Do Navajo verbs take theta roles, not words, as arguments? The case for and against
2026-11-11 audit pass 2. Apostrophe fixes only. Navajo glottal stops typed as ' or ’ (for example in prefix names like ʼa- and ʼeʼe-, or words like Awééʼ) now print as ʼ (U+02BC), and a build step no longer turns the closing quote of an English translation inside a Navajo row into a glottal stop (‘…the girls.’, not ‘…the girls.ʼ). No form, gloss or translation wording changed.
Verbs on this page 15 entries in the verb dictionary (show)
yiztał ‘kick’ · yiyiisxį́ ‘kill (one)’ · dabiiłtsą́ ‘see’ · nishhozh ‘tickle’ · nayiisniiʼ ‘buy’ · yoolóós ‘lead (one animate)’ · yíchxǫʼ ‘be ruined; (classifier) ruin’ · yíkʼą́ ‘grind (e.g. corn)’ · yiyííbį́į́ʼ ‘pick (round objects, e.g. berries)’ · yiyííłtaʼ ‘count (plural objects)’ · yóldon ‘shoot at’ · yoodlą́ą́ʼ ‘drink (it)’ · yikʼichʼínímááz ‘roll upon (it)’ · yikʼiikę́ę́z ‘fall upon (it)’ · nímááz ‘rolled; with yaa/baa ‘to’: rolled to’
Restated on the unified structure (2026-11-15). The framework is restated in §1: VPISH (arguments merged in the verb’s scope, subject in Spec,vP) with theta roles, and V→v→T with the subject raised to Spec,TP as the working default in tensed clauses; each pattern is what else moves, with an LF whose scope matches c-command. Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. Flat-S and other pre-VPISH notation below is labelled contrast. Derivations below in which the verb or subject does not raise are labelled contrast; the working default raises V→v→T and the subject to Spec,TP. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §1. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.
2026-11-11 audit pass 3. Verdict words on this page (SUPPORTS, AGAINST, CONTESTED, Neutral) are our tallies against the variants of the dropped thesis, not status labels: read SUPPORTS and AGAINST as HYPOTHESIS (our assessment) and CONTESTED as OPEN. See the pass 3 report.
Status badges now use only the label set SOURCED, HYPOTHESIS, SPECULATIVE, OPEN, NONE and GRANT'S CONJECTURE; verdict or task words that had been styled as badges (for example Opened, PRE-VPISH, DONE, SUPPORTS, PARTLY ANSWERED) are now plain text or a label with a qualifier.
2026-11-11 audit pass 4. Interlinear rows whose form and gloss do not line up one-to-one were re-checked against the source page. Each is marked † with a note under its table: SOURCED where the source prints it that way, OPEN where the source cannot confirm it. See the pass 4 report.
Note (2026-11-02): Grant has dropped the “roles, not words” thesis. Pages that test it are kept as a historical record. Current work: Light determiners.
This is the site's flagship page for Grant's central claim. It states the claim in several precise versions, then runs every diagnostic we could source, one at a time, with a Navajo example and a verdict for each. The aim is to say plainly where the claim holds and where it fails. Chemistry is deliberately left out until the linguistic claim is settled. Written 2026-10-14 (PT).
Bottom line, stated first and not softened. The strong version, that nouns are never arguments of anything in Navajo and that roles alone decide which noun goes with which slot, fails. Three things sink it. Bare nouns are true arguments of postpositions (Hale 2003 ms.:51). Word position, not role entailments, decides some links (Willie 1991, *(33c), (25)). And the slots are referential pronouns with person and number, not bare theta roles (Willie's ADJ-PHI, p. 53). A weaker version survives, and it is a serious, published position (Jelinek 1984; Willie 1991; Hale 2003). On that version, the verb's arguments are role-assigned pronominal slots inside the verb word (or inside the verb-plus-postposition unit), and clause-level nouns are adjuncts linked to those slots. It is not uncontested, though. A configurational analysis (Speas 1986/1990) and a binding study of possessor yi- argue that overt nouns sit in argument positions, and the decisive tests (Condition C, extraction, quantifier scope) are still undecided in the literature.
On this page: 1 The claim, in versions · 2 Where the claim comes from · 3 Diagnostics, one by one · D12 Rank reads the event · 3b Animacy as an extension of Dowty · 4 The case against, steelmanned · 5 VPISH-aware derivations · 6 Verdict by version · 7 What would settle it
1. The claim, in versions
SOURCED Baseline framework. This site takes Dowty's (1991) proto-role theory as its baseline. It is a mainstream cross-linguistic account of how participant roles map to syntax, not a rival to other theories. Arguments carry clusters of Proto-Agent entailments (volition, sentience, causation, movement, independent existence) and Proto-Patient entailments, and the Argument Selection Principle maps the argument with the most Proto-Agent entailments to subject (1991:572, 576). Other standard accounts make the same basic cut in different terms. Levin & Rappaport Hovav (Argument Realization, 2005) state realization over event-structure prominence. Role and Reference Grammar sorts arguments into two macroroles, Actor and Undergoer (Van Valin & LaPolla 1997; Van Valin 2005). (The last two are cited, not opened, on this site.) The Navajo questions on this page are about how Navajo extends that baseline, not whether it applies.
Notation as on the proto-role schema page. Σ(√) is the verb's set of role slots {ρA, ρP, …}, each with Dowty (1991) proto-role entailments. π(ρ) is a slot's person and number value. N is an overt noun phrase in the clause.
T-STRONG No N is ever an argument of any head. Every argument position of
every predicate (V or P) is a role slot ρ; N only ever links to a ρ.
Which ρ an N links to is decided by ρ's role content alone.
T-MID The arguments of V (and of grammatical P) are the pronominal slots
of Σ, valued word-internally. N is never an argument of V; it is an
adjunct linked to exactly one ρ. Linking may use position, voice,
person and number as well as role content.
T-WEAK The pronominal prefixes are true arguments (not agreement), and
overt N are optional and adjoined. Silent on whether the slots are
"roles" or just pronouns, and on how linking works.
