Prefix order and where d sits

Legacy link recovery
Legacy section links (scroll sideways); old articles remain available.
subject-headreflexive-reciprocalda-reachpl-heightpl-move-reasonnumber-scopesubject-d-sitevoice-allomorphydeictic-orderspace-dqualifier-d-orderdistr-future-dfuture-d-doo

On the spine (2026-11-14 summary)

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

  • HYPOTHESIS The verb word is the V+v+T complex formed by V→v→T. The plural da- is a head (Dist⁰, a placeholder) merged right above vP that agrees with the closest argument it c-commands, and the verb passes through it on the way to T (H21a): *NX-273. Results §7.
  • GRANT'S HYPOTHESIS The order of prefixes inside the word (restated 2026-11-15). Headedness is fixed locally by Merge site, and the word is linearized opposite to the clause, so the word is predicted to mirror the clause, with a higher head further left (Grant’s Merge-site hypothesis; this is the Mirror Principle’s order for prefixes, and sourced orders that go the other way are misfits, fact-check row 14). Fit against the sourced template is OPEN: the conjunct core fits; aspect, object/subject, distributive and preverb orders do not yet (unified structure §3; headedness). The prefix-order hypotheses in the sections below are tested against that statement, not results.
  • OPEN Light d was withdrawn by Grant on 2026-11-14: a silent determiner on bare nouns did no work in thirteen reviews. Whether the noun phrase has a light n, parallel to light v, is an open structural question, not a premise (results §13). The verbal d discussed below is a separate question.

Why ordinary subject prefixes sit on the tense–aspect head, and the fourth-person, indefinite and areal subjects do not: one feature bundle, two places to pronounce it (T35c, T36g)

2026-11-11 audit pass 2. Apostrophe fixes only. Navajo glottal stops typed as ' or ’ (for example in prefix names like ʼa- and ʼeʼe-, or words like Awééʼ) now print as ʼ (U+02BC), and a build step no longer turns the closing quote of an English translation inside a Navajo row into a glottal stop (‘…the girls.’, not ‘…the girls.ʼ). No form, gloss or translation wording changed.

Working note. Kept as written (URL, data and status labels unchanged). Filed under The word as a spell-out domain; the consolidated argument is the essay word-edge. Rule IDs: rules index.

Verbs on this page → dictionary: yiztał ‘kick’ · yishxash ‘bite’ · nishhozh ‘tickle’ · joolyé ‘be called (named)’ · yiʼash ‘walk (two)’

Restated on the unified structure (2026-11-15). The fourth person is restated as a person value on an argument merged in Spec,vP or VP, with nothing extra moving: the optative fourth bound by the wisher (H51), the respect/report/story ji- as an Agree mark (H57a, OPEN against H57b), the areal ho- as a personless area argument (H54a, H60). Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). Flat-S and other pre-VPISH notation below is labelled contrast. Derivations below in which the verb or subject does not raise are labelled contrast; the working default raises V→v→T and the subject to Spec,TP. Status on the spine: see unified structure, §4, fourth-person rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

2026-11-11 audit pass 3. A rule block written with Hale 2003’s M/S head was not marked as older notation; it is now labeled as a historical record. See the pass 3 report.

Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference.

On Willie's framework §5, rule WF-6 (SUBJ-HEAD) said that every subject prefix is pronounced on a head (Mode/Subject, M/S) and every object prefix is not. With the site's spell-out rule (CYC) that gives Willie's “the patient always to the left of the agent” (p. 50). But WF-6 was a stipulation, and it listed the fourth-person ji- and the agentive-passive ʼadi- among the head exponents. Task T35c asked for a derivation and a check of ji- and ho-/ha-. Task T36g, opened on reflexive vs reciprocal RD4, asked whether the deictic subject is on a head above DETR or an adjunct. This run answers both, from Hale 2001's section on the deictic subjects, Hale 1997's earlier treatment of the same prefixes, Hale 2003's condition on the Mode/Subject portmanteau, and Willie's fourth-person forms. All were read this run (Hale 2001 and Willie on page images with 300 dpi crops). Written 2026-10-18, 8:46pm run (PT).

Ground rules (Grant's). Hale 1997, 2001 and 2003, Willie 1991, Kibrik 2005 and this site's own rules are each treated as possibly partly right. A template position or a gloss label is an analysis, not data. Only contradicting data count against a hypothesis, and silence counts against nothing. Grant does not speak Navajo. Every Navajo form below is copied from an opened source with page and example number. Forms are segmented only where the source prints a segmentation. Two Hale 1997 forms printed in a damaged font are described, not reproduced. No form, gloss or judgment is invented. Schematic letters are used where a test needs a form no opened source prints. Dowty's proto-roles are the baseline, and “role in the action” names a participant's Dowty entailments. Every rule name coined here is ours and a HYPOTHESIS.

Notation. This is the site's convention, since no conventions file has arrived from the Syntax · Semantics · Interface bot (its BRIEF.md and VERB-QUEUE.md are unchanged since Oct 16). Σ = role-schema step. A = word level (cycle D1 = the MVS). B = clause level (D2 and up). Spine as on the clause spine §4 and reflexive vs reciprocal §5: V < v < M/S (= T) < Qual < (DETR) inside D1, then Foc < Top < C. ρ = a pronoun slot ⟨Eρ, posρ, φρ⟩ (proto-role schema); this page adds an optional fourth part, δρ (§5). ρA is first-merged in Spec,v (VPISH) and ρX in Comp,V. CYC = A/B as spell-out §3: C1 non-heads before heads; C2 non-heads higher-first; C3 heads higher-first.

Bottom line, in plain words

SOURCED The facts. The ordinary subject prefixes (1st, 2nd and 3rd person) are pronounced inside the Mode/Subject piece, to the right of the qualifier: Hale 2001 (24a) shidoołhosh ‘He/She will tickle me’ has the object sh, then the qualifier d, then w “3rd:Perf”. The three “deictic” subjects (Young & Morgan's position V: fourth person j(i)-, indefinite ’-, areal ho-) are pronounced to the right of the object but to the left of the qualifier: (22b) shiʼdooʼniid ‘I was told (lit. someone told me)’ is sh ’ d w d niid, and (22c) hodildóóh ‘It (e.g., cloud) starts to drift’ is hw d Ø l dóóh. And with a fourth-person subject a third-person object has no prefix (Willie (23b) jishxash ‘That person bit it’), while with a noun subject it has yi- (Willie p. 129 ʼashkii yishxash ‘The boy bit it’).

HYPOTHESIS Why ordinary subjects sit on the head. Hale 2003 states the condition himself: the subject joins Mode in one portmanteau whenever Mode “locally c-command[s] and govern[s] the pronoun”, and this is “obligatory” (n. 11, pp. 67–68). Under VPISH the subject slot is first-merged in Spec,v, which is exactly the position Mode governs (the specifier of its complement). No other slot is in that position. So the subject's person and number are pronounced on the head and every other slot is pronounced on itself. With the site's rule that non-heads come before heads, the object then comes first. WF-6 is no longer a stipulation. It follows from where VPISH puts the subject and from Hale's own locality condition (rule PORT, §5).

HYPOTHESIS Why the deictic subjects are elsewhere: one bundle, two exponents. A fourth-person, indefinite or areal subject slot carries two kinds of feature: ordinary person/number (which for these is the zero “third-person” value, as both Hale 1997 and Hale 2001 say) and a deictic feature (human, indefinite, or areal). The Mode portmanteau takes only person and number. The deictic feature is left over and has to be pronounced on its own, attached to the highest head of the verb word (the qualifier, or the reflexive/passive di- when present). That is where Hale put it twice: “a prefix to E” (1997) and “prefixes to the Qualifier element” (2001). With that, the site's spell-out rule derives the position V order with no exception, and Hale's “infixation … to the deictic prefixes alone” is no longer needed as a special rule. It also predicts something nobody stated: only subjects can have a separate deictic position, because only the subject slot has part of its bundle taken by a head. Fourth-person objects (ho-/ha-) keep their whole bundle and sit in the ordinary object position, as Willie's table (1) and (31f) hojishxash ‘That person bit that person’ show.

HYPOTHESIS Grant's feature-bundle idea gets a direct case. The fourth-person subject behaves like a third person for the portmanteau (zero agreement) and unlike one for the object prefix (no yi-). Hale 1997 n. 17 says exactly this: it “shares zero-AGRs with the third person” but “does not behave like a third person in obviation”. One slot, two feature groups, two pronunciation sites. That is easier to state if the verb's argument is a feature bundle than if it is a word.

OPEN What is still open. (1) Is the zero person value under ji- really Hale's third person? His test is the rounded [w] allomorph; Willie's joolyé ‘4.called’ and jooʼash ‘4s.walking’ show a rounded vowel after j-, but no opened source segments them (T37a). (2) A one-place verb whose single participant starts low (Hale's labile intransitives, (22) of Hale 2003) is not in the position Mode governs. Either it moves up inside the word (VPISH-type raising) or the locality condition is looser (T37b). (3) Hale's undated handout on the conjunct verb system lists the indefinite subject with 1st, 2nd and 4th as giving a zero 3rd-person object, but Kibrik's agentive-passive forms (from Young & Morgan) show bi- (T37c).

On this page: 1 Sources · 2 The problem · 3 Data · 4 What the analyses claim · 5 Formal rules · 6 Five analyses tested · 7 Derivations SD1–SD10 · 8 What changes on the site · 9 Gets right, misses · 10 Grant's thesis, Dowty, event structure, STRATA, the changeover · 11 Tasks T37a–h

1. Sources opened this run (2026-10-18, 8:46pm PT)

SourceWhat was read, and how
Hale, Ken 2001. “Navajo verb stem position and the bipartite structure of the Navajo conjunct sector.” Linguistic Inquiry 32: 678–693 (box [local working copy, not in the repository])Newly read on page images: §4 “The Deictic Subject Prefixes” and §5, pp. 689–692 (PDF pp. 12–15 at 110 dpi). 300 dpi crops of (22a–c), (23), (24a, b). The text layer drops the barred ł (it prints “shijihozh”), so forms were read from the images only.
Hale, Ken 1997. “Remarks on the syntax of the Navajo verb”, Parts I–III, NLA workshop notes (box [local working copy, not in the repository])Newly read on page images: §4 “The deictic subject prefixes”, PDF pp. 26–29, trees (17)–(19), nn. 16–18. 400 dpi crops of (16) and (20): both printed in a damaged font (“Shizhní¬ʼ²í”, “shidiʼnílʼ²í”), so they are not reproduced; only Hale's prose and tree labels are used. Re-read: p. 2 (the prefix order) and n. 9 (PDF p. 16), text layer of the born-digital PDF.
Hale, Ken 2003 ms. “On the significance of Eloise Jelinek's Pronominal Argument Hypothesis” (box copy hale-pa.pdf, born-digital)Newly read for this question: n. 11 (pp. 67–68), text layer; pp. 10–12 and 26–29 re-read (the portmanteau, (8), (20)–(23)).
Hale, Ken. “The Phonological Aspect of the Navajo Conjunct Verb System.” Handout, MIT (hale062.pdf; a scan with an OCR layer)Read on a page image: p. 9, (21) “Spellout (informal)” and (22) “Third Person”; p. 1, (2) the template list (text layer, checked on the image).
Hale & Keyser, “The basic elements of argument structure” (ms., hale012)Re-read, text layer of the born-digital PDF, PDF p. 16: what a head “locally c-commands (i.e., the specifier of its complement)”.
Willie 1991 (scan on the box; PDF = printed + 6)Re-read on page images, new 300 dpi crops: p. 124 (23a, b); p. 128 (30b), (31b–d); p. 129 (31e, f) and the repeated (28), (30) with third-person subjects; p. 130 (“permitted in context”); p. 113 (8) and the “deictic pronominal” sentence; p. 115 (11)–(12). Table (1), p. 28, and pp. 36, 125 used as recorded on the site.
Kibrik 2005 (box [local working copy, not in the repository])Re-read on page images: chart (6) positions 2–10 and (7a, b), (8a, b), (9a) (PDF pp. 17–19; new 300 dpi crops of (7a, b)).
Not openedYoung & Morgan 1987 (cited for every template label and for hodildóóh, p. 450); Rice 2000 (Hale 2001 reports her argument that position V lacks person); Speas 1990; Hale 2000; McDonough 2000; Neundorf 2000; Golla 1970; Kari 1989. Nothing is attributed to them beyond what the opened sources report.

2. The problem

WF-6 as written on Willie's framework §5: “Every SUBJECT-slot exponent is pronounced on a HEAD of D1 (M/S): -sh-, -ni-, -ii(d)-, -oh-, -0-, ji- (table (1)) and the agentive-passive ʼadi- (p. 36). Every OBJECT-slot exponent is a NON-HEAD of D1.” Three things were wrong or missing.

