Which noun is which? Nominal adjunction: nouns attach to role slots, and the verb can overrule word order
2026-11-11 audit pass 2. Apostrophe fixes only. Navajo glottal stops typed as ' or ’ (for example in prefix names like ʼa- and ʼeʼe-, or words like Awééʼ) now print as ʼ (U+02BC), and a build step no longer turns the closing quote of an English translation inside a Navajo row into a glottal stop (‘…the girls.’, not ‘…the girls.ʼ). No form, gloss or translation wording changed.
2026-11-11 audit pass 4. Interlinear rows whose form and gloss do not line up one-to-one were re-checked against the source page. Each is marked † with a note under its table: SOURCED where the source prints it that way, OPEN where the source cannot confirm it. See the pass 4 report.
Verbs on this page 15 entries in the verb dictionary (show)
yiztał ‘kick’ · néidííłtsʼin ‘hit with the fist’ · dabiiłtsą́ ‘see’ · yoolóós ‘lead (one animate)’ · yáłtiʼ ‘speak, talk (to)’ · yi-sh-cha ‘cry’ · neezdá ‘sit down; make sit; keep (somewhere)’ · yíchxǫʼ ‘be ruined; (classifier) ruin’ · yíkʼą́ ‘grind (e.g. corn)’ · yiyííbį́į́ʼ ‘pick (round objects, e.g. berries)’ · yiyííłtaʼ ‘count (plural objects)’ · yóldon ‘shoot at’ · yiyííłbéézh ‘cook (in something)’ · neezdá ‘sat down; with yikáá’ ‘on it’: sat down on it’ · sidá ‘sits; with yii’/bii’ ‘in’: sits in’
Restated on the unified structure (2026-11-15). Nouns are restated as arguments merged in their theta positions (VP, Spec,vP) and, for the subject, raised to Spec,TP by the working default; a lone noun is linked by the H4″ default. Adjunction and pronominal-argument analyses below are labelled contrast; where nouns sit for Condition C is to be re-run on the unified spine. Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §2 and §4, first rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.
This page reads the first two sections of Willie's (1991) Chapter 3, “The Third Person and Nominal Adjunction”: §3.0–3.1 (pp. 51–63) and §3.2, “Adjoining nominals to oblique arguments” (pp. 63–70). The site had used only the §3.3 inverse part of this chapter (pp. 71–100), and Willie's linking rule had come to the site mainly through Willie & Jelinek 2000:268. The test is a basic one for Grant's thesis. If the verb's arguments are role slots and not words, what decides which noun goes with which slot? Willie's answer, on pages opened this run, has three parts. (1) A noun “by itself cannot be an argument -- it must be licensed by a pronoun” (p. 52). The sentence is complete without it, and it is added “for reasons relating to the structure of the discourse” (p. 51). (2) By default, position decides. A lone noun goes with the object (17). With two nouns the inner one goes with the internal argument and the outer one with the external argument (20). With a postposition, the innermost noun goes with the postposition's object (31). (3) The verb's meaning can overrule position. ‘boy’ + ‘picked (round objects)’ means ‘the boy picked them’, not ‘he picked the boy’ (22). The run gives four results. (a) How many slots the verb has decides how a string of nouns is parsed. ‘[the old man's car] is ruined’ is one noun phrase. ‘[the old man] [his car] ruined’ is two, and the transitive verb also allows the one-phrase reading (§6, SCHEMA-PARSE). This supports Grant's thesis. (b) The default links are nested around the left edge of the verb-plus-postpositions unit (the MVS). The noun nearest the MVS goes with the argument whose prefix is leftmost in it (§8, MIRROR-LINK). Willie's own (36) breaks this nesting when meaning demands it. (c) Some overrides come from the verb's lexical meaning, and some come from knowledge about the world (“horses typically are led”). Dowty (1991:572, n. 16) separates these two, and only the first belongs to a role schema (§9). (d) Nouns attach to the MVS as a whole, never inside it, which bears on Willie's promise that the MVS “behaves as a unit under certain syntactic processes” (§10, MVS-UNIT).
On this page: 1 Question · 2 Sources · 3 Nouns are adjuncts · 4 Free and emphatic pronouns · 5 One noun, two nouns · 6 The schema parses the string · 7 Postpositions · 8 Nested links · 9 When meaning overrules order · 10 The MVS as a unit · 11 Rules · 12 Derivations L1–L8 · 13 VPISH · 14 Open
1. The question
This site writes a verb's arguments as a set of role slots, Σ(√) = {ρA, ρP, …}, each with Dowty entailments (proto-role schema). Prefixes value the slots inside the word (rule set A). Nouns are linked to slots at clause level (rule set B; schema C3, “each overt DP links to exactly one ρ, and its linear position is fixed by rule set B”). C3 was a HYPOTHESIS resting on Willie & Jelinek 2000 and §3.3. This page checks it against Willie's full account of how nouns are linked in the direct voice. It asks three things:
- Phenomenon 2 (roles, not words). Are nouns really outside the argument structure? And when a noun's link is not fixed by order, is it fixed by the slot's entailments?
- Phenomenon 1 (rule set A vs B). Clause-level order is head-final (nouns to the left), and the prefix order inside the word is fixed. How do the two orders line up, and around which boundary?
- VPISH. Does the order of the nouns show anything about where the slots first Merge?
2. Sources opened this run (2026-10-11, ~8:55pm PT)
- Willie, MaryAnn 1991. Navajo Pronouns and Obviation. PhD dissertation, University of Arizona (the scan on the box). Fresh 150 dpi page images of printed pp. 51–71 (PDF 57–77), i.e. all of §3.0–3.2 and the first page of §3.3. Page offset: PDF = printed + 6 for Chapter 3 (printed 52 = PDF 58, from the page headers). The printed number “51” occurs twice. PDF 56 is the last page of Chapter 2 and PDF 57 the first page of Chapter 3. That is why Chapter 2 runs at + 5 (as found on the pleonastic & psych page) and Chapters 3–4 at + 6. Every Navajo line of (1)–(36) was checked on 300 dpi crops, with extra zoom on yiyííbį́į́ʼ (22)–(23), the l/ł of the ‘ruin’ forms, shiiniltsą́ (9)–(10), and łííʼ (32), (36). No OCR was used for any form. The OCR layer was used only to find the chapter and section starts.