T-ROLE (independent of the three above) Some grammatical generalizations
are stated over role slots (ρA, ρP) rather than over grammatical
relations, case or position.
| Version | Status | Basis |
|---|---|---|
| T-STRONG | GRANT'S CONJECTURE in its strongest reading | Grant's project chat (Grok project “Navajo Verb Structure”), message 1: “their arguments aren't set within the domain of the VP, but rather only a schema of the theta roles … prior to realization as a lexical item.” Verdict: fails (§6). |
| T-MID | HYPOTHESIS, built on SOURCED statements | Willie 1991:51–52, 69–70; Hale 2003 ms.:5–7, 51; this site's ADJ-NOTARG and ADJ-* rules. Verdict: the best-supported version; contested. |
| T-WEAK | SOURCED as a published analysis (the Pronominal Argument Hypothesis) | Jelinek 1984; Willie 1991; Willie & Jelinek 2000; Hale 2003. Verdict: mainstream for Navajo, but disputed (Speas 1986/1990; possessor-yi- study). |
| T-ROLE | HYPOTHESIS | Best case da- (Willie 1991:82–84), which depends on how the inverse is analysed (T7). Verdict: plausible, not proved. |
HYPOTHESIS A distinction Grant's wording blurs. The published Pronominal Argument Hypothesis says the arguments are pronouns: Hale's (16) says “Person-Number morphology in the Navajo verb word has the status of full argument” (Hale, The Phonological Aspect of the Navajo Conjunct Verb System, handout p. 7). A pronoun is a referential element that bears a theta role. It is not the role itself. So the literature supports “the verb's arguments are role-bearing pronominal slots, not lexical nouns”. It does not support “the arguments are theta roles” in any literal sense. Willie links nouns to slots partly by person and number (ADJ-PHI, p. 53), which means a slot has referential features beyond its role. Grant's “schema of theta roles prior to realization as a lexical item” is best read as T-MID. Read literally, it says more than any source does.
2. Where the claim comes from
GRANT'S CONJECTURE Grant's Grok project “Navajo Verb Structure” has one document, Navajo Verb Structure — Exploratory Notes (Revised Working Draft, August 2026), and one chat, “Navajo Verbs: Proto-Roles in Morphology”. Both were mined for this page. The ideas they contribute are listed below. None of them is treated as sourced unless a published work below confirms it.
- Schema before lexical items (chat, message 1; Notes §8): the verb carries “a schema of formal features present on the Root prior to the full lexical realization of argument DPs”; “their features are valued by the prefixal morphology, and the full DPs may surface only as null pronouns, clitics, or externalized phrases ordered by topicality and hierarchy” (Notes p. 4). Partly confirmed: Willie 1991:51–52 and Hale 2003 ms.:5–7 say the prefixes are the arguments and the nouns are adjuncts. Not confirmed: that the prefixes are a schema of roles rather than pronouns (§1).
- Classificatory stems lexicalize Proto-Patient properties (Notes §3, §5.2): shape, rigidity, animacy and number of the Theme are built into the stem. Confirmed for linking: Willie's “round objects” and “plural objects” glosses block a noun from the patient slot ((22), (24), pp. 61–62; see D9). That is the one place where slot content, not the noun, decides a link.
- Local linearization: German top, Navajo bottom (chat, messages 2–3; Notes §6): HYPOTHESIS, handled on the headedness page and not tested here.
- Open question 6 (Notes §9): “What reconstruction, binding, or scope diagnostics are available inside the verbal complex itself …?” This page's diagnostics D6–D8 are a first answer.
3. Diagnostics, one by one
Each diagnostic asks: if Navajo nouns are adjuncts linked to argument slots, what should we see, and what do we see? Navajo forms are copied from the cited pages. Willie's glosses give object before subject (“3s.1s.ruined” = it, by me). We add no Navajo forms of our own.
D1. The verb alone is a complete clause, with pronominal marking always present
| (11) | (ʼashkii) | yicha |
| boy | 3s.crying | |
| ‘(The boy) he is crying.’ | ||
| (12) | yíchxǫʼ | |
| 3s.ruined | ||
| ‘It is ruined.’ | ||
SOURCED “The MVS is a complete functional complex, and by itself is a grammatical sentence. Nominal expressions are always added to the MVS for reasons relating to the structure of the discourse, not for reasons of grammaticality” (Willie 1991:51–52). Hale's (1) ni-sh-hozh ‘I tickle you’ has “no ‘small pro’ elements” (Hale 2003 ms.:5).
Verdict: SUPPORTS T-WEAK and T-MID. Counter-reading: a pro-drop language with rich agreement (Italian, or Navajo under Speas's analysis) shows the same thing, so D1 alone does not tell the two analyses apart.
D2. Nouns are optional and add no argument
SOURCED A noun “by itself cannot be an argument -- it must be licensed by a pronoun” (Willie 1991:52). (11) with or without ʼashkii; (15) (naadą́ą́) yíkʼą́ ‘(The corn) I ground it’ (p. 59). In possessor raising, the body-part noun inside the verb “is not an argument … but serves to derive a complex verb” (p. 187), so a lexical item can enter the verb without taking a slot.
Verdict: SUPPORTS T-MID. Optionality on its own is again compatible with pro-drop (CONTESTED as a discriminating test).
D3. A noun must match its slot in person and number; one free pronoun only
| (8) | shí | bi-chidí | yílchxǫʼ |
| I | 3.car† | 3s.1s.ruined | |
| ‘I (am the one who) ruined his car.’ | |||
| (9) | shí | shiiniltsą́ | |
| I | 1s.2s.saw | ||
| ‘I (am the one) you saw.’ | |||
| (10) | *ní shí shiiniltsą́ | ||
SOURCED † Segmentation note (2026-11-11 audit pass 4): Willie 1991 (8) prints bi-chidí with the gloss “3.car”: the gloss joins with periods or colons pieces the form separates with hyphens, so the two lines do not match one-to-one in the source. Kept exactly as printed.
SOURCED Free pronouns mark “referential contrast or emphasis”; “Constructions with two free pronouns are judged unacceptable” (pp. 53, 56). Hale gives the same pattern (Mary Willie's observation, handout (18): *Shi ni ni-sh-hozh vs Ni ni-sh-hozh ‘I tickle YOU’) and calls the free pronoun “a contrastive adjunct”, like English “As for you, I'm tickling you” (Hale 2003 ms.:5–6).
Verdict: SUPPORTS T-WEAK and T-MID. In an agreement language, both subject and object pronouns should be able to appear together in argument positions (English “I saw you”). The ban on two follows if free pronouns are dislocated topics or foci. AGAINST T-STRONG's “role alone”: shí links by person value (to the subject in (8), to the object in (9)), not by role.
