With D⁰ a φ-variable, there should be no weak crossover or Condition B restriction on a quantified or bound noun doubled by a prefix. Deciding data: a bound-variable reading of a possessor prefix with a quantified antecedent (tʼáá ʼałtso), in direct and inverse order. OPEN
  6. P-LK6 No possessor portmanteau. If the possessor’s D⁰ stays phrasal (like the object’s), it should never fuse with a noun-internal head as the subject fuses with mode. Deciding data: any possessed noun whose possessor prefix changes form with a noun-internal category. OPEN
  7. P-LK7 ji- never fuses with mode. If the deictic subject is the highest head of the head complex but outside M/S, no mode should have a ji- portmanteau. Deciding data: the deictic subject across all modes (Hale 2003 ms.:15 refers to Hale 2001 for deictic subjects; not opened here). OPEN

3b.6 Sources

SOURCED Opened for this section: Hale 2003 ms., “On the Significance of Eloise Jelinek’s Pronominal Argument Hypothesis”, pp. 5–12, 14–21, 26–40, 67–69; Willie & Jelinek 2000, “Navajo as a Discourse Configurational Language”, in Fernald & Platero (eds.), pp. 252–272; Jelinek 1984, Natural Language and Linguistic Theory 2, pp. 50–52; Aissen 2000, “YI and BI: Proximate and Obviative in Navajo”, pp. 1–2, 6; Baker & Kramer 2018, Natural Language and Linguistic Theory 36:1035–1088 (prepublication manuscript, Rutgers PDF), ms. pp. 2, 5, 12–13; Baker, “Agreement, Dislocation, and Partial Configurationality” (Rutgers ms.), pp. 2, 5, 31, 35, 47; Bresnan & Mchombo 1987, Language 63:741–782 (Stanford PDF), pp. 741, 745; Déchaine & Wiltschko 2002, Linguistic Inquiry 33:409–442, pp. 410, 432, 438. Abstract only: Preminger 2019, Glossa 4(1):13; Harizanov 2014, NLLT 32:1033–1088. Not opened (cited only through the sources above): Uriagereka 1995 and Torrego (big DP), Matushansky 2006, Speas 1990, Hale 1973, Hale 2001 on deictic subjects (as referred to by Hale 2003 ms.), Baker 1996, Diesing 1992, McDonough 1990, Platero (p.c. in Hale).

3b.7 Two-tier linking: follow-up (2026-11-15/16)

2026-11-15/16 · follow-up on the two-tier linking model. Grant (2026-11-15), after reading §3b.1–3b.6: “It looks like this is correct. Continue to look into it.” Then: “I thought the verb args had something to do with thetas rather than pronouns... It’s possible there’s still some connection... Perhaps it explains bi-/yi-?” This subsection tests P-LK1–P-LK7 against data already on the site and in free sources opened for it, tries to remove each strain of §3b without adding a stipulation, adds Grant’s theta-linking strand (TL) and tests it on yi-/bi-, and gives an LF for every derivation, with scope matching c-command. It adds no Navajo data. Every form is copied from the example index under its NX number, or quoted from an opened page with its page number. Forms that print in a broken font in a source are described and not quoted.

Verdict in plain words. GRANT'S HYPOTHESIS HYPOTHESIS The theta-linking strand works for yi-/bi-, with one borrowed ingredient. Read each prefix pronoun as the place where the verb discharges a theta role. Then yi- on the internal role’s pronoun means that the noun linked to the external role outranks it and c-commands it. bi- means it does not. Given a ranking of third persons (animacy, then topicality: Aissen 2000, already the site’s rule R4 on one proximate), one LF condition does the work of two old statements. It gives H9′ (“yi- needs a distinct third-person co-argument above it”) and the trigger of H3 (the patient moves above the agent when it outranks it). It also derives the starred inverse question *NX-073 against NX-074 and NX-068, the fourth-person facts, possessor b-/yi-, and Willie & Jelinek’s starred coreference in (70b) (as Condition C). The ranking itself is imported, not derived. Model changes: the extra rule LK-4 is gone. da- becomes an adjoined quantifier, H21a′, which is Hale’s own category for it. The deictic subject ji- is a referent, not a pronoun head (Hale 2001:690–691), which removes the respect-ji- strain. The split preverb gets a two-tier account: the stem leaves by head movement, the preverb stays behind as a phrase. Still unexplained is why the preverb is pronounced at the left edge: P-LK3 in its first form fails, because the preverb is inside the scope of da- yet left of it (Rice, as Hale 2001:689 states it). Predictions: P-LK7 passes on narrow evidence, P-LK3 fails as first stated and is replaced, P-LK4 is replaced, and P-LK1, P-LK2, P-LK5 and P-LK6 stay OPEN, each with the pair that would decide it. OPEN

What changed in the model

  1. TL replaces LK-4 (GRANT'S HYPOTHESIS, Grant 2026-11-15; our formalization, HYPOTHESIS):
    • TL-1. The verb’s theta grid is a set of D⁰ slots. Each theta role is discharged on one D⁰, the prefix pronoun, in its VPISH position. A noun has a role only through its link to a D⁰ (LK-1). So “the arguments are theta roles” and “the arguments are pronouns” are one claim: the pronoun is the theta slot as it is pronounced.
    • TL-2. Ordinary third-person referents are ranked: animacy first, then topicality (Aissen 2000, (6), (32); imported). Fourth persons are off the ranking (Aissen’s OP-2, one proximate §9).
    • TL-3. The link copies the result onto the internal role’s D⁰. That D⁰ is spelled yi- if the external role’s referent outranks its referent, and bi- otherwise. This is the site’s R4/OP-7, now stated through the link.
    • TL-4 (LF condition). At the A-positions, the outranking referent c-commands the outranked one. A-bar movement (focus, wh) is not counted.
  2. H21a → H21a′. da- is a distributive quantifier phrase adjoined to vP, above the outer specifier the inverse patient passes through (H21b, kept). It is not a head on the V→v→T path. Hale classes it this way, as the “argument quantifier (Pl)”, an adjunct of the disjunct sector (Hale 2003 ms.:16–17, (11)), and it can attach at different heights (2003 ms.:71 n. 14, citing Rice 2000:80–81). Its counting is semantic: it distributes over the closest plural λ-abstract it c-commands. It is not φ-Agree, which an adjunct cannot do.
  3. Deictic ji- is a referent. Following Hale 2001:690–691, the position-V prefixes “are not invested with a person feature”. They are “adjunct nominal expressions linked to true arguments”, third person only as a full DP is. The subject argument itself is in the Mode/Subj portmanteau, where third person has an overt allomorph (Hale 2001:691, (24)). On LK, ji- sits in the specifier of the subject shell, linked to a third-person D⁰ that has fused with T⁰.
  4. Preverb. A lexical preverb is Merged low, with V (Hale 2003 ms.:19–21, (14), (16)). V leaves by head movement (tier 1). The preverb is a phrase, so it stays behind (tier 2). This is the stranding of §3.4, moved into LK. It is pronounced in one left-edge slot of the word and interpreted in its low position (see P-LK3).