  1. No derivation. Why should subject φ be on a head and object φ not? The page itself said “WF-6 is a stipulation” (T35c).
  2. The deictic subjects are not where M/S is. If ji- and ʼadi- were exponents of M/S, CYC (C3) would put them after the qualifier, since Qual c-commands M/S. Hale 2001 (22b, c) put them before it. Reflexive vs reciprocal RD4 already found the 4th-person zh- before DETR in Kibrik (8a) and left two options open (T36g).
  3. The passive was split. Kibrik segments the agentive passive as indefinite subject (9) + DETR (8) (RR-5), so ʼadi- is not one subject exponent.

3. Data SOURCED

3.1 Hale 2001: the deictic subjects, a noun subject, and an overt third person in M/S

Hale prints a segment line and a category line under each form. Both are copied exactly. Our Leipzig line only relabels his categories: Object → OBJ with the person from the translation; Y&M's 3a (“also … termed the fourth person”), 3i and 3s → 4.SBJ, INDF.SBJ, AREAL.SBJ; Mode/Subj → M/S; Voice → VOICE; Qualifier → QUAL; 3rd:Perf → 3.PFV. The stem gloss comes from the translation.

Ex., pageForm (as printed)Hale's segments / categoriesLeipzig (ours, relabelled)Translation (Hale's)
(22a), p. 689shijiłhozhsh j Ø ł ghozh / Object 3a Mode/Subj Voice V1SG.OBJ-4.SBJ-M/S-VOICE-tickle‘He/She/They/One is tickling me.’
(22b), p. 690shiʼdooʼniidsh ’ d w d niid / Object 3i Qualifier Mode/Subj Voice V1SG.OBJ-INDF.SBJ-QUAL-M/S-VOICE-tell‘I was told (lit. someone told me).’
(22c), p. 690hodildóóhhw d Ø l dóóh / 3s Qualifier Mode/Subj Voice VAREAL.SBJ-QUAL-M/S-VOICE-drift‘It (e.g., cloud) starts to drift.’
(23), p. 690Ashkii shiłhozh.ashkii sh Ø ł ghozh / boy Object Mode/Subj Voice Vboy 1SG.OBJ-M/S-VOICE-tickle‘The boy is tickling me.’
(24a), p. 691shidoołhoshsh d w ł ghosh / Object Qualifier 3rd:Perf Voice V1SG.OBJ-QUAL-3.PFV-VOICE-tickle‘He/She will tickle me.’
(24b), p. 691shoołtééłsh w ł tééł / Object 3rd:Perf Voice V1SG.OBJ-3.PFV-VOICE-carry‘He/She is carrying me along.’

Hale's prose with these forms (pp. 689–692): position V “is to the right of Object and to the left of Qualifier (with some phonologically motivated exceptions)”. The 3a prefix is “simply a human subject marker construed with the third person in Mode/Subj”; 3i “is an indefinite subject marker, also construed with the third person”, and “in conjunction with the qualifier [d], is the grammatical subject in the agentive passive”; 3s is the “‘spatial’ or ‘areal’ subject” (see “the entry for hodildóóh in Young and Morgan 1987:450 and other relevant entries alphabetized with that verb”). “These elements are problematic precisely because they are subject prefixes. But they appear in the wrong position for subjects.” Following Rice, “the position V elements are not invested with a person feature. They are third person only in the sense that a full DP or NP is third person, like ashkii ‘boy’ in (23).” The rounded [w] “appears only in the third person”, so “the actual position of the third person, like that of local person subjects, is within the Mode/Subj portmanteau.” The deictic subjects “are adjuncts in the abstract representations of clauses, but they are spelled out as prefixes to the conjunct sector of the verb (or, perhaps more narrowly, as prefixes to the Qualifier element, as suggested in Hale 2000)”, by an “infixation” process restricted “(tentatively) to the deictic prefixes alone”. Full DPs, in contrast, “are spelled out as adjuncts to the verbal clause as a whole, and they assume there a linear ordering determined by discourse factors”.

Fidelity: Hale's straight apostrophe is written ’. His hw and sh are his segment spellings, not the surface strings. The final -ghozh/-hozh and -ghosh/-hosh alternations are as printed. Translations as printed, including ‘He/She/They/One’.

3.2 Hale 1997: the same prefixes, an earlier analysis

Prose and tree labels only; the two Navajo forms are in a damaged font and are not reproduced. (16) ‘One looks at me.’ shows “j- in its fricative alternant [zh-]” between AGRo and E (tree (17): AGRo sh-, E ní-, DP j-). (20) ‘I am being looked at.’ / ‘Someone is looking at me.’ is the agentive passive.

  • p. 2: “Preceding this position [E], a so-called “deictic” subject marker may appear—e.g., the indefinite ’- (generally cited as ʼa- in Young and Morgan, 1987, etc.), and the renowned Navajo fourth person j- (ji-). And to the left of this is the position in which object agreement appears (AGRo).”
  • p. 26: “the standard agreement for subjects, AGRs in M, is zero when the subject is a deictic, just as when the subject is third person.” In its impersonal use, j- “is in reality a DP occupying syntactic subject position at d-structure”. n. 16: with an overt nominal (stories, deference) “The use with overt nominals is a form of agreement and must be analyzed differently—e.g., perhaps with j- base-generated as an adjunct to E”.
  • p. 27, n. 17: “The fourth person shares zero-AGRs with the third person. It does not behave like a third person in obviation, however—only the “true” third person (that realized pronominally as bí, for example) requires that a clausemate object in its c-command domain show obviative (y-) agreement.”
  • p. 28: j- “appears as a prefix to E, i.e., to the head of the E-projection … If j- has the lexical property that it is a prefix to E, it will necessarily assume that position”; the move happens “in the morphological component”. n. 18: “a DP, being a phrase, cannot be adjoined to a head. But the DP in this case is also a head”.
  • p. 29: in the agentive passive the subject morphology ʼ-d- “is prefixed to E, i.e., to the head of EP, in the same manner as the fourth person impersonal j-”, and “the lexical object remains an object … construed with AGRo sh-”.
  • n. 9 (PDF p. 16): “If a third person object appears in the c-command domain of a first, second, or fourth person subject … If the third person object morphology appears in E, then it is realized as Ø; otherwise, it is realized as b-.” (Already the basis of WF-7.)

Hale's E is the qualifier/aspect position (Young & Morgan VI; “edge”, p. 2). His 1997 trees use d-structure and s-structure with raising and head movement. They are transitional (not flat-S) and are restated in §7, not used as evidence about where nouns attach.

3.3 Hale 2003 and the handout: the portmanteau condition

  • Hale 2003 pp. 10–11: “mode and the subject are assembled into a portmanteau (Mode/Subj …)”. n. 11 (pp. 67–68), on the intransitive of a labile verb, whose argument starts as the specifier of V: “the pronominal subject of the intransitive must be close enough to Mode to enter into the portmanteau relation with it … if the pronominal subject were raised to [Spec,v], the required portmanteau formation would be a perfectly local operation, since Mode would then locally c-command and govern the pronoun … Portmanteau formation seems to be obligatory (under the relevant condition, i.e., that Mode locally c-command and govern the pronoun) … On the other hand, the structure of the causative shows that the external argument (subject, surface causee object) in the embedded clause is a specifier, and not inherently a part of M/S.” Whether the intransitive's argument raises: “I do not know the answer”.
  • Hale & Keyser (hale012, PDF p. 16): a head assigns case to “the DP which it locally c-commands (i.e., the specifier of its complement)”.
  • Handout (hale062) p. 9: “(21) Spellout (informal): … (b) M and D (subject) fuse, forming a portmanteau-like unit.” “(22) Third Person: Third person subject is regularly Ø. Third person object (of verb) is also Ø if the subject is 1st, 2nd, 4th, or indefinite person. Otherwise a third person object is overt, i.e., if: (i) the subject is third person; (ii) the third person object in question is the object of postposition (P).” Its tree (23c) on the same page draws the 1st-person subject iid as a head D (VIII) between M and v.

3.4 Willie 1991: the object prefix under a fourth-person subject and under a noun

Willie's gloss order is object.subject.stem. Where she gives no gloss line, we add none. Her ' is written ’, and her barred ł as ł. Forms are not segmented.

Ex., pageForm (as printed)Willie's glossLeipzig (ours, relabelled)Translation (Willie's)
(23a), p. 124hashxash4.3.bit4.OBJ.3.SBJ.biteIt bit that person.
(23b), p. 124jishxash3.4.bit3.OBJ.4.SBJ.biteThat person bit it.
(30b), p. 128jiztał——That person kicked it.
(31b), p. 128ʼashkii mósí yishxash——The boy bit the cat.
(31d), p. 128mósí hashxash——The cat bit that person.
(31e), p. 129mosi jishxash——That person bit the cat.
(31f), p. 129hojishxash——That person bit that person.
(28), p. 129 (repeated number)ʼashkii yishxash——The boy bit it.
(30), p. 129 (repeated number)ʼashkii yiztał——The boy kicked it.
(8), p. 113shimá ʼéí Baaʼ joolyé1.mother 4.called1.POSS.mother (ʼéí, Baa’ unglossed) 4.SBJ.calledMy mother is called Baa’.
(11), p. 115 (one word)jooʼash=go4s.walking=when4SG.SBJ.walking=when(part of ‘some boy was walking with a girl when …’)

On (28) and (30), p. 129: they are Willie's third-person alternatives to the fourth-person replies (28b) haztał and (30b) jiztał. They “show incorrect word order, according to the principles of the voice alternation and the animacy hierarchy” (the agent noun stands where patient nouns normally do), but “this reading of these sentences is permitted in context” (p. 130). They are used here only for the shape of the object prefix, which the word-order problem does not affect. (31e) prints mosi without the accents of (31b, d). In (8) Willie's gloss line leaves ʼéí and Baa’ unglossed, and so does ours. Also p. 113: “fourth person is a deictic pronominal as stated in Young & Morgan (1987).” And p. 125 (as recorded on fourth person §3.3): with a 1st, 2nd or 4th argument present, “the theta role of the 1, 2 or 4 argument is clearly marked in the pronoun incorporated into the verb.”

3.5 Kibrik 2005 (as on reflexive vs reciprocal, re-read)

Chart (6): 10 “Accusative [direct object] pronoun”; 9 “3 person Nominative [deictic subject] pronoun”; 8 “transitivity decrease (Navajo, Hupa)”; 7 and 5 “qualifier”; 3 “mode”; 2 “1/2 person Nominative [subject] pronoun”. Forms with his segmentation and position numbers: (7a) táá-bí-ʼ-di-s-d-giz Pref-3.Acc-Indef.Nom(9)-TD(8)-Conj-TI↓-wring ‘He was washed’; (7b) bi-di-ʼ-ní-l-ʼį́ 3.Acc-TD(8)-Indef.Nom(9)-Pref-TI↑↓-see ‘He is looked at’; (8a) ʼá-zh-dí-l-zhééh Refl.Acc(15)-4.Nom(9)-TD(8)-TI↑↓-mow ‘He (4th person) shaves himself’. Kibrik's raised glottal stop is written ’. A Hupa form (9a) (not Navajo, Golla 1970) has the third-person animate subject prefix before the first-person object prefix, and (9b) after the second-person object; Kibrik calls this “variable order of the 3rd person Nominative and the Accusative pronouns in Hupa … and Ahtna”.

4. What the analyses claim, side by side

QuestionWillie 1991Hale 1997Hale 2001Hale 2003 / handoutKibrik 2005
Is ji- a subject pronoun?Yes: subject column of table (1), with -sh-, -ni-, -0-Impersonal use: the subject DP itself; with a noun: an adjunct to E, “a form of agreement”No: an adjunct nominal with no person feature, linked to the true (3rd-person) subject in M/S—Yes: “3 person Nominative [deictic subject] pronoun”
Where is it pronounced?Not stated beyond “subject” and “innermost” (p. 27)Prefixed to E (morphological movement)Infixed: prefix to the conjunct sector or to the Qualifier—Position 9, above 8 and the qualifiers; 1/2 subjects at 2
Ordinary subject φA pronoun in an A positionAGRs in MIn the Mode/Subj portmanteauM/S portmanteau, obligatory when Mode locally c-commands and governs the pronounPosition 2 (1/2 only)
3rd object under ji-Prints Ø forms ((23b), (30b), (31e))Ø (n. 9); 4th is not a “true” 3rd for obviation (n. 17)—Ø (handout (22))—
3rd object under the indefinite—n. 9 lists only 1st, 2nd, 4th—Ø (handout (22))bí-/bi- in the agentive passive ((7a, b))

5. Formal rules (ours; all HYPOTHESIS unless marked)

OPEN Older notation: the block below uses Hale 2003’s M/S (Mode/Subject) head, which is not a node in our current trees. It is kept as a historical record, not as one of our analyses.