D4. Free word order
SOURCED Navajo order is not free. With two nouns, the inner one goes with the internal argument and the outer one with the external argument, “by default” (Willie 1991:60, (20)). One noun with a transitive verb goes with the object (p. 59, (17)). Swapping the nouns changes the meaning, or forces the inverse bi- (Willie 1991:71; inverse page).
| (33) | ʼashkii | yá | yoolóós |
| boy | 3.for | 3.3.lead [sing. +animate] | |
| a. ‘He is leading the boy for her.’ b. ‘She is leading it for the boy.’ c. *‘The boy is leading it for her.’ | |||
Verdict: AGAINST the classic Jelinek/Hale “free order” prediction, and AGAINST T-STRONG: ρA of ‘lead’ fits a boy perfectly, yet (33c) is out. Position decides. Defence (steelman of PAH): Willie & Jelinek (2000) call Navajo “discourse configurational”: the order of adjuncts is fixed by discourse and animacy rules, not by argument structure. That is coherent, but it means order cannot be cited as evidence for the claim. At best it is neutral.
D5. The number of slots limits how the nouns are grouped
SOURCED The same noun string hastiin sání bi-chidí ‘man old his-car’ has to be one phrase with intransitive (14) yíchxǫʼ (‘The old man's car is ruined’) and can be one or two phrases with transitive yiyííłchxǫʼ ((19) ‘The old man ruined his car’; (21) ‘He/she ruined the old man's car’) (Willie 1991:58–61).
Verdict: SUPPORTS T-MID (SCHEMA-PARSE, HYPOTHESIS on the nominal adjunction page). The slots are fixed first, and the nouns are fitted to them. Counter-reading: the same holds in any language where the verb's valence limits its NP arguments, so this weakly favours, rather than proves, the adjunct analysis.
D6. Binding and Condition C
SOURCED Hale credits Ellavina Tsosie Perkins (1976) with “the Condition C Problem” (handout (17)): ‘My father will butcher the deer he killed (earlier) today’, where shizhéʼé ‘my father’ sits inside the relative clause and corefers with the matrix subject. Hale's structures: (b) [ec]j [today my fatherj deer it-he-killed]-RL it-he-will-butcher vs (c) my father, deer, [today it-he-killed]-RL it-he-will-butcher. (The Navajo line in our copy of the handout is corrupted by OCR, so it is not reproduced here.) If the matrix subject were an argument position c-commanding the noun, (b) would violate Condition C. If the nouns are adjuncts and the only arguments are prefixes, nothing c-commands a noun, so there is no violation (cf. Baker 1996).
SOURCED (secondhand, excerpt only) Hale & Platero (“Negative polarity expressions in Navajo”, MIT ms., hale014.pdf; earlier labelled “Hale & Perkins”, corrected 2026-11-13) say Speas (1990:232) explains the same facts in a configurational grammar with a Parallel Function Constraint. They leave the choice open: “It will be necessary to determine the correctness of one or another of the structural arrangements … we must look elsewhere”. They also write that, on one view, “Navajo is a standard configurational language (cf., Speas, 1990)”.
Verdict: CONTESTED. The facts fit the adjunct analysis, and a rival account exists. By Hale's own account the question is unresolved.
Update 2026-10-14 (8:46pm): Speas 1986 read in the original (page). SOURCED Platero's (4.15a) ‘Hei branded the horse that the mani roped’ is good, and Speas calls it a Condition C violation in her own structure, rescued only by an across-the-board representation (pp. 256, 266–267). (4.119)–(4.121) show the same exemption with no relative clause (pp. 319–320). Her one Condition C exclusion, (4.15d) ‘*Iti branded the man that roped the horsei’, has a horse acting on a man in a yi- verb, which the rank rule excludes anyway. HYPOTHESIS Revised verdict: LEANS SUPPORTS T-MID/FB. Speas's ATB rescue still exists, and a minimal pair with two participants of equal rank is needed to decide (T22b).
Update 2026-10-15 (12:46pm): Hale 1999 “Conflicting Truths” checked (page). SOURCED Hale's (7) gives the Perkins sentence in clean Navajo: [Jį́į́dą́ą́ʼ shi-zhéʼé łį́įʼ nayiisniiʼ-ę́ę] yídiʼdoołił ‘My fatheri will brand the horse hei bought (earlier) today’ (p. 171). His (10) Si-tsilí [tsʼídá shi-zhéʼé bi-łį́ʼ-ígíí] yeiníłtį́ ‘Hei gave my younger brother the very horse of my fatherj’ does not allow i = j (p. 172), and nor do the factive clauses in (11a,b). His conclusion: “(10) and (11) show that Condition C is obeyed in Navajo”, and (7) is possible only because Platero's parsing rule reads the nouns as shared by both verbs (p. 173). HYPOTHESIS Revised verdict (1:07pm, replacing the 12:46pm “CUTS AGAINST”): CONDITIONAL on assumptions A1–A10. Hale's two-people cases are in tension with nouns adjoined above the subject (the version of T-WEAK drawn in his tree (9), and FB-5's exemption) if Condition C is what excludes the readings, the judgments are stable, and an adjunct is out of the subject's reach. None of those is independently established for Navajo, and Speas's (4.119) has the opposite judgment on the same shape. D6 is compatible, undecided (page; derivations and audit). 8:46pm: Jelinek 1984 supports Hale's S-adjunct picture for Warlpiri; Baker's later footnote does not settle Navajo attachment (T25c). 2026-10-16 (4:46am): re-run site by site in a VPISH/TP/CP spine (Condition C by site). No attachment site predicts both Hale's and Speas's judgments. Hale's (10), (11a) are compatible with nouns as linked adjuncts (e.g. adjoined at vP with the subject bundle raising), so they bear on attachment height and subject movement, not on argument versus adjunct. D6 stays compatible, undecided.
Update 2026-10-21 (12:46pm): the same trip, three subjects (page). SOURCED In the NLA's travel sentences, the person who travels is in a ‘with’ phrase when the verb is ‘fly’, ‘float’, ‘crawl’ or ‘run/drive’ (Chidí naatʼaʼí nihił dah diit’a’ ‘airplane 1dpl-with up 3-fly.P’, ‘The plane took off with us’), the driver is the subject with the stem glossed ‘drive’ (Háí lá kintah góó sh ił ʼadooł bąs? ‘Who can drive me to town?’), and the traveller is the subject with -ee ‘by means of’ (Łééchąąłgaii b ee nááshdááł, ‘by Greyhound bus … 1-return’). In sh ił dah ʼadiilwod ‘so I just drove that way’ the subject is a third-person slot with no noun at all. HYPOTHESIS The subject is the participant the stem's own role-in-the-action entailments rank first (causation > movement > being carried; TP-4). This fits the weak version (role-bearing slots, nouns optional). It is not a test of the strong version.