The predictions, tested

PredictionEvidence openedVerdictMinimal pair that would decide it
P-LK1 Tier membership: only phrasal-tier pieces can be split off by a cliticSOURCED The only split on record is ka-[á]sh nanin[á] “are you sick?” (Platero p.c., Hale 2003 ms.:69 n. 12; vowels restored on where -ísh goes §2). There ka stands as its own word, so the split is between words. No opened form puts -ísh inside a verb word (EN-1′). New tier evidence that needs no clitic: an oblique-object pronoun is pronounced either leftmost (Young & Morgan’s Ia, before the iterative) or in the object position IV when a preverb of their class Ib is present. It is never inside the mode/subject complex (Young & Morgan 1987:187, as quoted by Hale 2003 ms.:69–70 n. 13; Hale’s example prints in a broken font and is not quoted). Willie’s glosses put the reflexive before the plural, ʼádadiidleesh “REF.PL.1dl.painting” (NX-493; Willie 1991:34 (14)), but the distributive before the reciprocal, daʼahiidleesh “DIST.REC.1dl.painting” (NX-497; 35 (18)).OPEN for the clitic test. A pronoun that can sit on either side of the adjunct prefixes is what a phrasal-tier pronoun can do and a head-adjoined one (LK-h) cannot. This favours LK-p (narrow claim).A verb with a separable preverb (like ka) plus da- or an overt object prefix, with -ísh: can it follow da- or the object prefix, and never the mode/subject prefix? Second pair: a reflexive verb with a class-Ib preverb and da-. If the reflexive competes for the edge like the oblique pronoun, it should move to the right of the preverb.
P-LK2 Reconstruction: a noun is Merged in its shell, so a moved noun has a lower copySOURCED NX-267 cannot decide this. It is subject-first, and both copies of the subject (Spec,TP and Spec,vP) c-command the object, so reconstruction changes nothing. Its “one boy” reading can also come from łaʼ alone (H22b). The inverse version, tʼáá ʼałtso ʼatʼééké ʼashkii łaʼ biztsʼǫs (NX-265; Willie 1991:79 (58)), translated “All the girls were kissed by a boy.”, is printed with one translation and no reading split. Binding: in Willie & Jelinek 2000:276 (70), “[All John’s children] saw him”, coreference of “him” with John is out under bi- (b), and the disjoint reading is fine (a).OPEN for reconstruction. (70b) shows the opposite side: the raised patient referent (silent) c-commands into the agent (Condition C; derived under TL below). So A-movement of the inverse patient creates c-command and does not reconstruct for Condition C. That is a fit for LK. On the pronominal-argument view (nouns adjoined, nothing moves), it needs Platero’s precedence condition instead (possessor command).An NX-265-type inverse where the agent is a quantifier that cannot take wide scope by itself (not łaʼ, léiʼ or a name), checked for the reading in which it scopes over the fronted patient. Or: a fronted bi- patient containing a possessor pronoun bound by a quantified agent (“his own mother, every boy bi-saw”).
P-LK3 Tier against height: a low preverb with a high da-SOURCED Rice’s scope order inside the disjunct zone, as Hale states it: “scope is right to left … an element with wide scope is to the right of elements appearing within its scope”; “the iterative … and the plural … [da] have the preverb within their scope” (Hale 2001:689). The same preverb yisdá ‘to safety’ is leftmost whether it scopes over subject and object (NX-107, Yisdánídashizhdoołtééł, “They (3a = 4th person) will get me back to safety.”; Hale 2003 ms.:14 (9), (10); scope stated p. 16) or over the object only (Hale 2003 ms.:18–19 (13), (14), “I rescue it.”).Fails as first stated. If the phrasal tier were ordered by height, with specifiers and adjuncts on the left, da- (wider scope) would be left of the preverb. It is right of it. Repair (HYPOTHESIS): the preverb is pronounced in the word’s single left-edge slot and interpreted in its lower copy, so scope = c-command holds at LF. The edge slot is real: it is the one the oblique pronoun competes for (P-LK1). Its trigger is not derived: OPEN.NX-107 under the object-only reading of yisdá (only “me” comes to be in safety). If the string allows that reading, a low preverb is pronounced left of da-, and the edge account is needed. If it does not, Hale’s high attachment for (9) and Rice’s scope conflict for this word.
P-LK4 Respect ji- is third personSOURCED Hale 2001:690–691: the deictic subjects (3a = fourth person, 3i, 3s) have no person feature of their own and are linked to the third-person subject in Mode/Subj. Hale segments shijiłhozh “He/She/They/One is tickling me.” as Object · 3a · Mode/Subj · Voice · V (2001:689 (22a)).Replaced. If every ji- is a referent linked to a third-person D⁰, LK-1 lets a noun co-occur with it whether it is respect or reference (NX-470). The choice between H57a and H57b no longer matters to LK. New P-LK4′: what separates respect ji- from reference ji- is whether it is ranked (TL-2), not its person. OPENA respect ji- subject (as in NX-470) in a transitive clause with a third-person object. Fourth persons are off the ranking, so the object should be bi-/default (Aissen 2000, Table 1). If respect ji- is ranked, yi- should be possible.
P-LK5 No Baker & Kramer effects (D⁰ is a φ-variable)No bound-variable datum with a doubled quantified noun has been opened. (70b) is Condition C with a name, not weak crossover.OPENA quantified tʼáá ʼałtso antecedent binding a possessor prefix, in the direct and the inverse, with the bound reading checked.
P-LK6 No possessor portmanteauNo possessed noun with a possessor prefix that changes with a noun-internal category was found in the index.OPEN (silence counts for nothing)Any possessed noun whose possessor prefix varies with a noun-internal category.
P-LK7 ji- never fuses with modeSOURCED Hale 2001:689–691: in (22a) j (3a) is a separate piece from Mode/Subj. Third person is inside Mode/Subj, with the rounded allomorph [w] in the future and progressive (24a, b).Passes, narrowly: one paradigm cell, and Hale’s segmentation, not a minimal pair.The fourth-person subject across all modes, against third person, looking for a mode whose fourth person has no separate j/zh piece.