SH-0  SLOT+   A slot is rho = <E, pos, phi, delta>. phi = person/number. delta is optional and is a
      (bundle) DEICTIC feature: delta in {4 (human), INDF, AREAL} (Y&M 3a, 3i, 3s as reported by
               Hale 2001). A slot with delta has phi = the zero ("third-person") value for agreement
               (SOURCED: Hale 1997 p. 26 "AGRs in M, is zero when the subject is a deictic";
               Hale 2001 pp. 690-691). Splitting the deictic off as delta is ours.

SH-1  LOC     A head H locally c-commands and governs XP iff XP is H's complement or the specifier
               of H's complement. (SOURCED wording: Hale & Keyser hale012 PDF p. 16; Hale 2003 n. 11.)

SH-2  PORT    At spell-out of D1, the phi of the slot that M/S locally c-commands and governs is
               pronounced INSIDE the head M/S (the Mode/Subject portmanteau). Obligatory.
               (SOURCED condition: Hale 2003 n. 11, pp. 67-68; handout (21b). Restricting it to phi,
               so that delta is not taken, is ours.)

SH-3  SUBJ-HEAD' (theorem, replaces WF-6). Under VPISH, rho_A is first-merged in Spec,v, the
               specifier of M/S's complement. So rho_A's phi is pronounced on a HEAD. Every other
               slot (rho_X in Comp,V, an oblique, a causee in a lower Spec,v) is not governed by M/S
               and is pronounced on itself, as a NON-HEAD. With CYC C1: every object exponent
               precedes M/S. No per-morpheme list.

SH-4  DELTA-HOST  A delta left over by PORT (that is, the delta of the slot whose phi M/S took) is
               pronounced as its own element, LEFT-ADJOINED to the highest nuclear head of D1:
               DETR if present, else Qual, else M/S. A head-adjunct counts as part of H(D1), and
               within a complex head the adjunct precedes its host (C3+).
               (SOURCED: Hale 1997 p. 28 "a prefix to E"; Hale 2001 p. 691 "prefixes to the
               Qualifier element". "Highest nuclear head" generalizes them to DETR, ours.)

SH-5  NO-SPLIT  A slot whose phi no head takes keeps its whole bundle (phi + delta) in one exponent.
               Corollary: only a SUBJECT can have a separate deictic position. A 4th-person object
               (ho-/ha-) is an ordinary object exponent. (Derived from SH-2 + SH-4.)

SH-6  3-OBJ'  (refines WF-7) A 3rd-person rho_X pronounced in D1 is
                 y-  if c-commanded in D1 by a slot with phi = 3 and NO delta  ("true third person",
                     Hale 1997 n. 17);
                 0   if c-commanded by a slot with phi in {1, 2} or delta = 4, and pronounced in E
                     (Hale 1997 n. 9);
                 b-  otherwise (includes delta = INDF: Kibrik (7a, b)).
               The handout's (22) puts INDF in the 0 group. That clashes with (7a, b) (T37c).

SH-7  CAUS    (corollary of SH-1, SH-2) In [M/S [vP Causer v_caus [vP Causee v VP]]] only the Causer
               is the specifier of M/S's complement. The Causee's phi stays on its slot, a non-head,
               and surfaces with object morphology. (SOURCED as a statement: Hale 2003 n. 11
               "not inherently a part of M/S".)

SH-8  UNACC   (OPEN) A labile intransitive's single slot is first-merged in Spec,V (Hale 2003 (22)),
               not the specifier of M/S's complement (v intervenes). Hale: it "must be close enough
               to Mode". U-1: it raises to Spec,v inside D1 (VPISH-type A-movement). U-2: LOC is read
               as "the closest slot in M/S's c-command domain" (Agree with the closest goal).
               Same strings; different predictions only for movement diagnostics (T37b).
RuleRests onStatus
SH-0Hale 1997 p. 26; Hale 2001 pp. 690–691; Willie table (1)HYPOTHESIS (the zero φ is SOURCED as Hale's claim)
SH-1, SH-2Hale 2003 n. 11; hale012 p. 16; handout (21b)SOURCED condition; the φ-only restriction is HYPOTHESIS
SH-3SH-1, SH-2 + VPISH (ρA in Spec,v); Willie p. 50; (24a, b); Hale 2003 (1)HYPOTHESIS; new; replaces WF-6
SH-4Hale 1997 p. 28; Hale 2001 p. 691; (22a–c); Kibrik (7a), (8a)HYPOTHESIS; new
SH-5derived; Willie table (1); (31f), (23a)HYPOTHESIS; new prediction, fits all printed forms here
SH-6Hale 1997 nn. 9, 17; Willie (23b), (30b), (31b, e), p. 129 (28), (30); Kibrik (7a, b)HYPOTHESIS; handout clash OPEN
SH-7Hale 2003 n. 11HYPOTHESIS (no causative form tested here)
SH-8Hale 2003 (22), n. 11OPEN Update 2026-10-19, 4:46am: U-2 restated as IN-4 PORT-CLOSEST and favoured on economy (not decided): the inchoative bleeds the low slot's subjecthood, and IN-4 covers both yishdlóóh and shiʼniidlóóh with no movement (the ‘start to’ test §5–6). Update 2026-10-19, 12:46pm: in the ‘start to’ unaccusative, Mode's third person is now read as a default. HMP: subject position VIII is “entirely empty … as it is regularly in persons other than first or second”. This is IN-7 (D). Areal ho- does not satisfy the inchoative's object need (HMP (15)), which fits treating it as a deictic subject, not an object (HMP page §6).

6. Five analyses tested against the data

Tests: t1 object before the deictic ((22a, b), (31f), Kibrik (7a)). t2 deictic before the qualifier or DETR ((22b, c), Kibrik (8a)). t3 an ordinary subject after the qualifier and after the object ((24a); Hale 2003 (1) ni-sh-hozh, as recorded). t4 a 3rd-person object is Ø under ji- ((23b), (30b), (31e)) but yi- under a noun-linked 3rd-person subject ((31b), p. 129 (28), (30)). t5 bi- under the indefinite in the agentive passive (Kibrik (7a, b)). t6 no separate deictic object position (Willie table (1); (31f) ho before ji). t7 an overt 3rd person inside M/S ((24a, b) w).

Analysist1t2t3t4t5t6t7
WF-6 as written (ji-, ʼadi- on M/S)✓✗ (C3 puts M/S material after Qual/DETR)✓ (stipulated)✓ if ji- is φ = 4✓ (WF-7 elsewhere case)— (stipulated)✓
Hale 1997 (j- = subject DP, prefixed to E; nn. 9, 17)✓✓✓✓ (n. 17)✓ (n. 9 omits INDF)— (silent)—
Hale 2001 (adjunct nominal, no person, infixed)✓✓✓✗ as stated: an adjunct “third person only in the sense that … ashkii” predicts yi-, as under ʼashkii (p. 129); data: Ø— (silent)— (silent)✓ (his evidence)
Kibrik template (positions 9 vs 2)✓✓✓— (silent)describes——
This page (SH-0 to SH-6 + CYC)✓ C1✓ SH-4✓ SH-3✓ SH-6✓ SH-6✓ derived (SH-5)✓ SH-2

HYPOTHESIS Verdict. WF-6 as written fails t2 on three printed forms ((22b), (22c), Kibrik (8a)), so it is replaced by SH-3 + SH-4. Hale 2001's “third person like ashkii” fails t4 as literally stated: that is contradicting data, not silence. It is repaired by keeping his position claim and his “no person feature” claim but making the deictic a feature of the slot (δ), present inside the word, rather than a noun-like adjunct that arrives with the nouns. That is close to Hale 1997's two-part treatment, minus its movements. The deictic is not on a head above DETR (T36g option (i)), and not a phrasal non-head (option (ii)). It is an adjunct to DETR or Qual, a third option that derives the same string.

7. Derivations SD1–SD10

Format as on Willie's framework §7. Where something merges or is pronounced is ours (HYPOTHESIS). Strings, glosses and segmentations are the sources' (SOURCED). Role labels follow the translations, read Dowty-style (role in the action). “Not segmented further” means the output is compared with the printed string as a whole. V-H (the subject stays in Spec,v) is assumed. V-P gives the same strings, because PORT reads φ, not position.

SD1. (24a) shidoołhosh ‘He/She will tickle me’: an ordinary 3rd-person subject on M/S, after the qualifier

  1. Σ Σ(√TICKLE) = {ρA (causes a sensation, acts volitionally: Proto-Agent), ρX (is affected: Proto-Patient)}. φ(ρA) = 3, no δ; φ(ρX) = 1sg.
  2. A Merge V ghosh with ρX (Comp,V). Merge v (ł); ρA in Spec,v (VPISH). Merge M/S (gamma perfective). Agree M/S ⟶ ρA. Merge Qual d (inceptive; Hale: the future is the gamma perfective “preceded by the inceptive qualifier [d]”).
  3. A PORT (SH-2): ρA is the specifier of M/S's complement, so its φ [3] is spelled inside M/S → w (“3rd:Perf”, Hale's rounded allomorph).
  4. A Spell-out D1 by CYC. N(D1) = {ρX} → sh. H(D1) = {Qual, M/S, v, V}; C3 → d ⌢ w ⌢ ł ⌢ ghosh. C1 → sh ⌢ d ⌢ w ⌢ ł ⌢ ghosh = Hale's segment line for shidoołhosh ✓. The subject exponent follows the qualifier; the object precedes everything.

SD2. Hale 2003 (1) ni-sh-hozh ‘I tickle you’ (as recorded on the site): a 1st-person subject on M/S, after a 2nd-person object

  1. Σ as SD1, with φ(ρA) = 1sg, φ(ρX) = 2sg.
  2. A As SD1 without Qual. PORT: ρA's φ on M/S → sh. ρX is in Comp,V, governed by V, not by M/S (v intervenes), so its φ stays on the slot → ni, a non-head.
  3. A Spell-out D1: C1 → ni ⌢ [M/S sh] ⌢ v ⌢ V = ni-sh-hozh ✓ (Hale's hyphens). This is SH-3 at work: the subject is not listed as a head exponent; it is pronounced on M/S because of where VPISH merges it.

SD3. (22a) shijiłhozh vs (23) Ashkii shiłhozh: the deictic is inside the word, the noun is outside it

  1. Σ Σ(√TICKLE) as SD1. (22a): ρA = ⟨φ = 3 (zero), δ = 4⟩. (23): ρA = ⟨φ = 3, no δ⟩. ρX = 1sg in both.
  2. A Merge, Agree as SD2. PORT: φ(ρA) on M/S → Ø in both (zero imperfective, 3rd person).
  3. A (22a) only: DELTA-HOST (SH-4). δ = 4 is left over. No DETR, no overt Qual, so the highest nuclear head is Qual (silent) or M/S. j left-adjoins to it.
  4. A Spell-out D1. N = {ρX} → sh. H = {[j + Qual/M/S], v, V}. (22a): sh ⌢ j ⌢ Ø ⌢ ł ⌢ ghozh ✓ (Hale's segment line). (23): sh ⌢ Ø ⌢ ł ⌢ ghozh ✓.
  5. B (23) only: Merge ashkii in D2; Link to ρA (φ-match 3). D2 spell-out: ashkii ⌢ [D1] ✓. In (22a) nothing happens at B: the subject's deictic content was pronounced at A.
  6. Contrast. Hale 2001 calls j- and ashkii the same category. On this derivation they differ in when they enter: δ is part of the slot from Σ on and is pronounced in D1 by rule set A; the noun enters in D2 and is ordered by rule set B (discourse). SD6 shows that the difference is visible.