Update 2026-10-21 (8:46pm): a role with no slot (page). SOURCED Dowty 1991 lists “incremental theme” among the Patient Proto-Role properties (p. 572) and, for motion verbs, puts it on the path, which “is … not a syntactically realized argument at all” (p. 569). HYPOTHESIS In the NLA's one mile sentence (Naakidi tsin sitą́ ądi … ʼahééniijééʼ ‘we ran about two miles …’) the verb word agrees only with the runners; the distance phrase has its own postposition and no slot. So the weak thesis is narrowed: the verb's slots are its participant roles; a dimension role like the path is entailed but only measured by a clause-level phrase (DM-2, DM-3). GRANT'S CONJECTURE The distance phrase is the incremental theme: supported for English (Hay, Kennedy & Levin 1999); the Navajo telicity test is designed but OPEN.
Update 2026-10-22 (4:46am): a noun between a postposition and the verb (page). SOURCED NLA (9)–(11b) print … bee dibé síníloh?. The object noun stands between bee ‘3-with’ (Willie's grammatical instrumental, p. 41 (28f)) and the verb, against Willie's “cannot be interrupted” (p. 50 (44a)). HYPOTHESIS Two derivations give the string. One (IP-1) puts the means PP on the clause side, so no noun is inside VP; that fits T-MID. The other (configurational, extrapolated from Speas's SOV) puts dibé in Comp,V, which goes against T-MID. Graded compatible, undecided, not contradicting. It would count against T-MID only if a noun turns up between a PP whose object is a participant of the verb and the verb itself (T49a). New rule IP-5: where a participant is pronounced depends on the stem's role schema, not on the noun or the postposition's form. The same -ee stands next to the verb for the possessor of ‘exist’ (NLA Q2) and on the clause side for the rope.
Update 2026-10-22 (12:46pm): which postpositions belong to the verb complex (postpositions by verb). SOURCED Willie 1991 prints voice pairs on ‘for’ with ‘lead’, ‘to’ with ‘roll’ and ‘to’ with ‘speak’, and says the inverse has scope over the MVS and its postpositional phrases (pp. 100, 208). On p. 66 she says the noun just before yá “is not coindexed with it, and cannot be governed by it”. HYPOTHESIS So the verb complex is the domain where role slots, the stem's own and those its postpositions add, are ranked against each other, and nouns are linked from outside it. That fits the weak thesis, and nothing opened contradicts it. It is not a separating test against a configurational account.
Update 2026-10-22 (8:46pm): does the voice reach the instrument? (Instrument and voice). SOURCED In NLA third-person sentences, ‘for’, ‘to’ and ‘with (someone)’ take the voice's y-form, and ‘with’ also takes the b-form. The instrument takes only the b-form bee (two sentences). GRANT'S CONJECTURE This fits the weak version: the voice compares role slots, and the slot of a means is outside the comparison. Nothing contradicts it, and nothing separates it from a configurational analysis.
D7. Extraction
SOURCED Baker's polysynthesis account predicts that if nouns are adjuncts, nothing can be extracted from them (CED). Willie 1991 Ch. 8 finds that question words behave like nouns. háísh ‘who’ is “coindexed with (licensed by) a third person pronominal” (p. 201). There are no long-distance questions ((21)–(32)), and a question word plus a noun form one adjunct with no extraction from it (WH-PHR; wh-questions page). SOURCED (abstract only) Speas (1986), in her dissertation abstract: “by standard syntactic tests, overt nominals do not behave like adjuncts, they behave like arguments”, and the prefixes are “infixes” that sit in argument positions syntactically.
Correction 2026-10-14 (8:46pm), from the dissertation itself: Speas raises the option that the prefixes are arguments that move into the verb at PF, and then declines it (1986:201). Her nouns are in θ-positions, and the prefixes are non-θ tails of a CHAIN (p. 202). Her tests are no weak crossover (2.95) and multiple questions (2.96c); neither involves extraction from a noun. Both are COMPATIBLE with FB (page).
Update 2026-10-15 (4:46am): SOURCED Hale & Platero 1974:13 report (from Platero 1973, data not reprinted) that relativization obeys Ross's island constraints but pronominalization does not. On their analysis relativization moves a determiner (p. 27). HYPOTHESIS This is COMPATIBLE with FB, because a D/φ element can be the thing that moves. It decides nothing until the island data are opened (T23d on the H&P 1974 page). The same paper adds two more cases against T-STRONG: in (32) and (46) the position of a noun overrides the reading the roles would favour.
Verdict: CONTESTED. Willie's data fit. Speas's tests, now opened, do not decide extraction either way. A further internal conflict: Willie 1991 (4a′) allows one reading of háísh John yiyiiłtsą́, while Willie & Jelinek 2000:273 (58b) allow two for the parallel string (T19a).
D8. Quantifiers and non-referential nouns
SOURCED Pronominal-argument languages are expected to lack true NP quantifiers and non-referential NP arguments, since an adjunct linked to a definite pronoun cannot be a bound variable (Baker's list includes “no true quantifiers”). Navajo has a universal quantifier on the external argument whose reading depends on voice (Willie 1991:79–80, (58)–(60)). Wh-words without =sh “can be interpreted as an indefinite in a non-interrogative sentence” (p. 210).
Verdict: CONTESTED, leaning AGAINST the strong form. Quantified and indefinite expressions do occur, and the scope facts for (58)–(59) are not yet stated by any source we opened (T5c). The adjunct analysis can survive only if the quantifier is an adverbial-type adjunct; that has to be shown, not assumed.
D9. Slot content overrules position
| (22) | ʼashkii | yiyííbį́į́ʼ |
| boy | 3.3.picked round objects | |
| ‘The boy picked them. (c.f. berries)’ | ||
| (24) | ʼashkii | yiyííłtaʼ |
| boy | 3.3.count plural objects | |
| ‘The boy counted them.’ | ||
SOURCED A lone noun should go with the object (ADJ-OBJ), but the stem requires round or plural objects, so the boy goes with the subject. Willie: “berries do not pick boys” (p. 61). But some overrides rest on world knowledge, not on the slot: “horses typically are led” (p. 66). And (25) ʼashkii yóldon ‘The boy is shooting at it’ has no visible slot property at all (p. 62).