Theta-linking and yi-/bi- (Grant, 2026-11-15)

GRANT'S HYPOTHESIS OPEN Grant asked whether bi-/yi- reflects which theta role’s pronoun the higher noun links to, together with animacy. On TL the answer is yes. yi- on the internal role’s D⁰ says that the external role’s referent outranks it (and, by TL-4, c-commands it at an A-position). bi- says it does not. So the prefix reports which theta slot the most prominent third person is linked to. Each row below runs the derivation and gives the LF. Scope = c-command; [prx]/[obv] are TL-3’s values.

DatumDerivation on LK + TLLFVerdict
łį́į́ʼ dzaanééz yiztał (NX-029; Willie 1991:71 (37a)), “The horse kicked the mule.” against dzaanééz łį́į́ʼ biztał (NX-030; 71 (37b)), “The mule was kicked by the horse.”Equal animacy, so topicality ranks them (Aissen (12); Aissen’s §1 on these sentences: (1b) is “about” the patient). NX-029: the horse outranks the mule and c-commands it from Spec,TP, so the object D⁰ is yi-. NX-030: the mule outranks the horse, so TL-4 forces the mule’s referent above the agent in Spec,TP (the H3 move), and the object D⁰ is bi-.NX-029: [the horse] λx [[the mule] λy ∃e[kick(e) ∧ Agent(e,x) ∧ Patient(e,y)]]. NX-030: [the mule] λy [[the horse] λx ∃e[kick(e) ∧ Agent(e,x) ∧ Patient(e,y)]]passes. H9′ and H3’s trigger follow from TL-4. The landing site is only “an A-position above Spec,TP” (OPEN).
*haʼátʼíísh John biiłtsą́ (*NX-073; Willie 1991:206 (10b)) [What was seen by John?], against NX-074 (206, (9), second on the page), “What did John see?”, with yi-, and háísh John biztsʼǫs (NX-068; 204 (8a)), “Which one (of you) was kissed by John?”*NX-073: bi- needs the wh referent to outrank John, and an inanimate cannot outrank a human (Willie gives the animacy hierarchy as the reason, p. 206). NX-074: John outranks ‘what’, so yi-. The wh word moves to Spec,CP by A-bar focus movement (Willie: “a kind of focus movement”, pp. 206–207), which TL-4 does not count. NX-068: the wh word is human and presupposed (“which one”), so it can outrank John, and bi- is fine. Tree below.NX-074: ASKy [thing(y)] [John λx ∃e[see(e) ∧ Agent(e,x) ∧ Theme(e,y)]]. *NX-073: the same with y[prx] above John, undefined because [prx](y) fails. NX-068: ASKy [y ∈ C] [y λy [John λx ∃e[kiss(e) ∧ Agent(e,x) ∧ Patient(e,y)]]]passes. The NX-073/074 pair falls out without a special rule for questions. On Aissen’s own linear constraint (proximate precedes obviative, her (11)), NX-074 needs an exception for focus. On TL-4 it does not.
ʼashkii ʼatʼééd léiʼ biztsʼǫs (NX-041; Willie 1991:77 (46)), “The boy was kissed by a girl.” against *ʼashkii léiʼ ʼatʼééd biztsʼǫs (*NX-042; 77 (47))NX-041: the definite boy outranks an indefinite girl by topicality, so bi-, and he is above her. *NX-042: bi- needs the indefinite boy to outrank the definite girl, and an indefinite cannot be the topic (Aissen’s (35) discussion).NX-041: [the boy] λy ∃x[girl(x) ∧ ∃e[kiss(e) ∧ Agent(e,x) ∧ Patient(e,y)]]passes, with the extra premise that an indefinite cannot be a topic (H6′ restated as a topicality fact, not derived).
Fourth person: John jiztał (NX-468; Willie 1991:112 (7b)), “She kicked him, John.”; John haztał (NX-469; 112 (7c)), “John kicked that certain person.”; yiztał=yę́ę jiiłtsą́ (NX-353; 172 (47)), “That person saw the one who was kicked.”, main verb glossed 4.3With the fourth person as subject or object there is only one ranked third person, so nothing outranks the object and no yi- appears. That is what NX-353’s main verb shows, against the 3.3 verb yiyiiłtsą́ (NX-072; 206 (10a)), which has it. Aissen 2000, Table 1, gives the object as BI with a 1st, 2nd or 4th person subject. John links to the only third-person shell (LK-1).NX-468: ∃e[kick(e) ∧ Agent(e,z) ∧ Patient(e,y)], y = John, z = the ji- referent. NX-469: ∃e[kick(e) ∧ Agent(e,x) ∧ Patient(e,z)], x = Johnpasses. Why fourth persons are off the ranking is not derived. It agrees with Willie 1991:127 (as Aissen reports it), where the fourth person “overrides” the animacy hierarchy.
Possessor: John háísh baghan=di yiztsʼǫs? (NX-381; Willie 1991:216 (28a)), “Who did John kiss at his house?”, against John háísh yighan=di yiztsʼǫs? (NX-382; (28b)), “Who did John kiss at her house?” (the index records speaker variation)The possessor D⁰ copies the value of its referent through the link. In NX-381 it is John, who is not outranked, so b-. In NX-382 it is the person kissed, outranked by John, so yi-.NX-381: ASKy [John λx ∃e[kiss(e) ∧ Agent(e,x) ∧ Patient(e,y) ∧ Loc(e, ιh[home(h) ∧ Poss(h,x)])]]. NX-382: the same with Poss(h,y)passes, for the speakers Willie reports
Willie & Jelinek 2000:276 (70): “[All John’s children] saw him”, with da- and bi-: disjoint reading (a) fine, coreference with John (b) outbi- says the silent patient referent outranks the agent, so by TL-4 it sits above the agent DP and c-commands John inside it. Coreference would violate Condition C. Ranking alone (Aissen) predicts (b) good, which is T58a on possessor command. TL with the H3 move gets the star.(70a): [prok] λy [[all Johni’s children] λX [DA [X λx ∃e[see(e) ∧ Agent(e,x) ∧ Patient(e,y)]]]], k ≠ i; (b) needs k = i, so prok c-commands Johni: Condition Cpasses, and it needs the movement, which the pure ranking account lacks (the deep-inverse question D2 in possessor binding; narrow claim, one datum)
*NX-345 against NX-346 (Willie 1991:169–170 (39), (40)): two inverse clauses, rescued by léiʼAn account by “one proximate per span” would star NX-345 (two disjoint bi- referents), but the site has already found that rule contradicted by two bi- objects in one clause (one proximate, OP-5). TL uses the subject-relative R4, which says nothing here.—NOT UNDERSTOOD (unchanged)