- Dowty, David 1991. “Thematic proto-roles and argument selection.” Language 67(3):547–619. Page image of printed p. 572 (the stored 150 dpi image
dpg-27.png): (27)–(28) and n. 16, which separates entailments of the predicate from inferences drawn from the whole sentence.
Transcription and glosses. Navajo as printed, with the glottal stop as ’ and the nasal hook as ą/į/ǫ. Willie's gloss lines are copied exactly, misprints included and marked [sic]. Her gloss order for verbs is object.subject (“3s.1s.ruined” = it, by me) and for postpositions object.P (“3.for”). Hyphens inside nouns (bi-chidí) are hers. Verbs are not segmented beyond her glosses.
Cited by Willie in these pages, not opened: Young & Morgan 1987:120 (“simple passive”, for (12)); Willie 1989 (Direct vs Inverse); Chomsky 1981.
3. Nouns are adjuncts, licensed by a pronoun (§3.0–3.1) SOURCED
P. 51: “In this chapter, I will describe the process whereby nominals are adjoined to the MVS, and how they are properly coindexed with the pronominal arguments.” Pp. 51–52: “The MVS is a complete functional complex, and by itself is a grammatical sentence. Nominal expressions are always added to the MVS for reasons relating to the structure of the discourse, not for reasons of grammaticality. When for discourse reasons the speaker desires to remove any potential ambiguity from the reference of a pronominal argument, a coreferential noun may be included in the utterance, and these nominals are in adjunct positions.” P. 52: “The noun by itself cannot be an argument -- it must be licensed by a pronoun; but it forms a complex expression with the pronoun, since it modifies the pronoun.” P. 52 (§3.1.1): “Navajo is a verb final language, and adjoined nominals appear at the left of the MVS. Except for a few terms that refer to humans, nouns are not marked for number, and nouns and free pronouns are never marked for case, since these features are marked in the incorporated pronouns.” Pp. 69–70: “nouns in Navajo are not in argument positions, where they are directly governed and assigned case by the verb or postposition, but are linked via predication to the incorporated pronouns, like "discontinuous" relatives. While a noun in and of itself cannot fill an argument position, it forms a complex constituent with its pronominal head at the level of the interpretation of the sentence.” P. 70: “a nominal may be adjoined for each of the pronouns in the MVS.”
GRANT'S CONJECTURE matched by SOURCED. Grant's thesis is that Navajo verbs take theta roles, not lexical items, as arguments. These statements are the most direct support for it that the site has page-checked. The arguments are the incorporated pronouns, one per slot. A noun is an optional modifier of one of them and is never the argument itself. The site's wording is that the slots are role slots with Dowty entailments, while Willie's is that they are “pronominal arguments”. Whether these differ is T7 (open questions).
4. Free and emphatic pronouns (§3.1.1–3.1.2) SOURCED
Willie first takes the free pronouns. They “have a particular function, that of marking referential contrast or emphasis, and therefore do not appear in "unmarked" constructions” (p. 53). With a 1st- or 2nd-person argument, “only a free pronoun with the same "phi"-features (person and number) may be adjoined” (p. 53). That gives her principle (2): “A adjoined [sic] nominal must agree in semantic ("phi") features with the pronoun it is coindexed with” (p. 53). It is “fundamental to the interpretation to all adjoined nominals in Navajo” (p. 54).
| (1) | nihí | yiicha | |
| we | 2d.crying [sic?] | ||
| ‘We (are the ones who) are crying.’ | |||
| (3) | Tʼáábí | Mary | yiztał |
| 3.EMPH | Mary | 3.3.kicked | |
| ‘He himself kicked Mary.’ | |||
| (4) | Tʼáábí | ʼádílcha | |
| 3 EMPH | self-3-cry | ||
| ‘He made himself cry.’ | |||
| (5) | Tʼáábí | yicha | |
| 3.EMPH | 3.cries | ||
| ‘He himself is crying. (on his own)’ | |||
| (7) | tʼááshí · tʼáání · tʼáábí | tʼáánihí · tʼáánihí · tʼáábí | |
| (two columns, no gloss line; “produced either by adding a particle tʼáá, meaning "just" or by a shortened form -a”, p. 55) | |||
| (8) | shí | bi-chidí | yílchxǫʼ |
| I | 3.car† | 3s.1s.ruined | |
| ‘I (am the one who) ruined his car.’ | |||
| (9) | shí | shiiniltsą́ | |
| I | 1s.2s.saw | ||
| ‘I (am the one) you saw.’ | |||
| (10) | *ní | shí | shiiniltsą́ |
SOURCED † Segmentation note (2026-11-11 audit pass 4): Willie 1991 (8) prints bi-chidí with the gloss “3.car”: the gloss joins with periods or colons pieces the form separates with hyphens, so the two lines do not match one-to-one in the source. Kept exactly as printed.
P. 56: “With transitives, it is possible to add a free pronoun that agrees in person and number with only one of the arguments.” “Constructions with two free pronouns are judged unacceptable, since the function of the adjoined free pronoun is contrastive reference.” P. 54: “It is the reflexive pronoun ʼádi on the verb that makes (4) a reflexive sentence, not the emphatic pronoun”.
HYPOTHESIS (ours) Two points for the site. First, a free pronoun is linked by its person value alone. In (8) shí ‘I’ can only go with the 1st-person slot, which is the subject, and in (9) only with the 1st-person slot, which is the object. That is PI-MATCH / SLOT-TYPE-LINK (pleonastic & psych §10), with Willie's (2) as its source. Order plays no part. Second, *(10) bans two free pronouns even though each would match a different slot. So the ban is not about slots. It is about one contrast per clause (T18h).