SD4. (22b) shiʼdooʼniid ‘someone told me’: the indefinite δ on the qualifier (the agentive passive)

  1. Σ Σ(√TELL) = {ρA (the teller: volition, causation), ρX (the one told), (what is told)}. ρA = ⟨φ = 3 (zero), δ = INDF⟩; ρX = 1sg. The third participant is not marked in the word and is not discussed by Hale.
  2. A Merge V niid + ρX; v = d (Hale n. 6: the agentive passive “requires the d-classifier”); ρA in Spec,v; M/S (w, as Hale segments it); Qual d (Hale: “The 3i prefix [’], in conjunction with the qualifier [d], is the grammatical subject”).
  3. A PORT: φ(ρA) on M/S. DELTA-HOST: δ = INDF left-adjoins to the highest nuclear head, Qual d: [’ + d].
  4. A Spell-out D1: N = {ρX} → sh; H = {[’ + d], M/S w, v d, V niid} → sh ⌢ ’ ⌢ d ⌢ w ⌢ d ⌢ niid ✓ (Hale's segment line).
  5. WF-6 as written: ’ (or ʼadi-) on M/S → C3 gives sh ⌢ d ⌢ [’ + w] … ✗ (the deictic would follow the qualifier). Willie's p. 36 “to the right of (internal to) the object pronoun” is kept: [’ + d] is part of H, so it follows every non-head.

SD5. (22c) hodildóóh ‘It (e.g., cloud) starts to drift’: the areal δ

  1. Σ Σ(√DRIFT) = {ρ1 (moves; the translation states no volition)}. ρ1 = ⟨φ = 3 (zero), δ = AREAL⟩. Whether ρ1 starts in Spec,v or Spec,V is not known (SH-8).
  2. A Merge V dóóh, v l, M/S Ø, Qual d. PORT (via U-1 or U-2 if ρ1 starts in Spec,V). DELTA-HOST: hw on Qual.
  3. A Spell-out D1: N = ∅; H = {[hw + d], Ø, l, dóóh} → hw ⌢ d ⌢ Ø ⌢ l ⌢ dóóh ✓.
  4. B A noun such as ‘cloud’ could link to ρ1 in D2 (Hale's “e.g., cloud” is in the translation only). Not tested.

SD6. (23b) jishxash and (31e) mosi jishxash vs p. 129 ʼashkii yishxash and (31b) ʼashkii mósí yishxash: the object prefix sees δ

  1. Σ Σ(√BITE) = {ρA (bites: volition, causation, contact), ρX (bitten: affected)}. φ(ρX) = 3. jishxash: ρA = ⟨3 (zero), δ = 4⟩. ʼashkii yishxash: ρA = ⟨3, no δ⟩.
  2. A Merge, Agree, PORT as SD3. DELTA-HOST for jishxash only.
  3. A Spell-out ρX by 3-OBJ′ (SH-6). jishxash: the c-commanding slot has δ = 4 → not a “true third person” → Ø (if in E). ʼashkii yishxash: the c-commanding slot is φ = 3 with no δ → y-.
  4. A Spell-out D1: j ⌢ … ⌢ shxash vs yi ⌢ … ⌢ shxash; printed jishxash ✓ and yishxash ✓ (not segmented further).
  5. B (31e): mosi links to ρX (3) ✓. (31b) and p. 129: ʼashkii links to ρA (3), mósí to ρX.
  6. Why this matters. The choice between Ø and yi- is made inside the word (D1), and it depends on whether the subject is deictic. A noun linked to the subject (ʼashkii) arrives in D2 and cannot change the object prefix (the site's “B is inert”, WF-5). So the deictic must be present in D1. Hale 2001 is right that it is not a person value, but it cannot be a noun-like adjunct that arrives with the nouns.

SD7. (31f) hojishxash and (23a) hashxash: a 4th-person object keeps its whole bundle

  1. Σ as SD6. (31f): ρA = ⟨3 (zero), δ = 4⟩; ρX = ⟨3 (zero), δ = 4⟩. (23a): ρA = ⟨3, no δ⟩; ρX = ⟨3 (zero), δ = 4⟩.
  2. A PORT takes φ(ρA) only. ρX's φ is taken by no head, so SH-5: its whole bundle is one exponent in the object position (Willie's table (1): object 4 = ho/ha). (31f): ρA's δ is hosted by the top head (SH-4).
  3. A Spell-out D1. (31f): N = {ρX ho}, H = {[ji + Qual/M/S], v, V} → ho ⌢ ji ⌢ … shxash = hojishxash ✓ (not segmented further). (23a): ha ⌢ … shxash = hashxash ✓. The ho/ha alternation is not derived (T8d).
  4. The prediction. There is a separate deictic subject position (V) but no separate deictic object position, because only the subject slot loses its φ to a head. The possessive paradigm fits too: Willie (2i) ha-má ‘4s-mother’ (p. 183, as recorded on possessor raising §3) has the 4th-person possessor in the ordinary possessor slot. No head takes a possessor's φ.

SD8. Kibrik (8a) ʼá-zh-dí-l-zhééh and (7a) táá-bí-ʼ-di-s-d-giz: δ on DETR, and bí- under INDF

  1. A (8a): as RD4, with ρA = ⟨3 (zero), δ = 4⟩. DETR di is the highest nuclear head, so SH-4 puts zh on it: [zh + DETR]. N = {REFLhi ʼá}; H = {[zh + dí], M/S, v l, V}. → ʼá ⌢ zh ⌢ dí ⌢ l ⌢ zhééh ✓. This is a third option for RD4 (not “a head above DETR”, not a phrasal non-head), and it gives the same string.
  2. A (7a): ρA = ⟨3 (zero), δ = INDF⟩; ρX φ = 3. SH-4: [’ + DETR di]. SH-6: δ = INDF is neither a true 3rd nor in the Ø set → b-. N = {Pref táá, ρX bí}; H = {[’ + di], M/S (Kibrik's Conj s), v d, V giz} → táá ⌢ bí ⌢ ’ ⌢ di ⌢ s ⌢ d ⌢ giz ✓.
  3. A (7b) bi-di-ʼ-ní-l-ʼį́: as (7a), then Kibrik's post-CYC metathesis (n. 5; RR-4) before ní → di ⌢ ’ ✓. The metathesis applies to a complex head [’ + di], i.e., inside H; it is local and phonological, as before.

SD9. A causative (schematic letters only): the causee is not part of M/S

  1. Σ A one-place verb V with ρ1, plus a causing layer: Σ′ = {ρC (causer), ρ1 (causee)}.
  2. A Merge [vP ρ1 v [VP … V]]; Merge vcaus (the ł-causer of plain ł-causer) with ρC in its specifier; Merge M/S.
  3. A PORT: only ρC is the specifier of M/S's complement. ρ1 is not (vcaus intervenes) → its φ stays on the slot, a non-head, with object morphology. Hale 2003 n. 11: the causee “is a specifier, and not inherently a part of M/S” ✓ (statement; no causative form is segmented on this page).

SD10. A labile intransitive (schematic): U-1 vs U-2

  1. Σ V with ρ1 in Spec,V (Hale 2003 (22)); v Ø above it.
  2. A U-1: Move ρ1 Spec,V → Spec,v (VPISH-style: the subject starts low and moves up, here inside the word). PORT then applies locally. U-2: no movement; M/S's Agree finds ρ1 as the closest slot, and PORT is read off Agree.
  3. Both give the same string. U-1 is A-movement, which Willie's “no movement to and from A positions” (p. 23) and Hale 2003's “no motivation for syntactic movement” (pp. 29–30) rule out unless that ban is only about nouns (as X1 reads it). This is T35e's question in a new place (T37b).

8. What changes on the site

  • HYPOTHESIS WF-6 is replaced by SH-3 + SH-4. Subject φ is on M/S because of PORT, not by a list. ji-, the indefinite ’- and the areal ho- are not M/S exponents; they are the left-over deictic part of the subject slot, adjoined to the highest head. The passive's ʼadi- is [’ + d/DETR], a complex head (SD4, SD8). T35c is closed with this derivation.
  • HYPOTHESIS T36g is closed: the deictic subject is an adjunct on DETR (or Qual), not a head of its own above DETR and not a phrasal non-head.
  • HYPOTHESIS T29f, deictic half: Hale's “infixation … to the deictic prefixes alone” is no longer an exception to CYC. It is SH-4 (head-adjunction) plus C1/C3. The oblique-object half of T29f (Hale 2003 n. 13) is untouched.
  • HYPOTHESIS WF-7 is refined as SH-6: the y- case needs a “true third person” (φ = 3 and no δ). The Ø list stays 1, 2, 4 (Hale 1997 n. 9); the indefinite falls under b-.

9. What each analysis gets right, and what it misses

AnalysisGets rightMisses or leaves open
Willie 1991Puts ji- with the subjects and ho/ha with the objects (table (1)): right about which role each marks. Groups 4th with 1st and 2nd (“clearly marked in the pronoun”, p. 125) and prints the Ø-object forms ((23b), (30b), (31e)). Calls the fourth person “a deictic pronominal” (p. 113).The table hides the fact that ji- is not pronounced where -sh-/-ni- are. Her framework has no statement of prefix order beyond patient-before-agent, so the position V puzzle does not arise for her.
Hale 1997Prefix to E (= our SH-4 with Qual as host). Zero subject agreement with deictics. n. 17: 4th is not a “true” third person for obviation (our SH-6). Treats the passive's ʼ-d- like j-.Needs d-structure raising (j- to Spec,EP for the EPP, the object to Spec,MP), head movement and a morphological move. Two analyses for one prefix (impersonal vs with a noun). These are transitional GB devices; the restatement here needs none of them.
Hale 2001The position (right of Object, left of Qualifier). No person feature on position V (after Rice). The overt [w] argument that the real 3rd person is in M/S. “Prefixes to the Qualifier element.”“Third person only in the sense that a full DP … like ashkii” predicts yi- under ji-, and Willie prints Ø (SD6). Infixation is restricted to the deictics by stipulation.
Hale 2003 / handoutThe portmanteau condition (local c-command and government; obligatory). The causee is not part of M/S. The handout's tree draws the subject as a head under M.Leaves the labile intransitive open (n. 11). The handout's (22) puts the indefinite in the Ø group, against Kibrik (7a, b) (T37c).
Kibrik 2005Positions 9 (deictic subject) vs 2 (1/2 subject) and the 9/8 metathesis. Notes that Hupa and Ahtna vary the order of the deictic subject and the object.A template, not a derivation. Calls position 9 a “3 person Nominative pronoun”, which the Ø-object facts and Hale's [w] argument do not fit as stated.
This site before this run (WF-6, WF-7, RD4)Saw that subject exponents behave as heads for order, and wrote the Ø/b-/y- rule from Hale 1997 n. 9.Listed ji- and ʼadi- as M/S exponents (fails (22b, c), (8a)). No derivation of WF-6. RD4 left the deictic's status open.
This page (SH-0 to SH-8)Derives SUBJ-HEAD from Hale's locality and VPISH. Derives position V with no exception. Predicts no deictic object position. Handles Ø vs yi- vs bi- with one feature.The δ host (“highest nuclear head”) is chosen to fit, following Hale. Whether φ under ji- is really 3rd (T37a), the labile intransitive (T37b) and the indefinite with a non-passive verb (T37c) are untested. The Hupa variation is not explained (T37f).

10. Grant's thesis, Dowty, event structure, STRATA, and the changeover

GRANT'S CONJECTURE → HYPOTHESIS Weak thesis (verbs take role-bearing slots; nouns are optional add-ons linked to them). Supported, and sharpened. The subject slot is pronounced in two places: its person/number inside the Mode/Subject head, its deictic feature on the highest head. The noun ʼashkii is linked to the same slot only at the clause level and does not affect the word (SD3, SD6). That is a slot-with-a-feature-bundle picture, not a word-as-argument picture. Nothing on this page contradicts the weak thesis. Strong thesis: not tested here.

HYPOTHESIS Feature bundles (Grant's idea). The fourth-person subject is the clearest case so far of a bundle whose parts behave differently: zero agreement like a 3rd person (Hale 1997 p. 26), but not a “true” 3rd person for the object prefix (n. 17; Willie (23b)). A word-sized argument would have one person value. A bundle can have φ for one rule and δ for another.

HYPOTHESIS Dowty. PORT puts the person/number of the slot in Spec,v on the head. Under VPISH that is the slot whose role in the action has the most Proto-Agent entailments in a transitive (tickle, bite, tell), which is Dowty's subject selection, the mainstream baseline. Navajo adds two things, both extensions rather than departures: a structural statement of where the selected subject is pronounced (on the tense–aspect head), and a deictic feature (human, indefinite, areal) that is about the referent, not about the role in the action.

HYPOTHESIS Theta roles and event structure as compatible descriptions (Grant's stance). SD9 shows them working together. The role schema says which participants there are. The event-structure layer (a causing vcaus) decides which of them is in the position Mode governs. So in a causative the causer, not the causee, is pronounced on M/S, and the causee surfaces as an object. Neither description replaces the other.