HYPOTHESIS Verdict on the theta-linking strand. On the tested data it passes. It does more than restate H3 and H9′: one LF condition (TL-4) yields H9′’s “above”, the trigger of H3, the animacy gaps and the question pair *NX-073/NX-074. Given the H3 move, it also yields the Condition C star in (70b), which the ranking-only account does not. It does not derive the ranking, which is Aissen’s and Creamer’s, imported. Nor does it derive why fourth persons are unranked, the landing site of H3, or *NX-345 against NX-346. Whether it is the theta slot or the pronoun that carries [prx]/[obv] makes no difference on TL-1, since they are one object. That is how Grant’s two descriptions, “thetas” and “pronouns”, connect. OPEN

Analysis A: NX-074 haʼátʼíísh John yiviiłtsą́ “What did John see?” (verb as printed). John is [prx] in Spec,TP; the wh referent is [obv] (yi-) and moves by A-bar movement to Spec,CP; working default Hypothesis

Analysis A: NX-074 haʼátʼíísh John yiviiłtsą́ “What did John see?” (verb as printed). John is [prx] in Spec,TP; the wh referent is [obv] (yi-) and moves by A-bar movement to Spec,CP; working defaultShells: [DP referent [D′ D⁰]], specifiers on the left; L is a relation, not a node (light d stays withdrawn). Theta-linking strand (GRANT'S HYPOTHESIS, OPEN; our formalization TL-1 to TL-4, HYPOTHESIS): each theta role is discharged on a pronoun D⁰ (the prefix); a noun gets its role only through the link to that D⁰ (LK-1). Among the third-person referents of one clause exactly one is proximate [prx], ranked animacy > topicality (Aissen 2000, as her ranking; imported, not derived); the link copies [prx]/[obv] onto the D⁰, and the object D⁰ spells it: bi- = [prx], yi- = [obv]. TL-4 (LF condition): at the A-positions (before any A-bar move) the [prx] referent c-commands the [obv] one. Working default: V⁰ head-moves through v⁰ to a final T⁰ (V→v→T), and the subject’s referent raises from the left specifier of its shell in Spec,vP to left Spec,TP; lower copies are silent. Panel A. Merge V ‘see’ with the object shell (Theme; the wh word in its specifier) and the subject shell in Spec,vP (Experiencer/Agent; John). Ranking: John (human) outranks ‘what’ (inanimate), so John is [prx] and the wh referent [obv]; the object D⁰ is spelled yi-. The wh referent then moves to Spec,CP; this is A-bar movement (Willie 1991:206–207: “a kind of focus movement of the Wh- word”), so it does not count for TL-4, which is checked at the A-positions, where John in Spec,TP c-commands the lower copy. Scope = c-command: Q in C⁰ over everything; John over the object’s lower copy. LF: ASK_y [thing(y)] [John λx ∃e[see(e) ∧ Agent(e,x) ∧ Theme(e,y)]] Full structure in the list below the figure.CPDPhaʼátʼííshC′TPC⁰Q (=sh pronounced on the first word)DPJohnT′vPT⁰DPv′DPJohnD′VPv⁰D⁰x: subject pronoun, 3rd, linked to John [prx]V′DPV⁰yiviiłtsą́DPhaʼátʼííshD′D⁰y: object pronoun yi- [obv]verb word (head tier; yi- on the phrasal tier, left)head-movehead-movehead-moveAgree link: 3rd person, [obv] → yi-MoveMove

Analysis B: *NX-073 haʼátʼíísh John biiłtsą́ [What was seen by John?]: the bi- derivation needs the wh referent to be [prx] and to c-command John at an A-position; ranking forbids it, so the derivation is not interpretable; working default Hypothesis

Analysis B: *NX-073 haʼátʼíísh John biiłtsą́ [What was seen by John?]: the bi- derivation needs the wh referent to be [prx] and to c-command John at an A-position; ranking forbids it, so the derivation is not interpretable; working defaultShells: [DP referent [D′ D⁰]], specifiers on the left; L is a relation, not a node (light d stays withdrawn). Theta-linking strand (GRANT'S HYPOTHESIS, OPEN; our formalization TL-1 to TL-4, HYPOTHESIS): each theta role is discharged on a pronoun D⁰ (the prefix); a noun gets its role only through the link to that D⁰ (LK-1). Among the third-person referents of one clause exactly one is proximate [prx], ranked animacy > topicality (Aissen 2000, as her ranking; imported, not derived); the link copies [prx]/[obv] onto the D⁰, and the object D⁰ spells it: bi- = [prx], yi- = [obv]. TL-4 (LF condition): at the A-positions (before any A-bar move) the [prx] referent c-commands the [obv] one. Working default: V⁰ head-moves through v⁰ to a final T⁰ (V→v→T), and the subject’s referent raises from the left specifier of its shell in Spec,vP to left Spec,TP; lower copies are silent. Panel B. Same Merge steps and theta roles as panel A. bi- spells [prx] on the object D⁰, so by TL-4 the wh referent must be [prx] and must c-command John from an A-position above Spec,TP (the H3 move, now triggered by TL-4). Two conditions fail: an inanimate cannot outrank a human (Aissen’s animacy ranking; Willie 1991:206 gives the animacy hierarchy as the reason for the star), and a plain question variable is not a topic. The [prx] presupposition on y is therefore not met and the LF is undefined; Willie stars the string (1991:206 (10b)). Contrast NX-068 háísh John biztsʼǫs “Which one (of you) was kissed by John?”, where the wh is human and presuppositional and bi- is fine. Scope = c-command, as the derivation would give it if defined. LF: ASK_y [thing(y)] [y_[prx] λy [John_[obv] λx ∃e[see(e) ∧ Agent(e,x) ∧ Theme(e,y)]]], undefined because [prx](y) fails Full structure in the list below the figure.CPDPC′TPC⁰Q (=sh pronounced on the first word)DPhaʼátʼííshTPDPJohnT′vPT⁰DPv′DPJohnD′VPv⁰D⁰x: subject pronoun, 3rd, linked to JohnV′DPV⁰biiłtsą́DPhaʼátʼííshD′D⁰y: object pronoun bi- [prx] (fails)verb word (head tier; bi- on the phrasal tier, left)head-movehead-movehead-moveAgree link: 3rd person, [prx] → bi- (ranking violated)MoveMoveMove
(2) Theta-linking and obviation on the two-tier model (GRANT'S HYPOTHESIS, OPEN): yi- kept under A-bar focus, bi- blocked for an inanimate question word Hypothesis NX-074: Willie 1991:206 ((9), second on the page): haʼátʼíísh John yiviiłtsą́ “What did John see?” (verb as printed); *NX-073: Willie 1991:206 (10b): haʼátʼíísh John biiłtsą́ [What was seen by John?]; NX-068 for contrast. Leaves are printed words. The shells, D⁰, the [prx]/[obv] values, the landing sites and the role labels are our assignments.
Bracket notation (copy)