5. One noun, two nouns (§3.1.3–3.1.4) SOURCED
| (11) | (ʼashkii) | yicha | |
| boy | 3s.crying | ||
| ‘(The boy) he is crying.’ | |||
| (12) | yíchxǫʼ | ||
| 3s.ruined | |||
| ‘It is ruined.’ | |||
| (13) | bi-chidí | yíchxǫʼ | |
| his car | 3s.ruined | ||
| ‘His car, it is ruined.’ | |||
| (14) | [hastiin sání bi-chidí] | yíchxǫʼ | |
| man old 3POSS.car | 3s.ruin PERF (p. 58); “3-3-ruined” when repeated on p. 60 [sic] | ||
| ‘The old man's car is ruined.’ | |||
| (15) | (naadą́ą́) | yíkʼą́ | |
| corn | 3.1.ground | ||
| ‘(The corn) I ground it.’ | |||
| (16) | (bi-chidí) | yílchxǫʼ | |
| (3.car)† | 3s.1s.ruined | ||
| ‘(His car) I ruined it.’ | |||
| (18) | chidí | yiyíílchxǫʼ | |
| car | 3s.3s.ruined | ||
| ‘He/she ruined the car.’ | |||
| (19) | [hastiin sání] | [bi-chidí] | yiyííłchxǫʼ |
| man old | 3POSS.car† | 3.3.ruined | |
| ‘The old man ruined his car.’ | |||
| (21) | [hastiin sání bi-chidí] | yiyííłchxǫʼ | |
| man old 3POSS.car | 3-3-ruined | ||
| ‘He/she ruined the old man's car.’ | |||
| (22) | ʼashkii | yiyííbį́į́ʼ | |
| boy | 3.3.picked round objects | ||
| ‘The boy picked them. (c.f. berries)’ | |||
| (23) | [ʼashkii bimá] | yiyííbį́į́ʼ | |
| boy 3s.mother | 3.3.picked round objects | ||
| ‘The boy's mother picked them.’ | |||
| (24) | ʼashkii | yiyííłtaʼ | |
| boy | 3.3.count plural objects | ||
| ‘The boy counted them.’ | |||
| (25) | ʼashkii | yóldon | |
| boy | 3.3.shooting | ||
| ‘The boy is shooting at it.’ | |||
SOURCED † Segmentation note (2026-11-11 audit pass 4): Willie 1991 (16) prints (bi-chidí) with the gloss “(3.car)”: the gloss joins with periods or colons pieces the form separates with hyphens, so the two lines do not match one-to-one in the source. Kept exactly as printed.
SOURCED † Segmentation note (2026-11-11 audit pass 4): Willie 1991 (19) prints [bi-chidí] with the gloss “3POSS.car”: the gloss joins with periods or colons pieces the form separates with hyphens, so the two lines do not match one-to-one in the source. Kept exactly as printed.
P. 57: “The single argument of an intransitive may have an agent, patient or theme theta role.” Of (12): “Young and Morgan (1987:120) call this the simple passive. Only one nominal, coindexed with the patient subject, is allowed in these constructions.” P. 59: “The nominal in both these constructions must be coindexed with the object, since the subject is first person.” P. 59, (17): “When a transitive sentence has only one adjunct, that adjunct is interpreted as coindexed with the object (patient) argument.” P. 60: “When a transitive verb has two third person arguments, it is possible to add two nominal adjuncts. Such sentences are not common in ordinary discourse.” P. 60, (20): “When a transitive sentence includes two nominal adjuncts, the inner nominal is coindexed with the internal argument, and the outer nominal (by default) is coindexed with the external argument.” P. 61: “If the transitive verb describes an action which can only be performed by an animate subject, these semantic features of the verb can "override" the expected interpretation principle stated in (20).” “Since berries do not pick boys, this sentence poses no problems of interpretation, and the expected word order can be suspended. This applies to long complex nominals as well.” P. 62: “In (23), the initial string cannot be interpreted as two constituents, even though the verb is transitive, because of the semantic features of the verb.” “(22) through (25) have the interpretation of the single nominal being coindexed with the subject. This coreferentiality between the nominal and the pronominal follows from what we know about what humans can do, such as picking, counting, and the like.”
Two small points as printed. Willie's text says the override suspends principle (20), the two-noun rule, but (22) and (24)–(25) have one noun, so the rule actually overridden there is (17). And p. 62 has “the expected work order” [sic].
6. The schema parses the string HYPOTHESIS (on SOURCED data)
Willie sets (14) beside (19) and (21) on purpose (pp. 60–61). The nominal string hastiin sání bi-chidí ‘man old his-car’ is the same in all three:
| Verb | Slots in Σ | Parse of the string | Reading |
|---|---|---|---|
| (14) yíchxǫʼ (intransitive) | 1: ρTh | one constituent, “obligatory” (p. 60) | ‘The old man's car is ruined.’ |
| (19) yiyííłchxǫʼ (transitive) | 2: ρA, ρTh | two constituents | ‘The old man ruined his car.’ |
| (21) yiyííłchxǫʼ (transitive) | 2 | one constituent, linked to the patient (p. 61) | ‘He/she ruined the old man's car.’ |
| (23) yiyííbį́į́ʼ (transitive) | 2 | one constituent only, because of “the semantic features of the verb” (p. 62) | ‘The boy's mother picked them.’ |
SCHEMA-PARSE (ours). The number of slots in Σ, and what each slot can hold, sets an upper bound on how many constituents the clause-level string can be split into. With one slot only one parse is possible. With two slots both parses are possible, unless the slots' content excludes one, as in (23). For Grant's thesis this is the clearest case on the site so far. The nouns do not decide the structure. The role schema decides how the nouns are grouped. It follows from Willie's p. 70 statement (“a nominal may be adjoined for each of the pronouns in the MVS”) plus p. 60. We did not find it stated as a rule in her text.