Verdict: SUPPORTS T-MID (slot content is one factor in linking) and is the best support for Grant's classificatory-stem idea. AGAINST T-STRONG: position (33c) and world knowledge (32) also decide links.
D10. The inverse and da-: generalizations stated over roles
SOURCED da- (three or more) counts the agent in both voices: “the agent falls under the scope of the distributive marker”, including when the agent is not the subject under bi- (Willie 1991:82–84). The yi-/bi- choice is made inside the word, by animacy ranking, before any noun is placed (Willie 1991 §3.3; Sandoval & Jelinek 1989; inverse page).
Verdict: SUPPORTS T-ROLE, CONTESTED. If Willie's case analysis is right (the inverse agent is internal and ergative, p. 72), da- can only be stated over the role. If W&J 2000's constant gloss (agent = 3NOM in both voices) is right, da- just tracks a case, and T-ROLE loses its best example (number & scope §4).
D11. Are there any nominal arguments at all?
SOURCED Hale: “Navajo does permit bare nominal arguments of postpositions to appear in their base positions (i.e., as so-called incorporates)” (Hale 2003 ms.:51). Willie's lexical postpositions take nouns in a way grammatical ones do not (GPP/LPP, pleonastic & psych page). And Navajo “does not itself have noun incorporation” into the verb (Hale 2003 ms.:13), which is why Baker (1996) does not count Athabaskan as polysynthetic in his sense.
Verdict: AGAINST T-STRONG (some nouns are arguments: of P). Neutral for T-MID, which is restricted to the arguments of V and grammatical P. Baker 1996 cannot be cited as direct support for Navajo.
D12. The ranking reads the event, not the word SOURCED (added 2026-10-14, 4:46am)
| (11) | Tsé | tʼiis | yikʼichʼínímááz |
| (rock) | (tree) | (it-upon-horizontally-it-rolled) | |
| ‘the rock rolled upon the tree’ — acceptable; *(12) Tʼiis tsé bikʼichʼínímááz unacceptable | |||
| (13) | Tʼiis | tsé | yikʼiikę́ę́z |
| (tree) | (rock) | (it-upon-it-fell) | |
| ‘the tree fell upon the rock’ — acceptable; *(14) Tsé tʼiis bikʼiikę́ę́z unacceptable | |||
If the ranking were a fixed order over lexical items, one of tsé and tʼiis would have to come first, or both voices would be free. Neither happens. The participant the verb says moves is the subject, in either order of the two nouns, and the inverse is out for both. Witherspoon: “the rock … is superior to the tree because of its locomotion” (p. 69). Rank also ignores lexical class for animates: awééʼ ‘baby’ ranks with animals (pp. 79–80).
Verdict: SUPPORTS T-ROLE (the ranking is computed over participants' proto-agent properties, here an event entailment, at the schema step; derivations C1–C2). Counter-reading: Hale's 1973:303 table (inanimate–inanimate: blocked) predicts the same stars without any ranking, and Witherspoon calls all of this “cultural rules” (pp. 66–67). So D12 shows the ranking is not lexical. It does not show that it is a grammatical rule over roles. One author, few consultants: CONTESTED as a grammatical diagnostic.
Summary
| # | Diagnostic | T-STRONG | T-MID / T-WEAK |
|---|---|---|---|
| D1 | Verb alone is a clause | SUPPORTS | SUPPORTS (not discriminating) |
| D2 | Nouns optional | SUPPORTS | SUPPORTS (not discriminating) |
| D3 | One free pronoun; φ-matching | AGAINST (links by φ, not role) | SUPPORTS |
| D4 | Free order | AGAINST | AGAINST classic PAH prediction; neutral under “discourse configurational” |
| D5 | Slots bound parsing | SUPPORTS | SUPPORTS (weak) |
| D6 | Condition C | CONTESTED | |
| D7 | Extraction | CONTESTED | CONTESTED |
| D8 | Quantifiers / indefinites | AGAINST (leaning) | CONTESTED |
| D9 | Slot content overrules order | AGAINST (not the only factor) | SUPPORTS |
| D10 | da- follows the agent | CONTESTED | SUPPORTS T-ROLE, CONTESTED |
| D11 | Any nominal arguments? | AGAINST | neutral |
| D12 | Ranking reads the event, not the word | neutral (not about linking) | SUPPORTS T-ROLE (Creamer 1974 added 12:46pm); CONTESTED as grammar |
3b. The animacy ranking behind yi-/bi- as an extension of Dowty's proto-roles
SOURCED Baseline: Dowty's Argument Selection Principle picks the subject and object by counting Proto-Agent and Proto-Patient entailments (1991:576), and it applies to Navajo direct clauses as it does elsewhere (the NAT rule on the inverse page). Dowty's principle selects only between subject and object, though. It says nothing about a voice alternation that ranks two participants against each other by kind. So the Navajo inverse needs the baseline extended in two ways, set out below: (a) a weighting of the Proto-Agent properties, and (b) person and topicality/definiteness factors. GRANT'S CONJECTURE The animacy scale itself is proto-agent weight. If that is right, the inverse ranks roles rather than words, and that strengthens T-MID and T-ROLE. Willie, Sandoval & Jelinek, and Creamer state the scale over kinds of referents. Update (2026-10-14, 4:46am): Witherspoon 1977 is now opened (page images, pp. 63–82), and he does tie the scale to control, intelligence, will and movement; see the next paragraph and the new page Who can act upon whom?. The link to Dowty's terms is still ours, and Hale 1973 is still known only through quotations (now page-checked in Witherspoon).