A: [CP [DP haʼátʼíísh] [C′ [TP [DP John] [T′ [vP [DP ⟨DP⟩ [D′ [D⁰ x: subject pronoun, 3rd, linked to John [prx]]]] [v′ [VP [V′ [DP ⟨DP⟩ [D′ [D⁰ y: object pronoun yi- [obv]]]] [V⁰ yiviiłtsą́]]] v⁰]] T⁰]] [C⁰ Q (=sh pronounced on the first word)]]] B: [CP DP [C′ [TP ⟨DP⟩ [TP [DP John] [T′ [vP [DP ⟨DP⟩ [D′ [D⁰ x: subject pronoun, 3rd, linked to John]]] [v′ [VP [V′ [DP ⟨DP⟩ [D′ [D⁰ y: object pronoun bi- [prx] (fails)]]] [V⁰ biiłtsą́]]] v⁰]] T⁰]]] [C⁰ Q (=sh pronounced on the first word)]]]

Structure as a list

Analysis A

  • CP
    • DP (id W-high; landing site of Move from Ro): haʼátʼíísh
    • C′
      • TP
        • DP (id S-high; landing site of Move from Rs): John
        • T′
          • vP
            • DP (id shellS)
              • DP (id Rs; struck copy; moved to S-high): John
              • D′
                • D⁰ (head; id Ds; head-moves to T): x: subject pronoun, 3rd, linked to John [prx]
            • v′
              • VP
                • V′
                  • DP (id shellO)
                    • DP (id Ro; struck copy; moved to W-high; goal of Agree link: 3rd person, [obv] → yi-): haʼátʼíísh
                    • D′
                      • D⁰ (head; id Do; probe for Agree link: 3rd person, [obv] → yi- with Ro): y: object pronoun yi- [obv]
                  • V⁰ (head; id V; head-moves to v): yiviiłtsą́
              • v⁰ (head; id v; head-moves to T)
          • T⁰ (head; id T; hosts the “verb word (head tier; yi- on the phrasal tier, left)” (V + v + T + Ds))
      • C⁰ (head; id C): Q (=sh pronounced on the first word)

Analysis B

  • CP
    • DP (id W-high; landing site of Move from O-high)
    • C′
      • TP
        • DP (id O-high; landing site of Move from Ro; struck copy; moved to W-high): haʼátʼíísh
        • TP
          • DP (id S-high; landing site of Move from Rs): John
          • T′
            • vP
              • DP (id shellS)
                • DP (id Rs; struck copy; moved to S-high): John
                • D′
                  • D⁰ (head; id Ds; head-moves to T): x: subject pronoun, 3rd, linked to John
              • v′
                • VP
                  • V′
                    • DP (id shellO)
                      • DP (id Ro; struck copy; moved to O-high; goal of Agree link: 3rd person, [prx] → bi- (ranking violated)): haʼátʼíísh
                      • D′
                        • D⁰ (head; id Do; probe for Agree link: 3rd person, [prx] → bi- (ranking violated) with Ro): y: object pronoun bi- [prx] (fails)
                    • V⁰ (head; id V; head-moves to v): biiłtsą́
                • v⁰ (head; id v; head-moves to T)
            • T⁰ (head; id T; hosts the “verb word (head tier; bi- on the phrasal tier, left)” (V + v + T + Ds))
      • C⁰ (head; id C): Q (=sh pronounced on the first word)

The strains, one by one

Strain (§3b)Principled repair triedLFResult
Inverse: LK-4 only restated H3 + H9′TL-1 to TL-4 (above). The yi-/bi- choice is the value of the internal role’s D⁰, read through the link from a ranking of referents. The landing site is a specifier at an A-position above Spec,TP (TL-4 counts A-positions only; Saxon’s A-like Spec,DiscP fits, possessor command update).As NX-030 aboveresolved in part: trigger derived, ranking imported, exact landing site OPEN
Respect ji- forced H57aji- is a referent linked to a third-person D⁰ in Mode/Subj (Hale 2001:690–691)NX-470: ∃s[called(s) ∧ Theme(s,x) ∧ Name(s,Baaʼ)], x = my mother = the ji- referent (two referents for one D⁰: a noun and ji-)resolved for LK-1. New question: how one shell takes both a noun and ji-, and whether respect ji- is ranked (P-LK4′) OPEN
da- had to be phrasal, against H21a’s Dist⁰H21a′: da- is an adjoined distributive quantifier phrase (Hale 2003 ms.:16–17 (11)), adjoined to vP above the inverse patient’s outer specifier. Its c-command domain is the same as Dist⁰’s, so H21a’s and H21b’s facts carry over unchanged. The verb no longer passes through a Dist⁰ head.NX-267 (many boys): [all the girls] λX [DA [X λx ∃y[boy(y) ∧ ∃e[kiss(e) ∧ Agent(e,x) ∧ Patient(e,y)]]]], with DA(P)(X) = ∀x[x ≤atom X → P(x)]. *NX-273: DA’s closest abstract is over “the boy”, which is singular, so DA is undefined.resolved (HYPOTHESIS). The exemptions for demonstrative, possessed and pronoun patients (H21b) stay OPEN.
Relatives need Agree + λ; *NX-345 vs NX-346No new repair. The enclitic’s link into CP is still λ-abstraction over the relativized D⁰’s variable (as in §3b.2).NX-312: ιx ∃e[kick(e) ∧ Agent(e,y) ∧ Patient(e,x)]NOT UNDERSTOOD (unchanged)
Split preverbTwo tiers. The preverb is Merged low with V. V head-moves V→v→T (tier 1), and the preverb, a phrase, is stranded (tier 2), so the two end up apart. The preverb is pronounced at the left edge, in the slot the oblique pronoun otherwise takes (P-LK1), and interpreted low.Hale (13)/(14) (2003 ms.:18–19), “I rescue it.”: [speaker λx ∃e[rescue(e) ∧ Agent(e,x) ∧ Theme(e,y) ∧ ∃s[RESULT(e,s) ∧ in-safety(s,y)]]]. The preverb’s predicate takes only y, as Hale argues (p. 19).split derived. Edge position: trigger OPEN. P-LK3 is the test.
Passive u: raising becomes head movementNo new data. With no referent there is nothing to rank, so TL predicts the default (non-yi-) form on the patient. That fits niʼdilzhééh (NX-020) and shiʼdidleesh (NX-021), which have a local patient.NX-020: ∃u∃e[shave(e) ∧ Agent(e,u) ∧ Patient(e,y₂)]OPEN: whether Spec,TP is filled

Next decisive data

  1. P-LK3: the object-only reading of yisdá in NX-107. If it is available, a low preverb sits left of da-, and the edge slot is needed.
  2. P-LK2/P-LK5: a bi- clause whose fronted patient contains a possessor pronoun bound by a quantified agent; an NX-265-type inverse whose agent cannot take wide scope by itself.
  3. P-LK4′: a respect-ji- subject with a third-person object: is yi- possible?
  4. TL, separating animacy from topicality: an inverse question with a presupposed (“which”) inanimate wh word against a human subject. Animacy predicts *, topicality alone predicts OK.
  5. P-LK1: -ísh in a verb with a separable preverb plus da- or an object prefix; a reflexive verb with a class-Ib preverb.