7. Postpositions (§3.2) SOURCED
| (26) | ʼatʼééd | yichʼįʼ | yáłtiʼ | |
| girl | 3.to | 3.speaking | ||
| ‘He is speaking to the girl.’ | ||||
| (27) | tsé | yikááʼ | neezdá | |
| rock | 3.on | 3.sat | ||
| ‘He sat down on the rock.’ | ||||
| (28) | ʼashkii | yikááʼ | neezdá | |
| boy | 3.on | 3.sat | ||
| ‘The boy sat down on it./He sat on the boy.’ | ||||
| (29) | yá | yoolóós | ||
| 3.for | 3.3.lead [sing. +animate] | |||
| ‘She is leading it for him.’ | ||||
| (30) | ʼashkii | ʼatʼééd | yá | néidííłtsʼin |
| boy | girl | 3s.for | 3.3.fist-hit | |
| ‘He hit the boy for the girl.’ | ||||
| (32) | łííʼ | yá | yoolóós | |
| horse | 3.for | 3.3.lead [sing. +animate] | ||
| ‘She is leading the horse for him.’ | ||||
| (33) | ʼashkii | yá | yoolóós | |
| boy | 3.for | 3.3.lead [sing. +animate] | ||
| a. ‘He is leading the boy for her.’ b. ‘She is leading it for the boy.’ c. *‘The boy is leading it for her.’ | ||||
| (34) | ʼazdzą́ą́ | ʼatʼééd | yá | yoolóós |
| woman | girl | 3.for | 3.3.lead [sing. +animate] | |
| a. ‘He is leading the woman for the girl.’ b. ‘The woman is leading it for the girl.’ c. *‘The woman is leading the girl for him.’ | ||||
| (35) | ʼasdzáá | yiiʼ | yiyííłbéézh | |
| woman | 3.in | 3.3.cooked | ||
| ‘The woman cooked it in it.’ | ||||
| (36) | ʼashkii | ʼatʼééd | łííʼ | yá yoolóós |
| boy | girl | horse | 3.for 3.3.lead [sing. +animate] | |
| ‘The boy is leading the horse for the girl.’ | ||||
P. 63: “The interpretive rule given in (17) applies also to sentences where the MVS is composed of an intransitive verb and a postpositional phrase.” “the single nominal is coindexed with the object of the verb if there is one, or with the postpositional object if the verb is intransitive.” “If the semantic features of the noun are incompatible with its being an adjunct to the object of the postposition, by default it may be interpreted as an adjunct to the subject.” P. 64: “Grammatical postpositional phrases may be phonologically integrated into the MVS, or they may form the first word in the MVS”. “There are three argument positions in this sentence; we can add a nominal to any of them.” P. 65: “If the semantics of the verb permit a human goal, then the nominal that immediately precedes the MVS will be interpreted as an adjunct to the postpositional object”. (31): “In an MVS with one Postpositional Phrase, the innermost nominal is coindexed with the postpositional object in the Direct Construction.” P. 66: “If the innermost adjoined noun has semantic features that make for its being interpreted as an adjunct to the object of the verb, then the generalization given in (31) will be overridden.” Of (32): “This interpretation follows from the fact that horses typically are led, and do not lead people. Note that on this interpretation, the noun preceding the inflected postposition (yá) is not coindexed with it, and cannot be governed by it.” P. 67, of (33): “On (33a), the nominal ’ashkii is interpreted as coindexed with the object of the verb, but it is not in a position to be governed by the verb. Note also that this single adjoined nominal cannot be interpreted as a subject adjunct; I will return to this point in Sec. 3.2. below.” P. 68: “Principle (31) above accounts for the exclusion of the reading shown in (34c).” P. 69: “it is entirely a matter of the semantic features of the verb and the adjoined nominal whether the single argment [sic] is to be coindexed with the verbal or postpositional object, or neither.” “When the single nominal cannot be coindexed with either object that it precedes, because of the semantic features of the MVS, it must be coindexed with the subject.”
Points as printed. P. 67 promises a return to the subject point “in Sec. 3.2. below”, but p. 67 is already in §3.2, and we found no later return on pp. 67–70. (36) is the same sentence as Willie's (85) on p. 95 (inverse §2.5), but (36) spells ‘horse’ łííʼ with no nasal hooks, as (32) does, while (85) and (37) print łį́į́ʼ. ‘woman’ is ʼazdzą́ą́ in (34) and ʼasdzáá, with no hooks, in (35). The ‘ruin’ forms with a 1st-person or 3/3 subject have plain l in (8), (16) and (18) (yílchxǫʼ, yiyíílchxǫʼ) but ł in (19) and (21) (yiyííłchxǫʼ). Willie p. 39 says the transitive is marked by the classifier, and (26) there has ʼííłchxǫʼ. So plain l may be a misprint, but we leave it as printed (T18f).
8. Phenomenon 1: the default links are nested around the MVS edge HYPOTHESIS
SOURCED inputs. (i) Inside the MVS the exponents of the arguments are in a fixed order: “the patient always to the left of the agent, and the postpositional phrases leftwards of the direct arguments” (p. 50, P-LEFT on the pleonastic & psych page). (ii) Outside it, the default links are (17), (20) and (31).
Put the two side by side. Number the exponents from the left edge of the MVS inwards: e1 = the postposition's object prefix (if any), then the verb's object prefix, then the subject. Number the nouns from the MVS outwards: N1 is the noun right next to it. The default links are then:
(30) 'ashkii 'at'ééd [ y-á néidííłts'in ]
N2 N1 [ e1 ... e2 ... e3 ... √ ] (prefix segmentation schematic)
boy ↔ e2 girl ↔ e1 PP obj V obj subj
└──────────────┼────────────┘ │
└───────────────────────┘
N1 ↔ e1 (31); N2 ↔ e2. Nested, not crossing.
(19) [hastiin sání] [bi-chidí] [ yi- ... (subj) ... √ ]
N2 N1 e1 e2
N1 ↔ e1 (20, "inner ↔ internal"); N2 ↔ e2 ("outer ↔ external").
MIRROR-LINK (ours). By default, the k-th noun counting outwards from the MVS links to the k-th exponent counting inwards from the MVS's left edge. The links are nested around that edge, so the clause-level order (rule set B) is the mirror image of the prefix order (rule set A plus the postpositions). Three things follow.
- The pivot is the MVS edge (B3), not the verb word (B2). In (30) the postposition's object is e1, outside the verb word, and the nesting counts it. A nesting centered on B2 would link N1 to the verb's object, which is reading *(34c)-type. That is a second datum for T2, in the same direction as WORD-EDGE′: the slots are counted up to B3.