| Rank (Willie's animacy scale; Creamer 1974's groups, simplified) | Proto-agent properties the referent can typically have (Dowty 1991:572) | Status |
|---|---|---|
| 1st / 2nd person | vol, sent, caus, mov, indep | Scale SOURCED (Willie 1991:86 and inverse page); mapping GRANT'S CONJECTURE |
| 3rd-person human | vol, sent, caus, mov, indep | as above. The ranking of 1/2 above 3rd humans is not predicted by these properties (see test 3) |
| Large animals, then smaller animals | sent, mov, caus; volition reduced | as above |
| Natural forces (lightning, wind, water) | caus, mov; no sentience or volition | as above (see test 1) |
| Inanimate things, abstractions | indep only | as above |
Witherspoon 1977, now opened. SOURCED (page images, printed pp. 63–81; full page). The ranking “deals with the phenomena of movement and action, referring particularly to who can control whom, or who can act upon whom” (p. 67). Water “does not possess intent or will and thus cannot control or act upon the girl” (p. 69). Who controls whom “is determined primarily by the level of one's intelligence and, to a lesser extent, by the potency of one's animation” (p. 77). Among inanimates, “movement and action are part of the considerations”: in (11) Tsé tʼiis yikʼichʼínímááz ‘the rock rolled upon the tree’ “the rock … is superior to the tree because of its locomotion” (p. 69). He reads bi- as ‘let’: (2) Łį́į́ʼ dzaanééz biztał ‘the horse let the mule kick it’ (p. 67), and *(8) Tó atʼééd boodlą́ą́ʼ ‘the water let (or caused) the girl drink it’ (p. 73). HYPOTHESIS Intelligence, will and intent map onto Dowty's volition and sentience; locomotion and animation onto movement and causation. So the source states both the proto-agent basis of the scale and its weighting. Two cautions: Witherspoon is an anthropologist giving an ethnosemantic account from a few consultants, and he says the patterns come from “cultural rules”, not “the deep structural level of Navajo grammar” (pp. 66–67).
The extensions, and what they must handle
- Extension (a): weighting. Natural forces. GRANT'S CONJECTURE Lightning and wind cause and move things, and causation and movement are proto-agent properties, so a high rank for forces follows from proto-role weight. That is consistent with Dowty and is not treated here as a problem. SOURCED One datum still has to be fitted: Willie finds (73a) ʼiiʼniʼ łį́į́ʼ yiyiisxį́ ‘The lightning killed the horse’ “objectionable, as an animacy violation, and (73b) [the bi- form] more preferable” (1991:90). So in her judgment the horse outranks the lightning. Under the mapping, that means sentience (which the horse has and lightning lacks) outweighs causation and movement (which lightning has in that event). HYPOTHESIS The mapping survives only if the proto-agent properties are weighted, with sentience and volition above causation and movement, rather than simply counted. Dowty's argument selection principle counts (1991:576). SOURCED Now stated by a source: control is set “primarily” by intelligence and “to a lesser extent” by animation (Witherspoon 1977:77), and Witherspoon's (9)/(10) repeat Willie's pattern with water: Atʼééd tó biisxį́ ‘the girl can be drowned (killed) by the water’ is fine, but “the water cannot drown (kill) the girl” with yiyiisxį́ (pp. 65–66). The weighting is therefore no longer a one-datum guess. Willie notes that speakers vary on (73), and the Dowty labels remain ours.
- Entailments vs capacities. HYPOTHESIS Dowty's properties are entailments of the predicate about a participant, not standing properties of a kind of thing (1991:572 n. 16). In ‘the horse kicked the man’ the verb entails the horse's causation and movement whatever horses are like. The Navajo scale ranks referents by kind, even when no nouns are present (Willie 1991:86–88). For the mapping to work, the ranking has to read a referent's potential proto-agent properties. That is a natural extension of Dowty, but it is an extension, and it should be flagged as such. SOURCED Witherspoon's data split this in two (§6.3 there). Among inanimates the deciding property is an event entailment: rock and tree swap places depending on which one the verb says is moving ((11) ‘the rock rolled upon the tree’, (13) ‘the tree fell upon the rock’, p. 66). Among animates it is a capacity, but not a word class: a baby ranks with animals, not with adult humans (Hastiin awééʼ biztał, pp. 79–80). Either way, the rank follows what the participant can do or does, not the noun that names it.
- Extension (b): person and definiteness. SOURCED “The subject of an Inverse construction is definite, and becomes the discourse topic” (Willie 1991:100), and *(47) bans an indefinite inverse subject (pp. 76–77). Proto-role weight says nothing about definiteness. 1st/2nd persons above 3rd-person humans also does not follow from vol, sent, caus, mov or indep, which humans share. The baseline does not cover these effects, as expected, since Dowty's principle only selects subject vs object. Extension (b) adds person and topicality/definiteness alongside proto-role weight.
- Extension (b), continued: equal-rank pairs, where topicality breaks ties. SOURCED “And if they are equal, the choice rests on discourse factors” (Willie 1991:86, (70)). Her horse/mule pair is equal-ranked, and both voices are fine (Willie 1991:71; inverse page). HYPOTHESIS (Grant's framing) When proto-role weight is tied, topicality decides. Willie's statement is close to a direct source for this, and her p. 100 remark (the inverse subject “becomes the discourse topic”) says which way the tie goes. SOURCED The equal-rank test, run on Witherspoon 1977: for animates it holds (horse/mule (1)/(2), p. 63; baby/horse both yi- orders, pp. 79–80). For inanimates it fails: rock and tree are equal (Creamer group 7), yet the inverse is out in both orders, *(12) Tʼiis tsé bikʼichʼínímááz and *(14) Tsé tʼiis bikʼiikę́ę́z (p. 66), and Hale's own earlier table has “inanimate–inanimate: blocked” (1973:303, quoted p. 69). Verdict, revised: SPLIT. SUPPORTS for animates; CUTS AGAINST for inanimates. The repair is HYPOTHESIS INV-CAP: the inverse subject must have potential sentience or volition, a capacity to have resisted (Witherspoon p. 74). That repair is itself stated over a proto-agent property, so it does not hurt Grant's mapping, but the old rule “equal rank → discourse decides” was wrong as stated (details). Revised again, 12:46pm, after reading Creamer 1974 in the original (Eight ranks): verdict now SUPPORTS. INV-CAP is falsified by Creamer's preferred Tʼiis są́ biisxį́ ‘The tree died of old age’ (1974:37), a bi- sentence with a tree as subject. And Creamer states the rock/tree rule herself: the noun the verb says is moving “takes higher status than the more stationary object” (1974:36). So rock and tree are not equal in (11)–(14); the old rule was right after all, and every equal-rank pair Creamer prints (groups 1–5, and a newborn with an elephant) allows both voices. Untested: two inanimates where neither moves.