HYPOTHESIS STRATA (theta roles first, then event-structure roles, then noun selection). The order fits. δ is part of the slot from the role step on (layer 1). The event-structure heads (v, vcaus, the inceptive qualifier, DETR) must be in place before PORT and DELTA-HOST apply, because PORT needs to know which slot is local to M/S (SD9), and DELTA-HOST needs to know which head is highest (SD4, SD8) (layer 2). Nouns come last, at D2 (layer 3), and cannot change the word (SD6). On Hale 2001's literal view, by contrast, a nominal (the deictic adjunct) would enter inside the word, putting a layer-3 item before layer 2 is spelled out. SD6 favours keeping the deictic in layer 1 as δ. This orders the layers. It does not by itself separate STRATA from the adjunct and inside-the-clause views of where ordinary nouns attach.

HYPOTHESIS The changeover point (research focus 1). The deictic subjects are the best case so far that the A/B line is a cycle, not a category. Hale 2001 says they are “to be identified with adjunct nominal expressions”, yet they are ordered by rule set A (fixed: after the object, before the qualifier), while full nouns are ordered by rule set B, “a linear ordering determined by discourse factors” (p. 692). What decides is whether the material is pronounced in D1, not whether it is noun-like. Inside D1 the left-headed order holds (heads in order Qual/DETR ⌢ M/S ⌢ v ⌢ V, non-heads first); outside it, the finished word is one non-head in a right-headed clause.

HYPOTHESIS Phrase-level packaging (Grant's reminder that lexicalization happens at the phrase level too). Hale reports that Young & Morgan list areal-subject verbs as dictionary entries, with “other relevant entries alphabetized with that verb” (hodildóóh, Y&M 1987:450, not opened). If so, the areal δ is part of a listed unit together with the verb: a lexicalized subject, like English weather it. Lexicalization can include an argument slot's feature, not just a stem and its prefixes.

11. Tasks T37a–h

#TaskNeeds
T37aOPEN Is φ under ji- Hale's 3rd person? Does a 4th-person gamma-perfective, progressive or future show the rounded [w]? Willie (8) joolyé ‘4.called’ and (11) jooʼash ‘4s.walking’ have a rounded vowel after j-, but are not segmentedYoung & Morgan 1987 paradigms (not opened)
T37bOPEN SH-8: does a labile intransitive's single slot raise to Spec,v inside the word (U-1, A-movement) or is PORT just Agree with the closest slot (U-2)? (joins T35e) Partly answered 2026-10-19, 4:46am (the ‘start to’ test): under the inchoative an unaccusative's participant becomes an oblique and Mode shows a third person; U-2 (IN-4) favoured, raising not excluded; decisive tests T38a–c.1st/2nd-person forms of labile intransitives, segmented; a movement diagnostic
T37cOPEN Indefinite subject with a 3rd-person object outside the agentive passive: Ø (handout (22)) or b- (Hale 1997 n. 9's list; Kibrik (7a, b))? Does DETR make the passive special?Young & Morgan; NLA materials
T37dOPEN An areal-subject verb with an object prefix: is the order object ⌢ ho(areal) ⌢ Qual, as SH-4 predicts? How do areal ho- and 4th-object ho- differ in one word?Young & Morgan 1987:450 and neighbouring entries (not opened)
T37eOPEN δ with disjunct adjuncts (da-, iterative, preverbs): does ji- stay right of the object and left of the highest head in every case? Collect da- + ji- formsWillie Ch. 4–6 forms; Young & Morgan
T37fOPEN Hupa and Ahtna put the deictic subject on either side of the object (Kibrik (9)). Is that head-adjunction (Navajo) vs phrasal adjunction, a parameter on SH-4?Golla 1970; Kari 1989:439 (not opened)
T37gOPEN Clean copies of Hale 1997 (16) ‘One looks at me’ and (20) ‘I am being looked at’ (damaged font), to add them to the dataanother copy of hale023, or the same verbs in Young & Morgan
T37hOPEN Restate the fourth-person, fourth-person-uses and pleonastic ho- derivations with δ (SH-0), and check that no earlier derivation relied on ji- being an M/S exponentsite pages only

Sources: Hale 2001 pp. 689–692 (box copy [local working copy, not in the repository], page images PDF pp. 12–15; crops under [local working copy, not in the repository], not committed). Hale 1997 PDF pp. 2, 16, 26–29 ([local working copy, not in the repository]). Hale 2003 ms. pp. 10–12, 26–29, n. 11 pp. 67–68. Handout hale062 p. 9. Hale & Keyser hale012 PDF p. 16. Willie 1991 pp. 113, 115, 124, 128–130 (page images and crops under [local working copy, not in the repository]); table (1) p. 28, pp. 36, 125, 183 as recorded on the site. Kibrik 2005 §3.2, chart (6), (7)–(9). Cross-links: Willie's framework WF-6, WF-7 (T35c), reflexive vs reciprocal RD4 (T36g), A/B as spell-out (T29f), fourth person, plain ł-causer, open questions (T37), sources.

Why the reflexive comes before plural da- and the reciprocal after it: a two-piece reflexive and a one-piece reciprocal (T35b)

2026-11-11 audit pass 2. Apostrophe fixes only. Navajo glottal stops typed as ' or ’ (for example in prefix names like ʼa- and ʼeʼe-, or words like Awééʼ) now print as ʼ (U+02BC), and a build step no longer turns the closing quote of an English translation inside a Navajo row into a glottal stop (‘…the girls.’, not ‘…the girls.ʼ). No form, gloss or translation wording changed.

Working note. Kept as written (URL, data and status labels unchanged). Filed under The word as a spell-out domain; the consolidated argument is the essay word-edge. Rule IDs: rules index.

Verbs on this page → dictionary: yi-sh-cha ‘cry’ · ʼádílzhééh ‘shave (reflexive: shave oneself; agentive passive: be shaved)’ · ʼádíshdleesh ‘paint (reflexive: paint oneself; agentive passive: be painted)’

Restated on the unified structure (2026-11-15). Prefix order is restated as Grant’s hypothesis that headedness is fixed locally by Merge site, so the sublexical domain is linearized opposite to the phrasal domain and the prefixes are successive word-internal Merge and Agree sites (GRANT’S HYPOTHESIS, tested on the sourced template in §3.2; fit/misfit OPEN). A possible sublexical V-to-T is a separate, untested GRANT’S HYPOTHESIS (§3.3). The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). Derivations below in which the verb or subject does not raise are labelled contrast; the working default raises V→v→T and the subject to Spec,TP. Status on the spine: see unified structure, §3. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

2026-11-11 audit pass 3. The A/B headedness changeover is a hypothesis and is not established; sentences that stated it as fact now say so. See the pass 3 report.

Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference.

On Willie's framework, WD6, the site's single spell-out rule (CYC) failed on one pair of Willie 1991's forms. The reflexive comes before the distributive plural da- ((14) ʼádadiidleesh, gloss REF.PL), but the reciprocal comes after it ((18) daʼahiidleesh, gloss DIST.REC). With both anaphors in one object position, CYC puts da- first in both. Three ways out were listed (R1 a two-piece reflexive, R2 a late morphological placement of da-, R3 a separate position for the reciprocal). Task T35b asked which one is right. This run answers it from sources already on the box: Kibrik 2005's segmented reflexives (from Young & Morgan 1987:70–71), Young & Morgan's prefix template as McDonough 2019 reproduces it, Willie's reflexive and passive forms re-read on page images, and Speas 1986's footnote on what the reflexive is. Written 2026-10-18, 12:46pm run (PT).

Ground rules (Grant's). Willie, Kibrik, Young & Morgan (as reported), Speas and this site's own rules are each treated as possibly partly right. A template or a label is an analysis, not data. Only contradicting data count against a hypothesis, and silence counts against nothing. Grant does not speak Navajo. Every Navajo form below was read this run on a page image, with page and example number (300 dpi crops for every form whose diacritics matter). No form, gloss or judgment is invented. Forms are not segmented beyond the source's own hyphens. Where a derivation needs pieces that a source does not print, it names them schematically (REFLhi, DETR, Pl) and only then compares the printed string. Dowty's proto-roles are the baseline, and “role in the action” names a participant's Dowty entailments. Every rule name coined here is ours and a HYPOTHESIS.

Notation. This is the site's convention, since no conventions file has arrived from the Syntax · Semantics · Interface bot. Σ = role-schema step. A = word level (cycle D1 = the MVS). B = clause level (D2 and up). Spine as on the clause spine §4: V < v < M/S (= T) < Qual (= D1 = MVS) < Foc < Top < C. ρ = a pronoun slot ⟨Eρ, posρ, φρ⟩ (proto-role schema). ρA is first-merged in Spec,v and ρX in Comp,V. CYC = A/B as spell-out §3: C1 non-heads before heads; C2 non-heads higher-first; C3 heads higher-first. Disjunct adjuncts (preverb, reversionary, plural) are adjoined to QualP inside D1, as in Hale 2003's tree (31b) (spell-out rules N4).

Bottom line, in plain words

SOURCED The reflexive is two pieces, not one. Kibrik 2005 segments two Young & Morgan reflexives as ʼá-zh-dí-l-zhééh ‘He (4th person) shaves himself’ and ná-ʼá-di-zh-ni-l-tsʼin ‘He (4th person) punches himself’. He labels ʼá- “Refl.Acc” (template position 15, above the distributive at 13) and di- “TD”, a transitivity-decrease marker in a special inner position (8). In the first form the fourth-person subject prefix zh- sits between the two pieces. Young & Morgan's template, as McDonough 2019 reproduces it, puts “(Reflexive)” in the outer (disjunct) zone at position Ic, before the iterative (II) and the distributive plural (III), while direct-object pronouns are in the inner (conjunct) zone at IV. Willie's own (14) ʼádadiidleesh has da between the letters ʼá and di, and her gloss puts REF before PL.

HYPOTHESIS So R1 is right, in a sharper form (R1′). The outer piece ʼá- is a high adjunct of the verb word, attached above plural da-. The inner piece di- is a head inside the word, and heads are pronounced after all the non-heads. The reciprocal ʼahi- is one piece in the ordinary object position, which is below da-. With these positions the site's existing spell-out rule gives ʼá ⌢ da ⌢ di … for (14) and da ⌢ ʼahi … for (18), with no new rule. R2 (moving da- after syntax) is not needed, and no source reports such a move. R3 is true but does not help: the reciprocal is simply in the object position, and that never was the problem. The problem was treating the reflexive as one prefix.

HYPOTHESIS Three side results. (1) Willie says the agentive-passive prefix is “identical in phonological shape to the reflexive anaphor” (p. 36). On Kibrik's segmentation only the inner di- is shared. The passive's first piece is the indefinite deictic subject ʼa- (his position 9), not the reflexive's ʼá- (15), and Willie's own page prints the two with different tone (ʼádi- vs ʼadi-). (2) The reflexive is spread across the disjunct/conjunct boundary (the “#”). Binding and valency must then be computed in one domain that contains both sides, so that boundary cannot be where rule set A hands over to rule set B. This is consistent with the whole verb word (MVS) as the changeover, which remains a hypothesis, not an established result. (3) In Willie Ch. 3 (4) Tʼáábí ʼádílcha ‘He made himself cry’, the emphatic pronoun refers to the same person as the reflexive, yet Willie allows it. That fits linking to slots, not to referents: the pronoun is linked to the subject slot, and the ban (p. 32) is on linking to the anaphor's slot. That is a small point for Grant's weak thesis.

OPEN What is still open. Is the outer ʼá- the object itself, pronounced high, or a separate reflexive marker that binds a silent object (Speas 1986 n. 4 names both options)? Why does the reciprocal have no di-? Does ʼá- scope over da- (Rice's scope principle, unopened)? Young & Morgan's own pages were not opened, so their position number for ʼahi- is known here only from a tertiary source.