- It is a default, and Willie's own data break it. In (36) the nouns are ‘boy girl horse’ and the links are boy ↔ subject (e3), girl ↔ PP object (e1), horse ↔ verb object (e2). So N1 ↔ e2 and N2 ↔ e1, and the links cross. Willie licenses this through (32): a horse is led and is not the one led for. (34b) is a milder case: N2 skips the unexpressed verb object and goes to the subject. So MIRROR-LINK describes the order nobody has to explain. It does not say what is possible.
- What it does not show. The nested order is also what one expects if nouns are ordered by how high their slots are (subject highest, outermost). The two descriptions make the same predictions on these data. The (36) crossing is a counterexample for both. It shows that, for nouns, order is overridden by meaning, not by structure.
The segmentation in the diagram is schematic. Willie does not segment néidííłtsʼin here, and the position of e2 and e3 inside it is read off P-LEFT, not off this form. That yá is y- + -á follows her gloss “3.for”. The inverse is left aside here (T18d).
9. Phenomenon 2: when meaning overrules order, and which meaning
9.1 Dowty's distinction SOURCED
Dowty 1991:572, n. 16: “It is important here to distinguish entailments of the PREDICATE from what follows from any one sentence as a whole (e.g. entailments that may arise in part from NP meanings, etc.). For example, if Mary slapped John is true, and John is a normal human, then, slapping being the kind of action it is, we would conclude that John necessarily perceives something … But it does not follow that the direct object of slap is entailed to have the P-Agent property of sentience, since we can also felicitously say Mary slapped the table or Mary slapped the corpse. However, the object of awaken does have the P-Agent entailment of sentience, as is revealed by the anomaly of #Mary awakened the table/the corpse.”
9.2 Willie's overrides, sorted by Dowty's test HYPOTHESIS
Willie gives one cover term, “the semantic features of the verb” (pp. 61–62, 66, 69). By Dowty's n. 16 her overrides are of two kinds. (The sorting is ours. Willie does not cite Dowty.)
| Example | What excludes the default link | Kind (Dowty n. 16) | Part of Σ? |
|---|---|---|---|
| (22), (23) yiyííbį́į́ʼ | The object must be “round objects” (gloss), and a boy is not one. Willie's reason is “berries do not pick boys” (p. 61). | Predicate property of ρTh (the classificatory stem). Willie's reason also uses the agent's volition (P-A (27a)). | Yes: a property of the slot |
| (24) yiyííłtaʼ | The object must be “plural objects” (gloss), and ʼashkii is one boy. Most nouns are unmarked for number, but “a few terms that refer to humans” are marked (p. 52), and ‘boys’ is ʼashiiké (Willie (63b), number & scope §2.4). | Predicate property of ρTh | Yes |
| (25) yóldon | Nothing in the gloss excludes ‘he is shooting at the boy’. Willie: “what we know about what humans can do” (p. 62). | Unclear: no predicate property is visible | OPEN (T18a) |
| (32) łííʼ yá yoolóós | “horses typically are led, and do not lead people” (p. 66). Both ‘it is led’ and ‘led for’ allow an animal, since the gloss is [sing. +animate] and grammatical postpositions allow animates (p. 65). | Sentence-level, world knowledge (“typically”) | No |
| (35) ʼasdzáá yiiʼ yiyííłbéézh | A woman is neither the thing cooked nor the container ‘in it’, so the noun goes to the subject (p. 69). | Mixed: ‘in’ needs a container (P property), and ‘cooked’ needs food (world knowledge, not entailed) | Partly |
HYPOTHESIS ENT-LINK (ours). A noun can be linked to slot ρ only if it is compatible with the predicate properties of ρ, meaning its Dowty entailments plus the stem's sortal class (round, plural, [sing. +animate]). Among the compatible slots, the defaults (17)/(20)/(31) choose, and world knowledge can choose against them. Only the first step belongs to the role schema. Grant's thesis predicts the first step. It does not predict the second, and it does not need to, because linking nouns is not argument structure on Willie's account (p. 70: the link holds “at the level of the interpretation of the sentence”).
9.3 What cuts against Grant's thesis here
- OPEN (25) ʼashkii yóldon ‘The boy is shooting at it’. If no slot property excludes ‘he is shooting at the boy’, then a linking choice that Willie states without qualification is not driven by the role schema. Either the schema has properties we cannot see in the gloss, or the linking is pragmatic (T18a).
- OPEN *(33c) ‘The boy is leading it for her’. A boy can lead, and ρA of ‘lead’ (vol, sent) fits a boy perfectly, yet the reading is out. So compatibility with a slot's entailments is not sufficient for a link. Position also restricts. A lone noun before a PP + transitive verb may go with either object but not with the subject, unless both objects are excluded ((35)). Yet in (28), with an intransitive verb + PP, the lone noun ʼashkii can be the subject (‘The boy sat down on it’) even though nothing excludes the PP reading, and the text on p. 63 says the subject reading needs incompatibility (T18b). We do not resolve this.
10. The MVS as a unit (44d) HYPOTHESIS
SOURCED Willie's (44) (p. 50) lists four properties of the MVS. (d) is that “it behaves as a unit under certain syntactic processes (to be discussed in Chap. 3)”. The process Chapter 3 opens with is nominal adjunction “to the MVS” (p. 51), with nouns “at the left of the MVS” (p. 52). Two of its statements locate the nouns outside both the verb and the postposition. The noun before yá in (32) “is not coindexed with it, and cannot be governed by it” (p. 66). In (33a) ʼashkii “is not in a position to be governed by the verb” (p. 67). And grammatical postpositions never take a noun as object (p. 39, GPP/LPP).
MVS-UNIT (ours). Rule set B adjoins nouns to the MVS as a whole. It never adjoins them to the verb word or to a grammatical PP inside it, so no noun is ever between a grammatical PP and its verb. This is our reading of (44d). Willie does not say on these pages that adjunction is the process (44d) meant. No starred example with a noun inside the MVS is printed on pp. 51–70 (T18g). If MVS-UNIT holds, rule set B “sees” the MVS as one head: B3 is where clause syntax starts. Together with WORD-EDGE′ (Σ closes at B3) and MIRROR-LINK (the nesting pivots at B3), that makes three properties that converge on B3, against one (prefix linearization) at B2.