Verdict, stated plainly (revised 2026-10-14, 4:46am, after opening Witherspoon 1977). With Dowty as the baseline, the Navajo inverse fits as an extension with three additions. (a) Proto-Agent properties are weighted, with sentience/volition above causation and movement. This was a hypothesis from Willie's lightning/horse judgment (p. 90); it is now SOURCED as Witherspoon's claim (control “primarily” by intelligence, “to a lesser extent” by animation, 1977:77). The Dowty labels are ours, and speakers vary. (b) Person (1/2 > 3) and topicality/definiteness, SOURCED as facts (Willie pp. 76–77, 86, 100), which the baseline does not address. (c) INV-CAP: the inverse subject must have potential sentience or volition. Withdrawn 12:46pm: Creamer 1974:37 has a tree as a bi- subject, and her movement rule (p. 36) explains the rock/tree data. Replaced by (c′) SOURCED (Creamer 1974:30, 36; second source) among plants and objects the event's movement entailment sets the rank, and (d) SOURCED one non-Dowty factor, “usefulness to man, or relatedness to man” (1974:30), which belongs with extension (b). One datum cuts against a per-event count: old age causes the tree's death in Tʼiis są́ biisxį́ but ranks below the tree, so the rank reads potential (capacity) more than the event's entailments (Eight ranks §6.3). Ranking by proto-role weight alone does not account for the scale. Effect on the thesis: the ranking reads the participants' proto-agent potential, and among inanimates the entailments of the event, inside the word, with no nouns present (Willie 1991:86–88; D12). That strengthens T-ROLE more than before and supports T-MID indirectly. It is still short of firm, for three reasons. It rests on one ethnosemantic source, whose author calls the rules cultural rather than grammatical. That source's ‘let’ reading of bi- is contested by Fernald, Hale and Sandoval & Jelinek. And the weighting has not been checked across speakers.
Related: Is event structure built inside the verb word? asks whether the word assembles viewpoint, situation category and causation alongside the role slots before any noun comes in. Partly, and mostly already known. Telicity can be fixed outside the word.
4. The case against, steelmanned
- The configurational analysis (Speas 1986, 1990). SOURCED (abstract) Speas grants that Jelinek's idea is “on the right track” but says it “makes the wrong prediction in certain cases” and “could allow violations of the Projection Principle”. By standard tests, overt nominals “behave like arguments”. Her solution (corrected 2026-10-14, 8:46pm, from the dissertation): the prefixes are syntactically accessible “infixes”, but she declines to make them arguments (1986:201). Nouns head a CHAIN in a θ-position, and the prefix is its non-θ tail inside the word (p. 202); the θ-role goes to the CHAIN (p. 324). Checked datum by datum on Speas 1986 on Navajo: no datum contradicts T-MID/FB. On this view Navajo is a pro-drop language with object clitics, and nouns are arguments when present. Strongest point: D1, D2 and D5 do not tell her analysis apart from Jelinek's, and D4 (fixed order) fits hers better.
- Possessor yi- binding. SOURCED (abstract) A thesis, Binding-theoretic analysis of Navajo possessor YI- (author not yet identified), argues that “the anaphoric properties of the possessive pronoun yi- provide an argument that lexical DPs are in A-position, i.e. they are not adjuncts”, against Jelinek 1984 and Willie 1991. It notes the relevant reading is “hard to get” for some speakers. Strongest point: it is a binding test that adjunction should block.
- Order is not free (D4). The textbook signature of a pronominal-argument language is missing. The discourse-configurational rescue (W&J 2000) works, but it gives up a prediction.
- Linking is not role-driven alone (D3, D4, D9). Position, person and world knowledge all pick the slot. This does not hurt T-MID, but it does refute T-STRONG.
- Willie vs W&J 2000. The same research team gives different judgments on wh-focus (T19a) and different analyses of the inverse agent (T7). Two of our tests (D7, D10) depend on which is right.
- Baker does not include Navajo. No verbal noun incorporation (Hale 2003 ms.:13), so the polysynthesis parameter cannot be invoked as independent support. 8:46pm: his later “Agreement, Dislocation…” n. 16 leaves Navajo open between IE-like agreement and wider (PP/vP) adjunction — page; Baker 1996 still unopened.
5. VPISH-aware derivations
Conventions as on the derivations page. Σ = schema step; A = word-internal linearization; B = clause level. V-P is the VPISH variant (the subject slot raises to Spec,M/SP inside the word); V-H is Hale's in-situ licensing. Entailment labels are ours.
TA1. (8) shí bi-chidí yílchxǫʼ ‘I (am the one who) ruined his car’ (Willie 1991:56)
Σ Σ(√RUIN-tr) = {ρA (vol, caus), ρTh (chg)} HYPOTHESIS (labels)
Merge ρTh in Comp,V; ρA in Spec,v VPISH: base positions
Agree π(ρTh)=3, π(ρA)=1 ("3s.1s") SOURCED (gloss)
Move V-P: ρA → Spec,M/SP [ <ρA> left in Spec,v ] HYPOTHESIS
V-H: no movement; ρA licensed in situ SOURCED (Hale 2003 ms.:29)
A spell-out yílchxǫ' — Σ closed; a complete clause SOURCED (pp. 51–52)
B adjoin bi-chidí, then shí, at the left edge SOURCED (ADJ-LEFT)
link by φ: shí(1) ↔ ρA ; bi-chidí(3) ↔ ρTh SOURCED (ADJ-PHI, p. 53)
Result: nouns enter AFTER Σ is closed; no noun is in Spec,v or Comp,V.
T-MID: ✓ T-STRONG ("role decides"): ✗ — φ decides here.
Under Speas's alternative, step B would instead be Merge of bi-chidí in Comp,V and shí in Spec,v, with the prefixes as clitics linked to them. The surface string is the same. Only D6–D8 can tell these apart.
TA2. (33) ʼashkii yá yoolóós, readings a/b vs *c (Willie 1991:67)
Σ Σ(√LEAD) = {ρA (vol, sent), ρTh [sing. +animate]} + P yá {ρBen}
Merge ρTh Comp,V; ρA Spec,v; ρBen Comp,P (PP adjoined to V') HYPOTHESIS
Move V-P: ρA → Spec,M/SP HYPOTHESIS
A spell-out yá yoolóós (MVS = PP + V) SOURCED (p. 50, (44))
B adjoin 'ashkii; candidate links: ρTh (a) ✓, ρBen (b) ✓,
ρA (c) ✗ SOURCED (p. 67)
Role check: 'ashkii is compatible with ρA (vol, sent) — yet ✗.
Result: a link that the roles allow is blocked by position.