On this page: 1 Sources · 2 The problem, restated · 3 Data · 4 What the templates say · 5 Formal rules · 6 R1, R2, R3 tested · 7 Derivations RD1–RD8 · 8 What the outer piece is: RX-0, RX-1, RX-2 · 9 Gets right, misses · 10 Grant's thesis, Dowty, event structure, STRATA, the changeover · 11 Tasks T36a–h

1. Sources opened this run (2026-10-18, 12:46pm PT)

  • Kibrik, Andrej A. 2005. “Inflection versus derivation and the template for Athabaskan verb morphology.” (Box copy [local working copy, not in the repository], born-digital, 40 PDF pages, no printed page numbers. Already on the site for ex. (2), cited by section.) Newly read on page images: §3.2 “Proposals for the template structure”, chart (6), examples (7a,b) and (8a,b), and n. 5 (PDF pp. 17–19, 110 dpi; 300 dpi crops of (7) and (8)). The text layer garbles the glottal stop, so every form was read from the image. Kibrik credits (8) to “Young and Morgan 1987: 70-71” and bases his Navajo positions on Young, Morgan & Midgette 1992.
  • McDonough, Joyce 2019. “How to use Young & Morgan's 1987 The Navajo Language.” (Box copy [local working copy, not in the repository]; already on the site as secondhand Young & Morgan.) Newly read on a page image: Table 2, “The position class template with the types of morphemes in each position”, and the paragraphs around it (printed p. 32 = PDF p. 6).
  • Willie, MaryAnn 1991. Navajo Pronouns and Obviation (box copy). Re-read on page images this run: pp. 32–37 (110 dpi; 300 dpi crops of (14), (15), the p. 36 sentence with ʼádi-/ʼadi-, and (20)–(22); (11), (13), (16)–(19), (23) at 110 dpi, earlier crop-checked by the 2026-10-11 and 2026-10-18 4:46am runs). Also re-read: Ch. 3 p. 54, (3)–(4) (300 dpi crop of (4)), which nominal adjunction §4 already lists. Its reflexive is used here for the first time. Page offset: PDF = printed + 5 for pp. 4–51, and PDF 60 = printed 54.
  • Speas, Margaret 1986. Adjunctions and Projections in Syntax, MIT PhD dissertation (box copy). Re-read on a page image: p. 251, (4.9)–(4.10) and nn. 4–5.
  • Wikipedia, “Navajo grammar”, revision 1375578439 (2026-09-18, 10:21am PDT), raw wikitext. Tertiary, used for one point only: its object-prefix table lists Reflexive “(ʼá)-di-” and Reciprocal “ʼahi-”, and it says the object prefixes “can occur in position 4 as direct objects”. The table carries no reference. Nothing else is taken from it.
  • Not opened: Young & Morgan 1987 (pp. 70–71 and the template pages; known here only through Kibrik and McDonough), Young, Morgan & Midgette 1992, Saxon 1986 (cited by Speas n. 4), Rice 2000, Kari 1989, Golla 1970. Nothing is attributed to them beyond what Kibrik, McDonough and Speas report. A text-only dump of Young & Morgan is on the box from an unofficial mirror. It was not used: it has no page images, and its OCR scrambles the columns.

2. The problem, restated

SOURCED Willie p. 32: anaphoric prefixes “appear to the left of the subject pronoun in the prefix array”, “mark arguments that have patient theta roles”, “do not vary with person or number”, and always require “a -d or -l classifier, marking the verb intransitive”. P. 33: “Examples of reflexives with the distributive plural subject:”, then (14)–(15) with gloss REF.PL. P. 35: (18)–(19) “have subjects that are three or more in number, since the Distributive da- is present”, gloss DIST.REC. So the same da- (she calls it “distributive plural” on p. 33 and “Distributive” on p. 35) is glossed after REF but before REC.

HYPOTHESIS WD6 (Willie's framework §7) put both anaphors in ρX (rule WF-8) and Pl da- adjoined to QualP (WF-9). Both are non-heads, and Pl is higher, so C2 gives Pl ⌢ anaphor for both: right for (18), wrong for (14). The three escapes were:

OptionWhat it saysWhat it needs
R1The reflexive is two pieces: ʼá- above Pl, di- lowa segmentation showing ʼá and di as separate prefixes in different positions
R2The reflexive is one piece in ρX; da- is placed after syntaxa post-syntactic rule that puts da- after a reflexive but before a reciprocal
R3The reciprocal is not in ρX but in some position below Pla position for ʼahi- different from the reflexive's

3. Data SOURCED

3.1 Willie 1991: reflexive, reciprocal, agentive passive

Willie's gloss line is copied exactly. Her verb gloss order is object.subject.stem. Our Leipzig column only relabels it (REF → REFL, REC → RECP, PL/DIST → DISTR, PASS kept). Forms are not segmented except where Willie prints an underlying form with hyphens herself ((20), (21)).

Ex., pageForm (as printed)Willie's glossLeipzig (ours, relabelled)Translation (Willie's)
(11), p. 33ʼádiistsą́REF.1s.sawREFL.1SG.SBJ.sawI saw myself.
(13), p. 33ʼádiichʼidREF-1-scratchingREFL-1.SBJ-scratchingWe are scratching ourselves.
(14), p. 34ʼádadiidleeshREF.PL.1dl.paintingREFL.DISTR.1DU.SBJ.paintingWe are painting ourselves. (14′) We are painting each other.
(15), p. 34ʼádadiilzhííhREF.PL.1dl.blackeningREFL.DISTR.1DU.SBJ.blackeningWe are blackening ourselves. (15′) We are blackening each other.
(16), p. 34ʼahiidleeshREC.1dl.paintingRECP.1DU.SBJ.paintingWe are painting each other.
(17), p. 35ʼahiichʼidREC.1dl.scratchingRECP.1DU.SBJ.scratchingWe are scratching each other.
(18), p. 35daʼahiidleeshDIST.REC.1dl.paintingDISTR.RECP.1DU.SBJ.paintingWe are painting each other.
(19), p. 35daʼahiichʼidDIST.REC.1dl.scratchingDISTR.RECP.1DU.SBJ.scratchingWe are scratching each other.
(20), p. 37ʼádílzhééh (“ʼádi-ni-l-zheeh in underlying form”)REF.2s.shavingREFL.2SG.SBJ.shavingYou are shaving yourself.
(21), p. 37niʼdilzhééh (“ni-’adi-l-zheeh in underlying form”)2s.PASS.shaving2SG.OBJ.PASS.shavingYou are being shaved.
(22), p. 37ʼádíshdleeshREF.1s.paintingREFL.1SG.SBJ.paintingI am painting myself.
(23), p. 37shiʼdidleesh1s.PASS.painting1SG.OBJ.PASS.paintingI am being painted.
Ch. 3 (4), p. 54Tʼáábí ʼádílcha3 EMPH self-3-cry3.EMPH REFL-3.SBJ-cryHe made himself cry.

Willie's comments, quoted from the page images. P. 36: “the pronominal prefix that marks the unspecified agent is identical in phonological shape to the reflexive anaphor described in Sec. 2.1.2.1. above: ʼádi-. But note that the form ’adi- in the Agentive Passive in Navajo occurs to the right of (internal to) the object pronoun, proving that it marks the unspecified agent, and is not an object.” (The first is printed with a high tone, the second without.) Pp. 37–38: “In both these construction types, the patient precedes the agent in the prefix array.” P. 54: emphatic pronouns “are frequently seen in reflexive constructions”; reflexives “have inherent focus on these coreferent arguments”; “It is the reflexive pronoun ʼádi on the verb that makes (4) a reflexive sentence, not the emphatic pronoun”. P. 55, (5): Tʼáábí yicha, 3.EMPH 3.cries, ‘He himself is crying.’ (the plain intransitive).

3.2 Kibrik 2005, from Young & Morgan 1987:70–71: segmented reflexives and agentive passives

Forms exactly as Kibrik segments them. His raised glottal stop is written ’ (site convention), and his straight apostrophe in tsʼin likewise. The position numbers under the pieces are his. Leipzig is ours: DETR for his TD (“transitivity decrease”), CL for his TI (“transitivity indicator” = classifier, arrows as he prints them), INDF.SBJ for “Indef.Nom”, 4.SBJ for “4.Nom”, REFL.OBJ for “Refl.Acc”.

Ex.Form (Kibrik's segmentation)Kibrik's gloss linePositionsLeipzig (ours)Translation
(7a)táá-bí-ʼ-di-s-d-gizPref-3.Acc-Indef.Nom-TD-Conj-TI↓-wringbí 10, ’ 9, di 8PREF-3.OBJ-INDF.SBJ-DETR-CONJ-CL↓-wring‘He was washed’
(7b)bi-di-ʼ-ní-l-ʼį́3.Acc-TD-Indef.Nom-Pref-TI↑↓-seebi 10, di 8, ’ 93.OBJ-DETR-INDF.SBJ-PREF-CL↑↓-see‘He is looked at’
(8a)ʼá-zh-dí-l-zhééhRefl.Acc-4.Nom-TD-TI↑↓-mowʼá 15, zh 9, dí 8REFL.OBJ-4.SBJ-DETR-CL↑↓-mow‘He (4th person) shaves himself’
(8b)ná-ʼá-di-zh-ni-l-tsʼinPref-Refl.Acc-TD-4.Nom-Pref-TI↑↓-punchʼá 15, di 8, zh 9PREF-REFL.OBJ-DETR-4.SBJ-PREF-CL↑↓-punch‘He (4th person) punches himself’

Kibrik's comments. “Consider the transitivity decrease marker di-. It is found in reflexives and semipassives [agentive passives] in Navajo and Hupa, and can hardly be associated with any of the traditional positions; we need to postulate a special position for it (position 8 in the above chart). However, in Navajo this morpheme sometimes occurs to the left of 3rd person Nominative [deictic subject] pronouns”. “In these kinds of examples (7b), (8b) the metathesis treatment, I believe, is quite appropriate”. N. 5: “A tentative formulation of the rule could be as follows: Ci9 + di8 → di8 + Ci9 / __ {di, ni, …}, that is: the order of a position 9 prefix and the position 8 di- prefix is inverted in front of the prefixes of the shape di-, ni- and perhaps several others.”

SOURCED What the strings show, before any analysis. (a) In (8a) the reflexive's two pieces are separated by the subject prefix zh-. (b) In Willie (14)/(15) the letters ʼá and di are separated by da, and her gloss puts REF before PL. (c) In (16)–(19) there is no di after ʼahi. (d) The passive in (7a)/(7b) and Willie (21)/(23) has the object prefix first, and then a ’/’a piece and di.

4. What the templates say SOURCED

SourceReflexiveDistributive plural da-Direct object (incl. reciprocal?)Inner pieces
Young & Morgan 1987 via McDonough 2019 Table 2, p. 32Ic “(Reflexive)”, disjunctIII “(Distributive Plural)”, disjunctIV “Direct Object Pronouns”, conjunctV deictic subject pronouns; VIa–c adverbial-thematic, aspect; VII mode; VIII subject pronouns; IX classifier; X stem
Kibrik 2005 chart (6), “Standard Average Athabaskan”, Navajo from Young, Morgan & Midgette 199215 “reflexive Accusative [direct object] pronoun (Navajo, Hupa)”13 “distributive”10 “Accusative [direct object] pronoun”9 3rd-person (deictic) subject; 8 “transitivity decrease (Navajo, Hupa)”; 7, 5 qualifiers; 2 1st/2nd subject; 1 classifier
Wikipedia “Navajo grammar” (tertiary)object-prefix table: “(ʼá)-di-”position 3 “plural”“ʼahi-” in the object-prefix table; object prefixes “in position 4 as direct objects”—

SOURCED Caveats from the sources themselves. McDonough: “The template is problematic on several levels, and Y&M do not use it in the grammar except as a point of reference”. Morphemes are identified by template position “for identification and reference purposes alone, not for use in word formation”. Kibrik: vertical bars in his chart “mark the pairs of positions that might be interchanged in some languages”, and di- needs “a special position”. So the templates fix order, not structure. Structure is our job (§5).

HYPOTHESIS What all three agree on. The reflexive's ʼá- is in the outer, disjunct zone, before the distributive plural. The plain direct object is in the inner zone, after it. Young & Morgan file “(Reflexive)” with the optional disjunct categories (reversionary, semeliterative, iterative, distributive), not with the object pronouns. Kibrik calls it an accusative pronoun in its own position. The two labels differ, but the positions agree. The tertiary table puts ʼahi- with the object pronouns, and Willie's (18)/(19) order is what that predicts.

5. Formal rules (ours; all HYPOTHESIS unless marked)

RR-1  REFL = REFL_hi + DETR   The reflexive value [ana, refl] is pronounced by TWO elements of D1:
      (two pieces)             REFL_hi  ('á-)  a NON-HEAD adjoined to the top of D1 (QualP, or DETRP when
                                              DETR is present: Hale's "highest nuclear element"), ABOVE
                                              every other such adjunct from the reversionary down to Pl
                                              (Y&M Ic > Id, Ie, II, III; Kibrik 15 > 14, 13);
                               DETR     (di-)  a HEAD of D1, below the deictic-subject position and above
                                              the qualifiers (Kibrik 8: 9 > 8 > 7);
                               plus v = d/l  (Willie p. 32; Kibrik TI in (8)).
                               Segmentation and order: SOURCED (Kibrik (8a,b) < Y&M 1987:70-71;
                               McDonough Table 2). Head vs non-head, attachment: HYPOTHESIS.