11. Rules
Notation as on the pleonastic & psych page. π(ρ) is a slot's value. Nk is the k-th adjoined noun counting leftwards from the MVS. ek is the k-th argument exponent counting rightwards from the MVS's left edge (P-LEFT).
ADJ-NOTARG A noun is never an argument: it "must be licensed by a pronoun",
is "linked via predication" to it, and is added for discourse,
not grammaticality. (pp. 51–52, 69–70)
ADJ-LEFT Adjoined nominals appear at the left of the MVS; not governed or
case-marked by V or P. (pp. 52, 66, 67, 69)
ADJ-ONE At most one nominal per pronoun in the MVS. (p. 70; (36))
ADJ-PHI An adjoined nominal agrees in "phi" features with its pronoun.
(p. 53, (2); (8), (9), (15), (16))
ADJ-FREE Free pronouns mark contrast/emphasis; at most one per clause.
(pp. 53, 56; (1), (8)–(10))
ADJ-OBJ One nominal, transitive verb: it links to the object (patient).
One nominal, intransitive verb + PP: it links to the PP object.
(p. 59, (17); p. 63, (26))
ADJ-TWO Two nominals: inner ↔ internal argument; outer ↔ external argument
"by default". (p. 60, (20))
ADJ-PP MVS with one PP (direct voice): the innermost nominal ↔ the PP
object. (p. 65, (31); *(34c))
ADJ-SEM "Semantic features" of verb, P and noun can override ADJ-OBJ,
ADJ-TWO and ADJ-PP. (pp. 61–62, 66, 69; (22)–(25), (32))
ADJ-SUBJ A lone nominal that fits neither object links to the subject;
with a PP + transitive verb it cannot otherwise be the subject.
(pp. 67, 69; *(33c), (35))
ADJ-AMB A lone human nominal before PP + transitive verb is ambiguous
between the two objects. (p. 67; (33))
| Rule | Status | Basis |
|---|---|---|
| ADJ-NOTARG, ADJ-LEFT, ADJ-ONE, ADJ-PHI, ADJ-FREE, ADJ-OBJ, ADJ-TWO, ADJ-PP, ADJ-SEM, ADJ-SUBJ, ADJ-AMB | SOURCED | Willie's statements and examples as cited. The rule names are ours. ADJ-OBJ and ADJ-TWO match INV-6 on the inverse page (there from pp. 72–73 and W&J 2000:268). |
| SCHEMA-PARSE. |Σ| and the slots' content bound the number of constituents in the adjoined string. | HYPOTHESIS | (14) vs (19)/(21); (23). §6. |
| MIRROR-LINK. Default: Nk ↔ ek (nested around B3). | HYPOTHESIS, default only | ADJ-TWO + ADJ-PP + P-LEFT. Broken by (36) and (34b). §8. |
| ENT-LINK. A link to ρ needs compatibility with ρ's predicate properties (Dowty entailments + sortal class). The defaults and world knowledge choose among compatible slots. | HYPOTHESIS | §9.2; Dowty 1991:572 n. 16. Not sufficient: *(33c). Unexplained: (25). |
| MVS-UNIT. Rule set B adjoins nouns to the MVS as a whole, never inside it. | HYPOTHESIS | (44a, d); pp. 51–52, 66–67; GPP/LPP. §10. |
| Schema C3 (schema): “linear position is fixed by rule set B” | HYPOTHESIS, now refined | Position fixes a default link only. ADJ-SEM and ENT-LINK can choose another slot. |
12. Derivations L1–L8, step by step
Conventions as on the derivations page: Σ schema step, A word-internal linearization (prefixes left of the stem), B clause level (complements and adjuncts left of their head). V-P is the VPISH variant, in which the subject slot raises to Spec,M/SP inside the word. V-H keeps it in situ (Hale). Entailment labels are ours, from Willie's translations (Dowty 1991:572 (27)–(28)). Verbs are not segmented beyond Willie's glosses.
L1. (11) (’ashkii) yicha ‘(The boy) he is crying’ and (1) nihí yiicha
- Σ Σ(√CRY) = {ρA}, one slot (vol?, sent; ours).
- Merge ρA in Spec,v. Agree: π = 3 (gloss “3s.crying”). (V-P) Move ρA to Spec,M/SP. (V-H) no move. Same string.
- A Spell-out yicha. The word closes (B2). No PP, so the MVS closes too (B3), and Σ is closed (WORD-EDGE′).
- B Optional: adjoin ʼashkii at the left of the MVS (ADJ-LEFT). There is one slot, so the only link is ρA. ADJ-PHI: a noun counts as 3, and π = 3, so the link is allowed.
- (1): π(ρA) is non-singular (gloss “2d” [sic?], translation ‘We’). B adjoin the free pronoun nihí. ADJ-PHI: its person and number must match the slot. ADJ-FREE: the reading is contrastive, ‘we are the ones who …’.
L2. (8) shí bi-chidí yílchxǫʼ, (9) shí shiiniltsą́, *(10) ní shí shiiniltsą́
- Σ Σ(√RUINtr) = {ρA (vol, caus), ρTh (chg)}.
- Merge ρTh in Comp,V and ρA in Spec,v. Agree: π(ρTh) = 3, π(ρA) = 1 (gloss “3s.1s”). (V-P) Move ρA to Spec,M/SP.
- A Spell-out yílchxǫʼ (as printed; T18f). B2 = B3, so Σ closes.
- B Adjoin bi-chidí, then shí, both on the left. Linking is by π, not by order. shí is 1, so it goes to ρA. bi-chidí is 3, so it goes to ρTh (ADJ-PHI; Willie p. 59 for (16)). Here the order (N2 = subject, N1 = object) happens to match MIRROR-LINK, but nothing depends on it.
- (9): π(ρTh) = 1, π(ρA) = 2 (“1s.2s.saw”). shí links to the 1st-person slot, which here is the object: ‘I (am the one) you saw.’ So the same free pronoun goes with a different role in (8) and (9), and the role is decided by the value of the slot.