Sources for this follow-up

SOURCED Opened: Hale 2003 ms. (box copy), pp. 14–21 and 29–34 in the text layer, and pp. 14, 69–71 as page images (n. 12, 13, 14). Hale 2001, “Navajo Verb Stem Position and the Bipartite Structure of the Navajo Conjunct Sector”, Linguistic Inquiry 32(4), pp. 678, 689–692 (p. 689 as a page image). Willie 1991, printed pp. 78–80 (text layer, forms checked against the index entries NX-261–267). Willie & Jelinek 2000, pp. 275–277 (OCR of the scanned collection; their Navajo forms are not quoted). Aissen 2000, “YI and BI: Proximate and Obviative in Navajo” (author’s RTF; cited by section, example and table, because printed pages are not available), §§1, 3, 4, (6), (11), (12), (32), (35), Table 1. Speas 1986 dissertation, p. 275, §4.5.2 (page image): she adopts “the analysis of bi-verb forms proposed by Hale, Jeanne and Platero (1977) and Perkins (1978), in which such forms involve topicalization”. Through the site’s own pages: one proximate (R4/OP-7, OP-5), possessor command (Platero 1978; Saxon 2003), possessor binding, where -ísh goes. Not opened (cited only through the sources above, or not freely available here): Young & Morgan 1987 (pp. 187, 215 via Hale), Rice 2000 (via Hale), Kari, Faltz 1998 (p. 173 via Aissen), McDonough 2000, Speas 1990, Hale, Jeanne & Platero 1977, Perkins 1978 (the box file under that name is an unrelated thesis), Perkins & Fernald 2014 (not re-opened for this section).

4. Every phenomenon on the structure

Each row leads with the meaning: the LF (scope = c-command), then what the structural choice means for interpretation (scope, binding, theta roles, event structure; HYPOTHESIS, drawn from the row’s own analysis, with no new data), then where the phenomenon sits on §2–3, what moves beyond the default (V→v→T and subject to Spec,TP), its status, where it is argued, and what still does not fit. Word order enters only through “where” and “what moves”. “Review n” links are to the dated sections listed on Results on the spine.