RR-2  RECP = ONE PIECE        The reciprocal value [ana, recip] is pronounced in rho_X itself ('ahi-),
                               a NON-HEAD in the ordinary object position (Y&M IV per the tertiary
                               table; Willie (18)/(19) order). No DETR in (16)-(19) as printed.
                               v = d/l (p. 32). Needs #(rho_A) >= 2 (p. 34).

RR-3  PL  (= WF-9)            da- is a non-head adjoined to the top of D1 (QualP, or DETRP) and reads
                               rho_A's number (p. 32).

RR-4  CYC + MET               Spell-out of D1 by CYC (C1, C2, C3) as before, then ONE post-CYC,
                               phonologically conditioned reordering (Kibrik n. 5, "tentative"):
                                 MET   [X_9  di_8]  ->  [di_8  X_9]   /  __ {di, ni, ...}
                               (X_9 = a deictic subject: zh-, '-/'a-). Like SB5 (-ísh), it applies
                               after CYC and only to adjacent pieces.

RR-5  SHARED DETR             Agentive passive  = DEIC_indef ('-/'a-, Kibrik 9) + DETR (di-) + v = d/l.
                               Reflexive         = REFL_hi    ('á-,  Kibrik 15) + DETR (di-) + v = d/l.
                               Willie's "identical in phonological shape" (p. 36) = the shared DETR, plus
                               a first syllable that differs in tone in her print ('ádi- vs 'adi-) and in
                               position (Ic vs V).

RR-6  WHAT REFL_hi IS         OPEN between RX-1 and RX-2 (§8):
                                RX-1  rho_X[ana] is first-merged in Comp,V and silent; REFL_hi binds it.
                                RX-2  there is no separate rho_X; DETR identifies the internal role with
                                      rho_A, and REFL_hi marks that the identified role is the patient one.
RuleRestates / rests onData it coversStatus
RR-1Kibrik (6), (8); McDonough Table 2Kibrik (8a,b); Willie (11), (13)–(15), (20), (22), Ch. 3 (4)SOURCED order / HYPOTHESIS structure
RR-2Willie pp. 32–35; Wikipedia (tertiary)(16)–(19)HYPOTHESIS (position number OPEN, T36a)
RR-3Willie p. 32; Hale 2003 pp. 17, 39 (as recorded)(9), (14), (15), (18), (19)HYPOTHESIS (WD6 position question now settled: no clash left)
RR-4Kibrik n. 5(7b), (8b)SOURCED rule (Kibrik's, tentative) / HYPOTHESIS that it is post-CYC
RR-5Kibrik §3.2; Willie pp. 36–38(7a,b), (20)–(23)HYPOTHESIS on SOURCED segmentation
RR-6Speas 1986 n. 4none decides itOPEN (T36b, T36f)

6. R1, R2, R3 tested against the data

TestR1′ (two-piece reflexive, RR-1)R2 (one-piece reflexive in ρX; da- placed after syntax)R3 (reciprocal in its own position)
t1 (14)/(15): ʼá … da … di✓ by CYC (RD2)✓ only with a rule that puts da- inside the reflexive, unreported✗ silent on the reflexive
t2 (18)/(19): da … ʼahi✓ (RD1)✓✓, but WF-8 already derived it
t3 Kibrik (8a): zh- between ʼá and dí✓ (RD4)needs a second unreported rule to put the subject inside the reflexive✗ silent
t4 Kibrik (7b), (8b): di before the deictic subject, before ni/ní✓ with MET (Kibrik n. 5; RD5)✓ with MET—
t5 Y&M template: Ic (reflexive) before III (distributive)✓ that is the base order✗ R2 treats this order as derived; the template lists it as the base order—
t6 no di in (16)–(19)✓ RR-2 (the reciprocal has no DETR)says nothingsays nothing
t7 the passive shares di- (Willie p. 36; Kibrik §3.2)✓ RR-5 (RD6)says nothing—

HYPOTHESIS Verdict. R1′ derives every row with the site's existing spell-out rule plus Kibrik's own post-CYC metathesis. No datum contradicts R2, but R2 needs two reorderings that no source reports (t1, t3), and it treats as derived an order that the template lists as basic (t5). R3 is true (the reciprocal is in the ordinary object position), but it was never the issue, because that position is below Pl anyway. T35b is closed in favour of R1′. The structural questions inside R1′ (§8) stay open.

HYPOTHESIS What changes on the site. WF-8 on Willie's framework §5 is right for the reciprocal and wrong for the reflexive: the reflexive is not one exponent of ρX. WF-6 (SUBJ-HEAD) listed the passive's ʼadi- as one subject exponent on a head. On Kibrik's segmentation it is the indefinite deictic subject (position 9) plus DETR (8). WF-6 survives only if the deictic subject is pronounced on a head above DETR. If deictic subjects are adjuncts (Hale 2001, spell-out rules §3 (f2)), it is the known exception. This feeds T35c.

7. Derivations RD1–RD8

Format as on Willie's framework §7. Where something merges, moves or is pronounced is ours (HYPOTHESIS). Strings, glosses and segmentations are the sources' (SOURCED). Role labels follow the translations, and their entailments are read Dowty-style (role in the action). “Not segmented further” means the derivation's output is compared with the printed string as a whole.

RD1. (18) daʼahiidleesh ‘We (3+) are painting each other’: one-piece reciprocal under Pl

  1. Σ Σ(√PAINT) = {ρA (volition, causation: Proto-Agent), ρX (changed in appearance: Proto-Patient)}. ρA = [1, non-sg]; ρX = [ana, recip], which needs #(ρA) ≥ 2 (p. 34).
  2. A Merge V with ρX (Comp,V). Merge v = d-type (p. 32). ρA in Spec,v (VPISH). Merge M/S; Agree M/S ⟶ ρA [1, non-sg] (Willie's “1dl”). Merge Qual (no overt qualifier). Adjoin Pl to QualP; Agree Pl ⟶ ρA: number ≥ 3 (p. 35). Bind ρX to ρA inside D1.
  3. A Spell-out D1 by CYC. N(D1) = {Pl, ρX}; C2: Pl is higher → da ⌢ ʼahi. H(D1) = {M/S, v, V}; C3. C1: N before H. Output: da ⌢ ʼahi ⌢ [M/S ⌢ v ⌢ V] = printed (18) daʼahiidleesh ✓ (not segmented further).
  4. B No noun. None could be linked to ρX[ana] (p. 32).

RD2. (14) ʼádadiidleesh ‘We (3+) are painting ourselves / each other’: the outer piece above Pl, the inner piece a head

  1. Σ as RD1, but the internal role is reflexive. Under RX-1, ρX = [ana, refl] (silent). Under RX-2, ρX is identified with ρA (§8).
  2. A Merge V (+ ρX under RX-1). Merge v = d-type. ρA in Spec,v. Merge M/S; Agree M/S ⟶ ρA. Merge DETR (RR-1; Kibrik 8) above Qual (Kibrik 8 > 7) and below the deictic-subject position (9).
  3. A Adjoin Pl to QualP (RR-3). Adjoin REFLhi to QualP above Pl (RR-1; Y&M Ic > III). Hale: disjunct adjuncts are “adjoined to the maximal projection determined by the highest nuclear element” (2003 p. 39, as recorded on N4). With DETR present, that projection is DETRP rather than QualP. Their relative height is the template order.
  4. A Bind (RX-1: REFLhi binds ρX, and ρX is identified with ρA) or Identify (RX-2: DETR identifies the internal role with ρA). Either way, a plural ρA allows the reflexive and the reciprocal reading ((14), (14′); p. 34).
  5. A Spell-out D1. N(D1) = {REFLhi, Pl, (ρX silent)}; C2: ʼá ⌢ da. H(D1) = {DETR, M/S, v, V}; C3: di ⌢ M/S ⌢ v ⌢ V. C1: all N before all H. Output: ʼá ⌢ da ⌢ di ⌢ [M/S ⌢ v ⌢ V] = printed (14) ʼádadiidleesh ✓ (not segmented further; Willie's gloss REF.PL.1dl in the same order).
  6. A Contrast: WD6 with a one-piece reflexive in ρX gave Pl ⌢ REFL = *da-ʼádi-… ✗. The only change is RR-1.

RD3. (22) ʼádíshdleesh ‘I am painting myself’ and (20) ʼádílzhééh (“ʼádi-ni-l-zheeh”): no Pl, so the two pieces meet

  1. A As RD2 without Pl. N = {REFLhi}; H = {DETR, M/S [1sg] or [2sg], v, V}. Output: ʼá ⌢ di ⌢ M/S ⌢ v ⌢ V.
  2. A (20): Willie's own underlying form ʼádi-ni-l-zheeh has exactly this order, with her ʼádi = REFLhi + DETR, ni = M/S [2sg] and l = v ✓. (22): printed ʼádíshdleesh ✓ (not segmented further).
  3. A So Willie's single unit ʼádi- is what R1′ predicts whenever nothing intervenes. It looks like one prefix only because the two pieces are adjacent. (14) and Kibrik (8a) are the cases where they are not.

RD4. Kibrik (8a) ʼá-zh-dí-l-zhééh ‘He (4th person) shaves himself’: the subject between the two pieces

  1. Σ Σ(√SHAVE) = {ρA [4] (volition, causation), ρX (affected)}.
  2. A As RD3. The 4th-person subject is the deictic zh- (Kibrik position 9 = Y&M V). Two options for its status: (i) a head above DETR (WF-6); (ii) an adjunct-like non-head (Hale 2001, f2 (ii)).
  3. A Spell-out D1. (i): N = {REFLhi}, H = {DEIC zh, DETR, v, V} → ʼá ⌢ zh ⌢ dí ⌢ l ⌢ zhééh ✓. (ii): N = {REFLhi, zh}; C2 (REFLhi higher) → ʼá ⌢ zh; then H → same string ✓. MET does not apply: the piece after dí is l, not di/ni.
  4. A A one-piece reflexive cannot have a subject prefix inside it. (8a) is the strongest single datum for RR-1.
  5. A Update 2026-10-18, 8:46pm (T36g): a third option, (iii) zh left-adjoined to DETR as part of the head complex [zh + DETR], gives the same string and also fits Hale 2001 (22b, c), where the deictic precedes the qualifier. See subject on the head SD8 HYPOTHESIS.

RD5. Kibrik (8b) ná-ʼá-di-zh-ni-l-tsʼin ‘He (4th person) punches himself’: CYC, then Kibrik's metathesis

  1. A As RD4, plus the prefix ná- (Kibrik: “Pref”, no position given) as a higher non-head, and the prefix ni- (also “Pref” in Kibrik's gloss), which follows the subject.
  2. A Spell-out D1 by CYC gives ná ⌢ ʼá ⌢ zh ⌢ di ⌢ ni ⌢ l ⌢ tsʼin.
  3. A Post-CYC MET (RR-4; Kibrik n. 5): zh9 di8 before ni → di zh. Output ná-ʼá-di-zh-ni-l-tsʼin ✓.
  4. A So a late, local reordering does exist in this zone. But it is triggered by a following di/ni, swaps only positions 9 and 8, and never moves da- or ʼá-. It is not the rule R2 needs.

RD6. Agentive passive: Willie (21) niʼdilzhééh (“ni-’adi-l-zheeh”), (23) shiʼdidleesh; Kibrik (7a) táá-bí-ʼ-di-s-d-giz: the shared piece is DETR only

  1. Σ Σ(√SHAVE) = {ρA (indefinite human agent, p. 36), ρX [2sg] (affected)}.
  2. A Merge V with ρX; v = l/d-type; ρA = [indef] in Spec,v. The indefinite deictic subject ʼa-/’- (Kibrik 9, “Indef.Nom”). Merge DETR (Kibrik 8).
  3. A Spell-out D1. ρX is a non-head → first (C1) → ni/shi/bí before everything else (Willie: “to the right of (internal to) the object pronoun”, p. 36 ✓; “the patient precedes the agent”, pp. 37–38 ✓). Then ’a ⌢ di ⌢ v ⌢ V. (21) Willie's underlying ni-’adi-l-zheeh ✓; (7a) táá-bí-ʼ-di-s-d-giz ✓ (MET does not apply before s). (7b) bi-di-ʼ-ní-l-ʼį́: MET before ní ✓.
  4. B No noun may be linked to the indefinite agent (p. 36).

HYPOTHESIS Reflexive and passive share DETR and a d/l classifier, the valency-decrease part. They differ in the first piece. The reflexive's ʼá- is high (disjunct, position 15/Ic) and is about the patient role. The passive's ʼa- is a deictic subject (position 9/V) and is about the agent. That is why the passive's piece is “to the right of (internal to) the object pronoun” and the reflexive's is far to the left of everything. Willie's p. 36 observation is right about the subject, and the “identical shape” is only partial (her own print differs in tone).