- *(10): ní would link to ρA (2) and shí to ρTh (1). Both links are allowed by ADJ-PHI and ADJ-ONE. The clause is out by ADJ-FREE: there can be only one contrast.
L3. One string, two parses: (14) yíchxǫʼ vs (19)/(21) yiyííłchxǫʼ
- Σ (14): Σ(√RUINintr) = {ρTh}. (19)/(21): Σ(√RUINtr) = {ρA, ρTh}.
- Merge / Agree: (14) π(ρTh) = 3. (19)/(21) π = 3, 3 (“3.3”). (V-P) Move the subject slot (ρTh in (14), which is a patient subject, Willie p. 57; ρA in (19)) to Spec,M/SP. Prefix order is unchanged.
- A Spell-out yíchxǫʼ / yiyííłchxǫʼ. Σ closes.
- B (14): there is one slot, so ADJ-ONE allows one adjunct. The string must be parsed as one DP, [hastiin sání bi-chidí], and it links to ρTh. The bi- inside the DP refers to the old man, DP-internally.
- B (19): there are two slots. Parse as two DPs. ADJ-TWO: inner [bi-chidí] ↔ ρTh, outer [hastiin sání] ↔ ρA: ‘The old man ruined his car.’ (21): parse as one DP, and ADJ-OBJ links it to ρTh: ‘He/she ruined the old man's car.’ SCHEMA-PARSE: the verb's |Σ| sets which parses exist.
L4. (18) chidí yiyíílchxǫʼ vs (22) ʼashkii yiyííbį́į́ʼ, (24) ʼashkii yiyííłtaʼ
- Σ (18): {ρA, ρTh (chg)}. (22): {ρA (vol), ρTh: [round objects]}. (24): {ρA (vol, sent), ρTh: [plural objects]}. The sortal classes are from Willie's glosses.
- Merge / Agree π = 3, 3 in all three. (V-P) Move ρA. A Spell-out. Σ closes.
- B Adjoin one noun. ADJ-OBJ proposes ρTh. (18): chidí ‘car’ is compatible, so the link holds: ‘He/she ruined the car.’
- B (22): ʼashkii is not a round object, so ENT-LINK blocks ρTh. The only other slot is ρA, and a boy is compatible with it, so ADJ-SEM gives ‘The boy picked them.’ (24): one boy is not “plural objects”, so the same reasoning gives ‘The boy counted them.’ (23): the same block applies to the two-DP parse, since [ʼashkii] [bimá] would put ‘mother’ in ρTh. So only the one-DP parse is left (SCHEMA-PARSE + ENT-LINK).
L5. (30) ʼashkii ʼatʼééd yá néidííłtsʼin ‘He hit the boy for the girl’
- Σ Σ(√FIST-HIT) = {ρA (vol, caus), ρP (aff)} + Obl-á(ρBen) (sent?; ours). Under WORD-EDGE′, ρBen is part of Σ through the grammatical PP (T17d).
- Merge P = -á ‘for’ with ρBen as its object. Agree: π(ρBen) = 3, spelled y- (Willie: “3s.for”). A yá.
- Merge ρP, ρA. Agree π = 3, 3 (“3.3”). (V-P) Move ρA to Spec,M/SP. A néidííłtsʼin. B2 closes.
- B-ordered, inside Σ: [yá] to the left of the verb word (P-LEFT). B3 closes, and Σ closes.
- B Adjoin ʼatʼééd, then ʼashkii, at the left of the MVS. ADJ-PP: N1 ʼatʼééd ↔ ρBen. The verb “permit[s] a human goal” (p. 65), so ENT-LINK is satisfied. N2 ʼashkii ↔ ρP, as Willie's translation shows. MIRROR-LINK: N1 ↔ e1, N2 ↔ e2, nested around B3. Note that ADJ-TWO's “outer ↔ external” is not what happens here: the outer noun is the hit boy, not the hitter.
L6. (33) ʼashkii yá yoolóós: two readings, and not a third
- Σ Σ(√LEAD) = {ρA (vol, sent, caus, mov), ρP (mov; [sing. +animate])} + Obl-á(ρBen). Merge / Agree all 3. (V-P) Move ρA. A yá + yoolóós. B3 closes.
- B Adjoin one noun, ʼashkii. ENT-LINK: a boy is compatible with all three slots.
- ADJ-AMB: it may link to ρP, giving (33a) ‘He is leading the boy for her’, or to ρBen, giving (33b) ‘She is leading it for the boy.’ So (31)'s “innermost ↔ PP object” is not forced for a lone noun.
- ADJ-SUBJ: ρA is not available while an object is, so (33c) *‘The boy is leading it for her’ is out. Compatibility with the slot was not enough (§9.3).
L7. (34) ʼazdzą́ą́ ʼatʼééd yá yoolóós: a, b, *c
- Σ, Merge, Agree, Move, A as in L6.
- B Adjoin ʼatʼééd (N1), then ʼazdzą́ą́ (N2). ADJ-PP: N1 ↔ ρBen in both licit readings.
- N2 may link to ρP, giving (34a) ‘He is leading the woman for the girl’ (nested: N2 ↔ e2). Or it may link to ρA, giving (34b) ‘The woman is leading it for the girl’ (N2 ↔ e3, skipping the unexpressed e2; still non-crossing).
- *(34c) ‘The woman is leading the girl for him’ needs N1 ↔ ρP. That violates ADJ-PP, and Willie says so (p. 68). Contrast (32)/(36), where N1 ↔ ρP is allowed because the noun is a horse. So ADJ-PP yields to ADJ-SEM only when the noun does not suit the PP object. A girl suits both.
L8. (35) ʼasdzáá yiiʼ yiyííłbéézh and (36) ʼashkii ʼatʼééd łííʼ yá yoolóós
- Σ (35): Σ(√COOK) = {ρA (vol, caus), ρTh (chg)} + Obl-ii’ ‘in’ (container). Willie's gloss “3.in” has a pronominal object, so we treat it as grammatical. The p. 39 lists were not re-checked for -ii’. Merge / Agree 3, 3, 3. (V-P) Move ρA. A yii’ + yiyííłbéézh.