PhenomenonLF (scope = c-command)What it means for interpretationWhere on the structureWhat else movesStatusSources · NXWhat does not fit yet
Basic transitive clause∃e[kick(e) ∧ Agent(e,horse) ∧ Patient(e,mule)]Theta: Agent from Spec,vP, Patient from VP. Scope: the subject in Spec,TP c-commands the object. HS-1 against HS-2 is linearization only: the same LF either way.Patient in VP, Agent in Spec,vP, verb in TNothing (HS-1); remnant VP above T (HS-2)HYPOTHESISresults §1 · NX-029, NX-001HS-1 vs HS-2 OPEN
One noun, two participantsas above, role by linkingTheta by linking, not by position: the one noun gets the lower role. A default, not a condition on LF. Two-tier linking (§3b): the noun may sit in either third-person shell; nothing in LK chooses, so a default is what LK predicts. Fits.The noun is linked to the lowest third-person participant unless its meaning rules that out (H4″)NothingHYPOTHESIS default onlyresults §1; review 14 · NX-027, NX-444H4″ is a default, not grammar (NX-444 “in context”)
Inverse yi-/bi-Same event LF as the direct; the patient takes surface scope over the agentTheta roles unchanged; c-command changes. The moved patient c-commands the agent, so it scopes over it. Here order is the visible result of a scope-changing movement. Two-tier linking (§3b): the patient’s referent moves on the phrasal tier; its pronoun stays and is spelled bi- by where the referent landed (LK-4). Needs extra: LK-4 restates H3 + H9′.Patient moves via outer Spec,vP to above the raised agent (H3, H21b); yi- needs a distinct third-person co-argument above it (H9′)The patient (definite or “which” only, H6′)HYPOTHESIS Move; OPEN landing site, triggerresults §2; review 14 (H53) · NX-030, NX-053, NX-085/*086, NX-419–421Trigger only described by the animacy ranking (H53, a stipulation); *NX-278 not understood
Fourth person ji-/ho- (object)WISH_speaker(∃e[sleep(e) ∧ Agent(e,z₄)]), z₄ = speaker, for NX-415 (review 14)Binding: the optative fourth is bound by the wisher (z₄ = speaker), from the wish operator above TP into vP. Two-tier linking (§3b): a fourth-person shell holds only a silent same-person referent, so a noun is linked to the other participant (NX-468, NX-469). Fits.A person value on an argument in Spec,vP or VP; the optative fourth is bound by the wisher (H51)NothingHYPOTHESISreview 14 · NX-415–418H44 (respected sayer forces shift) NOT UNDERSTOOD
Respect, report and story ji-as a third personNo new participant and no scope change: respect or report is an agreement feature on a third-person argument (on H57a). Two-tier linking (§3b): LK lets a noun into a ji- shell only on H57a, so LK forces H57a (P-LK4).ji- as an Agree mark of respect/report on a third-person argument (H57a)NothingOPEN H57a favoured over H57breview 15; Japanese comparison (H57a-J, honorific agreement, favours H57a) · NX-470, NX-473, NX-476Contradicts H31 as an exceptionless rule
Areal / pleonastic ho-∃s[deep(s) ∧ Theme(s,a)] (review 15, (1c))Theta: H54a puts a referential area argument in the event; H54b makes ho- role-less. The choice decides what the event description contains.A referential area argument (H54a); the role-less ho- of ‘know’ in Spec,TP (H54b vs H54′)NothingHYPOTHESISreview 15 · NX-446–450*NX-453 NOT UNDERSTOOD (H55 a stipulation)
Indefinite ʼa-, reflexive ʼádi-, reciprocal ʼahi-∃e[see(e) ∧ Agent(e,x₁) ∧ Theme(e,r) ∧ r = x₁]Binding inside vP: the Agent in Spec,vP c-commands and binds the anaphor in VP before anything moves (r = x₁).ʼa- and the anaphors in VP; anaphor bound by Spec,vP inside vP; all personless (H60)NothingHYPOTHESISreview 16 · NX-490, NX-492–499Classifier with anaphors but not with ʼa- (H59) NOT UNDERSTOOD
Agentive passive ʼadi-∃e[shave(e) ∧ Agent(e,u) ∧ Patient(e,y₂)]Theta: the Agent role is still assigned, to a silent existentially closed u. Nothing is demoted or promoted, so the patient stays below the subject position. Two-tier linking (§3b): a personless D⁰ can hold no referent, so the agent can never be named; patient pronoun outer (phrasal tier), agent inner (head tier) derives H58. With no referent, u’s D⁰ reaches T⁰ by head movement; whether Spec,TP is then empty is OPEN.Silent indefinite human agent u in Spec,vP (H10b); the patient keeps its object prefixu raises to Spec,TP by the default; no object promotionHYPOTHESISreview 16 · NX-020, NX-021, NX-499, NX-500Simple passive: H10a vs H10b OPEN
Passive questions and relatives?λl.∃x∃e[sell(e) ∧ Agent(e,x) ∧ Theme(e,sheep) ∧ Loc(e,l)]The silent agent is closed off inside the question; the wh is a variable bound from C⁰[Q].Silent indefinite agent in Spec,vP; wh in situNothingHYPOTHESISresults §12 · NX-213, NX-215, NX-315Whether the theme can raise instead OPEN
Wh questionsASK_WH(λx ∃e[kiss(e) ∧ Agent(e,John) ∧ Patient(e,x)])Question scope comes from C⁰[Q], not from where the wh word is pronounced: in situ, scope over the whole clause.Wh word in its theta position = indefinite + Q (H11); C⁰[Q] linked by AgreeNothing; fronting (H14) = one Move to Spec,CPHYPOTHESISresults §3 · NX-067, NX-069, NX-084H12-move vs H12-pf; H13-int vs H13-host
Polar questions -ísh/=sh, daʼASK_POL(p)Q in C⁰ scopes over the whole proposition, negation included. The host of -ísh marks focus, not scope (H12).Q in C⁰ above NegP; -ísh pronounced on a hostHost placement only (H12)HYPOTHESIS; OPEN daʼresults §2, §6 · NX-053, NX-119–134, NX-308daʼ: head or Spec,CP (the HS pivot)
Sentential negation doo … da¬∃x[dollar(x) ∧ MIN(x) ∧ ∃s[have(s) ∧ Holder(s,speaker) ∧ Theme(s,x)]]Scope = c-command: da in Neg⁰ fixes which clause is negated (NX-180 against NX-181); a minimizer left of doo is still under negation.Neg⁰ da final above TP; doo adjoined inside TP (H23); -da on an h-word = Agree (H17)A minimizer moves left of doo inside TP; -í- argument into Neg’s domain (H18a)HYPOTHESISresults §4 · NX-180/181, NX-205, NX-218, NX-228/229, NX-287/288NX-292 ʼádin (reconstruct or no raising, H26); site of doo
Constituent negation haniiassertion ¬[ιP.P(speaker) = sleep]Constituent negation: it scopes only over its focus XP; the rest of the clause is asserted.Right after an XP, which is its focus (H20); not on the spineNothingHYPOTHESISresults §5 · NX-242, *NX-246, *NX-285, *NX-298Two follow-up predictions of *NX-246 need a source
Prohibitive lágo, wish laanaaDIR_hearer(¬∃e[sleep(e) ∧ Agent(e,z₄)]) for NX-416 (review 14)Mood heads above the optative TP: lágo puts negation over the whole wished-for proposition.Final heads above an optative TP (H52); lágo is sentential negation without dooNothingHYPOTHESIS; OPEN H52a/b/creview 14 · NX-415, NX-416Exact category of the two particles
Questions with negation?r. REASON(r, ¬∃e[go(e) ∧ Mover(e,you) ∧ Goal(e,hospital)])Q above Neg: a reason wh above NegP asks for the reason for a negated event.C⁰[Q] over Neg⁰; reason/manner wh above NegP (H15)Reason wh Merged or moved above NegPHYPOTHESISresults §6 · NX-138, NX-184, NX-210, NX-247Question force of NX-247 OPEN
Distributive da-…Agent(e,X), |X| ≥ 3Distributivity over the closest c-commanded argument. Which argument is counted follows from height: the inverse patient is counted from outer Spec,vP.Dist⁰ just above vP agreeing with the closest argument (H21a); inverse patient counted from outer Spec,vP (H21b)Nothing extra (verb passes through Dist⁰)HYPOTHESISresults §7; review 16 · NX-272/*273, NX-274, *NX-277/278, NX-488Exemptions not understood; word-internal position misfits P1 (§3.2)
Quantifier scope∀x[girl(x) → ∃y[boy(y) ∧ …]] and ∃y[boy(y) ∧ ∀x[…]]Surface scope = c-command (subject over object). The inverse reading needs covert Move (H22a) or a choice function (H22b). The open question is about LF, not order. Two-tier linking (§3b): prefixes link to quantified and indefinite nouns, so the prefix is a φ-variable; LK predicts that a moved noun reconstructs (P-LK2), against the pronominal-argument view.Subject in Spec,TP c-commands object in VPInverse scope: covert Move (H22a) or choice function (H22b)OPEN H22a vs H22bresults §7 · NX-267—
Relative clauses =yę́ęιx ∃e[kick(e) ∧ Agent(e,he) ∧ Patient(e,x)]The D⁰ enclitic takes the CP and is linked to one participant inside it by Agree: the role of “the one” is read inside the clause, and nothing moves out. Two-tier linking (§3b): the enclitic’s link reaches into CP (Agree + λ-abstraction), so it is an extra on LK, not its spec–head link; H31 for relatives follows from LK-1.D⁰ over a full CP, Agree with one participant (H27b, H28′, H31)Nothing moves outHYPOTHESISresults §8 · NX-312–315, NX-319, NX-331*NX-345 vs NX-346 NOT UNDERSTOOD
Fact clauses =ígíí, =hágííthe fact pAn unlinked D⁰ denotes the fact itself: no participant is abstracted.Unlinked D⁰ over CP (H34)NothingHYPOTHESISresults §8 · NX-372, NX-385—
=go clausesWHEN(p, q), the =go CP c-commanding the main TPThe =go CP c-commands the main TP: WHEN(p, q), with the two clauses sharing participants (H45).CP with =go in C⁰, adjoined left at TP or higher (H47); participant pairing (H45)Nothing (forced by NX-397)HYPOTHESISresults §11 · NX-397, NX-399, *NX-403, NX-407/*408, *NX-409/NX-410, NX-411/412Core of H45 NOT UNDERSTOOD; NX-401/402 source conflict
Reported speech, “you think”p = …; ∃e1[say(e1) ∧ Agent(e1,x₃) ∧ Theme(e1,p)]H40a and H40b give different LFs for the reported clause: a linked silent Theme against the complement of ‘say’. The fourth person here concerns reference, not attachment (review 12).Two CPs: H40a (the say/think CP adjoined, silent Theme) vs H40b (complement of V)Nothing, or one leftward Move under H40aOPEN H40a slightly aheadresults §9–10; review 14 · NX-378, NX-384–394, NX-417/418H44 NOT UNDERSTOOD; obligation reading of (28)
PostpositionsPP roles inside the event (schematic)Theta: a grammatical P relates its object through its own agreement; a lexical P contributes a place. Order inside the PP is linearization only. Two-tier linking (§3b): a grammatical P takes an argument shell (NX-505), a lexical P a place DP with no shell (NX-503); *NX-506 still needs H64’s selection.Selected PP in VP; grammatical P agrees with Spec,PP, lexical P has no probe (H64)Nothing; “cannot be interrupted” follows from the default + HS-1 + left specifiersHYPOTHESISreview 16 · NX-314; Willie (31)–(36)Lexical P stacking NOT UNDERSTOOD; inclusory ‘with’ (H65) OPEN
Possessors—The possessor in the left specifier c-commands the possessum. Light n would introduce the possessor as an argument, as v does the agent (OPEN). Two-tier linking (§3b): the possessor is a shell in the left specifier; its prefix looks like an object prefix because both are phrasal-tier pronouns (NX-380–383, NX-508). This removes one argument against light n, and gives none for it.Left specifier of the noun phrase; prefix = agreementNothing (possessor raising not yet restated)HYPOTHESIS; OPEN light nreview 16 · Willie (36)Possessor raising (ch. 7), possessor yi- (Horseherder 1998) await restatement
Causatives and classifiers—Event structure: a causer adds an external argument in Spec,vP. The classifier spells a value of v⁰ tied to whether the doer is distinct (H59α′).Classifier = a value of v⁰ (H59); a causer is an external argument in Spec,vPNot yet stated on the spineOPEN not restatedH59α′ (Japanese comparison, 2026-11-15) HYPOTHESIS; older pages causative-split, causee-bi, voice-allomorphyWhich intransitives take ł-causatives (T4); H59 NOT UNDERSTOOD
Verb word: prefix orderprefix order = height order under P1, so a scope orderIf prefix heads are ordered by height (P1), prefix order is scope order, with the leftmost prefix widest. Rice’s generalization is the reverse. So the word-level headedness choice has interpretive content, and scope data can test it. On the two-tier model (§3.3), argument prefixes are exempt: roles come from links, not from prefix order. Two-tier linking investigated (§3b): with the subject pronoun on the head tier and the object pronoun on the phrasal tier (LK-p, Hale’s nuclear/nonnuclear planes), object-left-of-subject is derived; with both on the head tier it is not. The split preverb stays NOT UNDERSTOOD; da- fits only if phrasal.Word-internal Merge sites linearized leftward (§3.1); two-tier linking (§3.3) or sublexical V-to-T (§3.4)Prefix tier head-to-head, noun tier spec-to-spec (§3.3); or the stem (§3.4)GRANT'S HYPOTHESIS tested; OPEN fit§3; headedness §2.2–2.3; review 16 · NX-478–483, NX-493, NX-497Aspect, object/subject, distributive orders; split preverb (P1 alone)
Telicity—Event structure (telic against activity) is OPEN; the =go, laanaa/lágo and stem contrasts tested so far are not telicity-driven (SOURCED, narrow).——OPENresults §13; reviews 13–16No minimal pair; two narrow claims contradicted