T-STRONG: ✗. T-MID: ✓ (linking may use position).
TA3. (22) ʼashkii yiyííbį́į́ʼ ‘The boy picked them’ (Willie 1991:61)
Σ Σ(√PICK.ROUND) = {ρA (vol), ρTh [round objects]} SOURCED (gloss) / HYPOTHESIS (labels)
Merge/Move as TA1
B adjoin 'ashkii; default ADJ-OBJ proposes ρTh SOURCED (p. 59)
ρTh requires [round]; 'ashkii ✗ → link to ρA SOURCED (p. 61)
Result: the stem's Proto-Patient class decides the link.
Supports T-MID and Grant's classificatory-stem conjecture (Notes §3).
TA4. Where VPISH could tell the analyses apart
OPEN If the inverse bi- raises ρP above ρA word-internally (V-P), a quantified patient-linked noun should scope over the agent under bi- but not under yi- (T5c, Willie (58)–(60)). Under T-MID the scope must come from the slot, since the noun is outside. Under Speas, it would come from the noun's own position. A clean scope judgment on (58)–(59) is the single most useful datum for both questions.
6. Verdict by version
Update 2026-10-14 (afternoon): the feature-bundle version. Not quite a word states a sharper version of T-MID, labelled FB: the verb's argument is a φ-level bundle [person, number, capacity rank κ, role] spelled out as the prefix, and nouns link to it by matching features. No Navajo datum on the site contradicts FB. It keeps the parts of Jelinek, Willie, Hale and Speas that work, and Creamer's capacity-based ranking fits it (κ is a capacity of the referent as construed, not a per-event count). It survives Speas's general framework, whose own Warlpiri analysis uses the same linking mechanism, but her Navajo chapter has not been read, so that result is provisional. 8:46pm: the 1986 dissertation's Navajo chapters are now read, and none of their data contradicts FB (Speas 1986 on Navajo). Methodological note (Grant): no single account is treated as the judge, and only contradicting data count against a hypothesis.
| Version | Verdict | Why |
|---|---|---|
| T-STRONG | FAILS | Nouns are arguments of postpositions (D11); position and person, not just roles, decide links (D3, D4, TA2); quantifier and indefinite facts are hard for it (D8). Grant's claim in this form is wrong. |
| T-MID (sharpened as FB, feature bundles) | Survives; no contradicting data found; key tests pending | Directly stated by Willie (pp. 51–52) and Hale (ms.:5–7, 51), and supported by D1–D3, D5, D9. Not uncontested: Speas 1986/1990 and the possessor-yi- study say nouns are in argument positions. 8:46pm: Speas 1986's Navajo data are now checked: 0 cut against FB, 3 support, 15 compatible, and D6 leans toward T-MID. FB borrows two positional rules (LINK-NEAR, PAR) from Platero and Speas (page). D7–D8 are undecided. |
| T-WEAK | Survives; the mainstream analysis, contested | The published Pronominal Argument Hypothesis. Same rivals as T-MID. 2026-10-15 (12:46pm, corrected 1:07pm): Hale 1999 (10), (11a,b) (Condition C) and (5b) (extraction) argue against nouns adjoined above the subject, conditional on assumptions A1–A10 and, for (5b), an unopened analysis. Low adjunction is not excluded. Its core claim, that the prefixes are the arguments, is untouched (page, audit). |
| T-ROLE | Plausible, not shown; strengthened by §3b and D12 (revised 2026-10-14, 4:46am) | Depends on the analysis of the inverse (D10, T7). The animacy ranking reads as weighted proto-agent potential plus topicality (§3b), and a published source now says so in its own terms (Witherspoon 1977:67–81). Among inanimates it reads the event's entailments (D12). Not settled: that source treats the ranking as cultural, not grammatical, and its ‘let’ reading of bi- is contested. |
| Literal “arguments are theta roles” | Not supported | The sources say the arguments are pronouns that bear roles (§1). A slot carries person and number as well as a role. |
7. What would settle it
| # | Test or reading | Decides | Status |
|---|---|---|---|
| TA-a | Read Speas 1990 (Phrase Structure in Natural Language, Kluwer), ch. 4–5, pp. 232–240, for her argumenthood tests and the Parallel Function Constraint | D6, D7 | PARTLY DONE 2026-10-14 (8:46pm): Speas 1986 Ch. 2 §2.2.9, Ch. 3–4 read on page images (page); D6 → leans SUPPORTS. Speas 1990 itself OPEN (T22d) |
| TA-b | Read Jelinek 1984 (NLLT 2:39–76) for the original diagnostics and Navajo data | §1, D4 | PARTLY DONE 2026-10-15 (8:46pm): opened for Warlpiri tree (28) and (73b); no Navajo adjunction-height data (only an NP-hierarchy cite via Hale, Jeanne & Platero 1977). 9:20pm: tree (28) is pre-VPISH notation for Warlpiri; no verdict on A3 (T26g). page |
| TA-c | Read Barss, Hale, Perkins & Speas 1989, “Aspects of Logical Form in Navajo” (in Cook & Rice, pp. 317–334) | D6, D8 | OPEN |
| TA-d | Identify and read the possessor-yi- binding thesis | Argument against T-MID | OPEN |
| TA-e | Willie & Jelinek 2000 §§ on discourse configurationality and quantifiers; Hale & Platero NPI paper in full | D4, D8, T19a | OPEN |
| TA-g | Open Witherspoon 1977 (open access) pp. 63–82, and Hale 1973 “A note on subject-object inversion in Navajo” | §3b: does a source tie the scale to control and agency? | Done for Witherspoon (2026-10-14, 4:46am): yes, control, intelligence, will and movement, with a weighting (page). Hale 1973 itself OPEN (pp. 303–306 now known through page-checked quotations) |
| TA-h | Find a bi- sentence with an inanimate subject, or confirm there is none (INV-CAP); judgments on bi- for equal-ranked animates (T20a–b) | §3b extension (c); the equal-rank test | First half DONE 12:46pm: yes, Creamer 1974:37 Tʼiis są́ biisxį́; INV-CAP withdrawn (page). Equal-ranked animates: Creamer prints both voices for groups 1–5 (pp. 33–36). Two non-moving inanimates OPEN |
| TA-f | Speaker judgments: scope of (58)–(59); a noun inside the MVS; Condition C minimal pairs | TA4, D6 | OPEN (needs native intuitions) |