- B (35): one noun, ʼasdzáá ‘woman’. ADJ-PP proposes the PP object, but a woman is not a container, so the link is blocked. ADJ-OBJ proposes ρTh, and a woman is not what is cooked, so that is blocked too (world knowledge, §9.2). ADJ-SUBJ: link to ρA. ‘The woman cooked it in it.’
- B (36): three nouns, one per slot (ADJ-ONE). N1 łííʼ: ADJ-PP proposes ρBen, and ADJ-SEM (the (32) reason) overrides that, so it links to ρP. N2 ʼatʼééd ↔ ρBen. N3 ʼashkii ↔ ρA. The links N1 ↔ e2 and N2 ↔ e1 cross, so MIRROR-LINK is overridden (§8). Same string and reading as (85), p. 95.
13. VPISH
HYPOTHESIS Three points.
- Nouns never sit where VPISH puts subjects. On Willie's account (ADJ-NOTARG, ADJ-LEFT) no noun is ever in Spec,vP or in a raised subject position. Nouns are outside the MVS and “not in a position to be governed by the verb” (p. 67). So VPISH, if it applies, applies to the slots and their prefixes inside the word (V-P), never to nouns. That is the PAH reading the site has used since the derivations page. Chapter 3 states it outright.
- Noun order cannot tell V-P from V-H. MIRROR-LINK puts the subject's noun outermost. That is what one expects if the subject slot is the highest, and it is the highest in both V-P (after raising) and V-H (Spec,v above Comp,V). And (36) shows that noun order can be overridden by meaning, so it is not a reliable readout of slot height anyway.
- The patient-subject intransitive. (12) yíchxǫʼ ‘It is ruined’ has a patient subject (p. 57), which in VPISH terms is an internal argument promoted to subject. One noun links to it ((13), (14)). Nothing on these pages shows whether the slot moves. The prefix gives only a 3, and noun order has one noun. This is the unaccusative question again (plain ł-causer), with no new test.
14. Open questions
| ID | Question | What would settle it |
|---|---|---|
| T18a | (25) ʼashkii yóldon is read ‘The boy is shooting at it’. What excludes ‘He is shooting at the boy’? If nothing in the verb's slots does, the link is pragmatic, and one of Willie's override cases is not schema-driven. | A speaker's judgment on the object reading; other sources for yóldon (Y&M 1987). OPEN |
| T18b | *(33c): a lone noun cannot be the subject of PP + transitive verb even when it suits ρA. In (28), with PP + intransitive verb, it can (‘The boy sat down on it’), although p. 63 says the subject reading needs incompatibility with the PP object. Is the difference transitivity? And why does p. 63 illustrate “incompatible” with (27), where the noun is the PP object? | More lone-noun sentences with PP + intransitive verbs; Willie's promised return “in Sec. 3.2” (not found on pp. 67–70). OPEN |
| T18c | Is an override from world knowledge ((32) “horses typically are led”) grammar or processing? Dowty's n. 16 test: does the default reading come back in a context where a horse could be the beneficiary? | Speaker judgments in context; texts. OPEN |
| T18d | Does MIRROR-LINK hold in the inverse? On the inverse page, INV-6 links the noun next to the verb to the internal slot, which under bi- is ρA. In (86) ʼashkii ʼatʼééd łį́į́ʼ bá yoolóós ‘The girl is leading the horse for the boy’, the order is PP object, subject, object (inverse §2.5). Is that nesting with the internal/external labels swapped, or a different pattern? | Where bi- sits in the template; Willie §3.3 re-read with this question (pp. 94–96). OPEN Partly answered 2026-10-13: no. In Willie's Ch. 8 (13b) John háísh bichʼįʼ yáłtiʼ the noun next to the inverse postposition is the speaker, and the outer noun goes with the postposition's object (wh-questions §7, MIRROR-LINK′). |
| T18e | (1) nihí yiicha is glossed “we 2d.crying” and translated ‘We …’. Ch. 7 (2d)–(2e) give nihi- as both “1-dual” and “2-dual”. Is “2d” a misprint for 1d, or a reading ‘you two’ that the translation does not show? | Y&M paradigms for ‘cry’ (dual); NLA Verb Modes (dictionary: cry). OPEN |
| T18f | ‘ruin’ with an agent is printed with plain l in (8), (16), (18) and with ł in (19), (21) and p. 39 (26). Misprint, or a real difference between the 1st-person and 3/3 forms? Also, (14) is glossed “3s.ruin PERF” on p. 58 but “3-3-ruined” on p. 60, a transitive gloss for an intransitive verb. | Y&M 1987 for -chxǫʼ. OPEN |
| T18g | MVS-UNIT predicts that no noun can stand between a grammatical PP and its verb. Is *N [P-obj-P] N [V] (a noun inside the MVS) actually starred? (44a) says the MVS “cannot be interrupted”, but no starred example is printed on pp. 51–70. | A starred example (Willie elsewhere; W&J 2000); a speaker. OPEN |
| T18h | *(10) bans two free pronouns even though each matches a slot. Is the limit one contrastive focus per clause, tied to W&J's focus position “immediately preceding the verb” (2000:268)? In (8) the free pronoun shí is not immediately before the verb. | Free pronouns in other positions; W&J 2000 §11 re-read. OPEN |
Fidelity notes. Forms are as printed. (36) łííʼ vs (85) łį́į́ʼ; (34) ʼazdzą́ą́ vs (35) ʼasdzáá; plain l vs ł in the ‘ruin’ forms; (14)'s two glosses. (4) has “3 EMPH” without a period, while (3) and (5) have “3.EMPH”. Willie's “A adjoined nominal” (2), “interpretation to all adjoined nominals” (p. 54), “expected work order” (p. 62), “single argment” (p. 69), “pronouns, like” spelled “proncuns” (p. 70, a scan artifact or misprint), “Sec. 3.2. below” (p. 67, written inside §3.2) and the unclosed parenthesis after “horse"” (p. 69) are all as printed. (7) has no gloss line, and its two columns are not labelled. We read them as singular | non-singular but do not print that as Willie's. Printed page 51 occurs twice in the scan (§2).