5. Notes by phenomenon

Inverse. dzaanééz łį́į́ʼ biztał (NX-030; Willie 1991:71 (37b)), translated “The mule was kicked by the horse.” has the same Merge steps and theta roles as NX-029; the patient then moves above the raised agent, and the verb is spelled with bi-. The inverse changes which phrase is highest, not who does what to whom; its English passive translation is not evidence of Navajo passive morphology. Order of operations: the H3 Move precedes placement of =sh: Díí=sh háí yiztał (NX-085; Willie 1991:211 (16)), translated “This person kicked who?”, against *NX-086.

Passive and the reflexive. ʼádílzhééh (NX-499; Willie 1991:37 (20)), translated “You are shaving yourself.” and niʼdilzhééh (NX-020; Willie 1991:37 (2.21)), translated “You are being shaved.” use one prefix shape in two places: outer when it stands for the Patient (VP), inner when it stands for the Agent (Spec,vP). Where a prefix goes depends on the structural position of the participant it stands for (H58), which is P3 of §3.1 restated for agreement.

Negation. Béeso łáʼí ndi doo naashʼáa da. (NX-287; Perkins & Fernald 2014:104 (20 (§5.3))), translated “I haven’t so much as one dollar.”: the holder raises to Spec,TP, doo is adjoined inside TP, the minimizer moves left of doo but stays under da. Sentential negation (Neg⁰) and constituent negation (hanii on an XP: ʼAshxosh hanii; shinááʼ yeeʼ hanályįįh. (NX-242; Perkins & Fernald 2014:93 (4 A (§3.2))), translated “I am not sleeping; I was just resting my eyes.”) differ in what they license, not only in meaning.

Wh in situ. John háísh yiztsʼǫs (NX-067; Willie 1991:204 (7b)), translated “John kissed who?”: the question word is merged where a noun with its role would be, and nothing else has to move.

=go above the main subject. yah ʼííyáa=go ʼatʼééd dah neezdá (NX-397; Willie 1991:134 (4)), translated “When he came in, the girl sat down.”: the girl is the main-clause subject in Spec,TP, and the =go clause is pronounced before her, so it is adjoined at the TP level or higher.

Optative and prohibitive. ʼajoołháásh laanaa (NX-415; Willie 1991:120 (17a)), translated “That certain person wants to sleep.” / “(I want to sleep.)” against ʼajoołháásh lágo (NX-416; Willie 1991:120 (17b)), translated “Don’t let him go to sleep.”: same verb word; the wish particle binds the fourth person to the speaker, the prohibitive does not. The difference is in the final head, not in telicity.

6. What is open

7. Older pages restated

On 2026-11-15 every older working-note page, essay and topic section listed below received a short dated note at its top, “Restated on the unified structure (2026-11-15)”, saying which part of this page it now sits under and what its status is on the spine. Nothing else on those pages changed: no Navajo form, gloss, judgment, translation or citation. Their own analyses stay as the record of what was argued then, and as labelled contrast where they used flat S, position-class templates, rule sets A/B, a no-raising verb, or light d.

OPEN Not yet done: these notes place each page; they do not re-derive each page’s own examples on the spine. That per-page re-derivation (possessor raising and possessor yi-, causatives, the 3i pages, the deictic-d pages, Condition C) continues in the scheduled reviews.