RD7. Willie Ch. 3 (4) Tʼáábí ʼádílcha ‘He made himself cry’: a reflexive over a causer, and a pronoun linked to the binder's slot

  1. Σ Σ(√CRY) = {ρ1 (sentience)}: the plain verb is one-place ((5) Tʼáábí yicha ‘He himself is crying’, p. 55).
  2. Σ/event structure: the translation ‘made … cry’ adds a causer: {ρA (causation), ρX = the crier (sentience; causally affected)}. As on plain ł-causer, the causer is an event-structure addition. Willie does not segment the verb, so which prefix or classifier carries the causation is not asserted here.
  3. A RR-1: REFLhi + DETR identify the crier with the causer. ρA [3] → silent subject (gloss “3”). Spell-out: ʼá ⌢ di ⌢ M/S ⌢ v ⌢ V = printed ʼádílcha ✓ (not segmented further; Willie's gloss “self-3-cry” is in this order).
  4. B Merge the emphatic pronoun Tʼáábí in D2. Link it to ρA, the binder's slot, not to the reflexive's. Willie p. 32: “No pronoun or nominal that is coreferent with the reflexive may be adjoined to the sentence”. Tʼáábí is coreferent with the reflexive's referent, so p. 32 holds only if “coreferent with the reflexive” means “linked to the reflexive's slot”.

RD8. Schematic test (letters only): what each view of ʼá- predicts for a reflexive with a postposition object

  1. A Let P-obj be a postposition's object in Y&M 0/Ia, and REFLhi in Ic. All three views below put P-obj ⌢ REFLhi (template order, C2).
  2. A RX-0 (ρX itself pronounced at Ic) predicts that whatever lets oblique objects “assume the highest prefixal position” (Hale 2003 n. 13; f2 (i)) could also lift ρX[refl] past a preverb when none competes. RX-1 and RX-2 predict that REFLhi stays at Ic.
  3. Needs a printed reflexive with a disjunct preverb and no oblique (T36d). No such form was found in the sources opened.

8. What the outer piece is: RX-0, RX-1, RX-2

SOURCED Speas 1986 n. 4 (p. 251), on (4.10) Ashkii ádiiłtsą́ ‘The boy saw himself’ (Platero 1978:108): “It is not obvious that there really is an empty argument in the reflexive case; we might say that the verbal morphology removes the object theta role in some way. However, see Saxon (1986) for some quite convincing arguments that an equivalent construction in Dogrib, another Athabaskan language, does contain an empty pronominal which is also anaphoric.”

ViewWhat ʼá- isWhat di- isFitsProblem
RX-0 SLOT-HIGHthe object slot ρX itself, pronounced at IcDETRKibrik's label “Refl.Acc”; Willie's “anaphors mark arguments” (p. 32)Needs the slot to move up, or to be pronounced out of height order. The first clashes with Willie's “no movement to and from A positions” (p. 23) if that covers slots (T35e). The second adds a third exception to CYC beside f2 (i), (ii)
RX-1 BINDER + SILENT SLOTa reflexive operator adjoined high that binds a silent ρX[ana]DETR (valency marker)Saxon's “empty pronominal which is also anaphoric” (via Speas); Willie's chain view (p. 25); keeps two role slotsWhy is the operator so high (above Pl), when its binder ρA is lower? Only if order here is just template order (C2), not scope
RX-2 ROLE IDENTIFICATIONa disjunct marker saying the patient role is the identified one (Y&M file it with other disjunct categories)the identification itself (“removes the object theta role in some way”, Speas n. 4; Kibrik's “transitivity decrease”)Hale & Keyser on O'odham: the reflexive there “is not a "true" reflexive. Rather … it is a morphological reflex of formal detransitivization” (ms. hale012, PDF p. 29, re-read this run); one-slot reflexiveKibrik calls ʼá- an accusative pronoun (an analysis, not data). If the reflexive participant is one slot, “the patient precedes the agent” (pp. 37–38) needs restating for reflexives

HYPOTHESIS R1′ lets Speas's two options both be partly right. The two pieces divide the labour Speas describes. The inner di- (with the d/l classifier) is the “morphology removes the object theta role” part, and it is shared with the passive. The outer ʼá- is the anaphoric part, and it is unique to the reflexive. No printed datum separates RX-1 from RX-2 (T36b, T36f). RX-0 is kept as an option but is the costliest.

9. What each analysis gets right, and what it misses

AnalysisGets rightMisses or leaves open
Willie 1991The anaphors are left of the subject, “do not vary with person or number”, require d/l, and exclude a linked noun (p. 32). Reflexive vs reciprocal readings with a plural subject (p. 34). The passive's prefix is internal to the object and marks the agent (p. 36). Prints the order evidence ((14) REF.PL vs (18) DIST.REC) faithfully.Treats ʼádi- as one unit ((20) “ʼádi-ni-l-zheeh”), which her own (14) contradicts. “Identical in phonological shape” overstates the identity (tone, position). p. 32's “coreferent with the reflexive” is too strong for Ch. 3 (4).
Kibrik 2005Segments the reflexive into two positions (15 and 8) and the passive into 9 and 8. Identifies di- as shared transitivity decrease. States the 9/8 metathesis with its trigger (n. 5).A template, not a derivation: why 15 is above 13, and why di- needs “a special position”, is not explained. The metathesis rule is “tentative”. The reciprocal is not discussed.
Young & Morgan 1987 (via McDonough)The reflexive is disjunct (Ic), before the distributive plural (III); direct objects are conjunct (IV).By McDonough's own account, positions are reference labels, “not for use in word formation”, and the template is a backformation. Not opened directly.
Speas 1986 (n. 4)Names the two live options for the reflexive (role removal vs anaphoric empty pronominal).Does not decide. Reflexive morphology not segmented.
This site before this run (WF-8, WD6)The reciprocal as an anaphor value on ρX, below Pl. Saw that CYC failed on (14), and named R1 as one way out.Treated the reflexive as one exponent of ρX. WF-6 treated the passive's ʼadi- as one subject exponent.
This page (RR-1 to RR-6, CYC + MET)Derives (14), (15), (18), (19), (20)–(23), Kibrik (7a,b), (8a,b), Ch. 3 (4) with the existing rule plus Kibrik's metathesis. Explains Willie's apparent unit ʼádi- as adjacency.The attachment height of REFLhi and the position of DETR are taken from the template, not derived. RX-0/1/2 are not separated. The reciprocal's lack of DETR is unexplained (T36c).

10. Grant's thesis, Dowty, event structure, STRATA, and the changeover

GRANT'S CONJECTURE → HYPOTHESIS Weak thesis. Nothing here contradicts it. Under RX-1 the reflexive verb still has two role-bearing slots, one of them anaphoric and silent. Under RX-2 one slot carries both roles. Either way the verb's arguments are slots or roles, and no noun is involved. RD7 adds a small positive point. In Tʼáábí ʼádílcha the emphatic pronoun is the same person as the reflexive but is allowed, so linking targets a slot (the binder's), not a referent. That is what “nouns and free pronouns are optional add-ons linked to slots” predicts. A referent-based linking rule would wrongly block (4). Strong thesis: not tested here.

HYPOTHESIS Dowty. A reflexive is one individual who has the Proto-Agent entailments of one role (volition, causation) and the Proto-Patient entailments of the other (changed or affected). Navajo marks this with a piece that targets the more patient-like role (Willie: “patient theta roles”, p. 32; WF-8's restriction to the most Proto-Patient slot). Dowty's selection principle is the baseline. Navajo's extension is that the anaphoric marking itself is tied to the Proto-Patient side. That is an extension, not a departure.

HYPOTHESIS Theta roles and event structure as compatible descriptions (Grant's stance). R1′ shows the two descriptions as two pieces of one word. The outer ʼá- says which role is anaphoric, a fact about the roles. The inner di- with the d/l classifier says that the valency is reduced, a voice/event-structure fact shared with the passive. Neither replaces the other. Both are true of the same reflexive, and the morphology pronounces them separately.

HYPOTHESIS STRATA (theta roles first, then event-structure roles, then noun selection). RD7 gives an ordering fact. In ‘He made himself cry’ the reflexive identifies the crier with a causer, a participant the event-structure layer adds to a one-place verb. So reflexive identification applies at or after the event-structure layer, and the emphatic pronoun comes after that (layer 3). This is consistent with STRATA's order. It concerns roles, not nouns, so it does not separate STRATA from the adjunct and inside-the-clause views of noun attachment. Those three still give the same judgments here. Caveat: the causative reading rests on Willie's translation. She does not segment the verb (T36h).

HYPOTHESIS The changeover point (research focus 1). The reflexive's outer piece is disjunct (Y&M Ic) and its inner piece is conjunct (Kibrik 8, within the conjunct zone). Binding or role identification must see both pieces and the subject slot in one domain. If the disjunct/conjunct boundary (“#”, B1 on headedness §3) closed a cycle, the outer piece would be merged after the inner zone was sealed and could not bind into it. So B1 is not the A/B changeover. That agrees with the site's D1 = MVS, which includes the disjunct adjuncts, and with the B3 data under T2. Inside D1, rule set A orders the pieces by head vs non-head and then by height. Both reflexive pieces obey that, and the only exception is Kibrik's local, phonological metathesis after spell-out.

HYPOTHESIS Phrase-level packaging (Grant's reminder that lexicalization happens at the phrase level too). The reflexive is a discontinuous unit listed as one item (Willie's ʼádi-; the tertiary table's “(ʼá)-di-”), whose two parts land in different positions and can be split by other prefixes. That is the word-internal analogue of a separable phrasal verb: listed together, placed apart by the syntax.

11. Tasks T36a–h

#TaskNeeds
T36aOPEN Confirm Young & Morgan's position for the reciprocal ʼahi- (IV?) and for the reflexive di-, from the book's own pagesYoung & Morgan 1987 template and pp. 70–71 (not opened)
T36bOPEN When does the reflexive appear without ʼá- (the tertiary table writes “(ʼá)-di-”)? A di--only reflexive would favour RX-2Young & Morgan; Kibrik 1996/2001a (not opened)
T36cOPEN Why does the reciprocal have no DETR di- in (16)–(19)? Collect more reciprocal forms, including reciprocals with postpositionsYoung & Morgan; Willie Ch. 3–9
T36dOPEN A reflexive with a disjunct preverb or oblique: does ʼá- stay at Ic (RX-1/2) or behave like Hale's oblique objects (RX-0)? (RD8)printed forms (Young & Morgan; Hale 2003 n. 13 sources)
T36eOPEN Scope of REFLhi vs Pl: does (14) mean ‘each of us paints himself’ or another reading, and does Rice's “greater scope to the right” reverse C2 here? (joins T29e)Rice 2000 (not opened); speakers
T36fOPEN RX-1 vs RX-2: Saxon 1986's Dogrib arguments for an empty anaphoric pronominal; does any Navajo test (number, linking, binding into a postposition) separate them?Saxon 1986 (not opened)
T36gDONE 2026-10-18, 8:46pm (subject on the head SD8): neither option; the deictic subject is left-adjoined to DETR (the highest nuclear head), part of the head complex, which gives the same strings for (7a), (8a) HYPOTHESIS. Original task: WF-6 after RR-5: is the deictic subject (ʼa-, zh-, ji-) on a head above DETR, or an adjunct (Hale 2001)? (joins T35c, T29f)Hale 2001, 2003 n. 11 (on the site)
T36hOPEN Is (4) ʼádílcha a reflexive built on a causative? Do reflexives always apply after causativization (a STRATA ordering test)?a segmentation of (4); Goertz et al. 2006 forms (on the site)

Sources: Kibrik 2005 §3.2, (6)–(8), n. 5 (box copy [local working copy, not in the repository], page images PDF pp. 17–19; crops under [local working copy, not in the repository], not committed). McDonough 2019, Table 2, p. 32 (box copy [local working copy, not in the repository]). Willie 1991 pp. 32–37, 54–55 (page images and crops under [local working copy, not in the repository]). Speas 1986 p. 251. Wikipedia “Navajo grammar” rev. 1375578439 (tertiary, one point). Hale 2003, Hale & Keyser (hale012) and Dowty 1991 as recorded on the site. Cross-links: Willie's framework WD6 (T35b), A/B as spell-out, open questions (T36), sources.

Can da- count the patient? Two hooks on one plural prefix (T53d)

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

Verbs on this page → dictionary: yiztsʼǫs ‘kiss’ · yishxash ‘bite’ · dabiiłtsą́ ‘see’ · yíghááh ‘join; with -tah ‘among’: go among’

Restated on the unified structure (2026-11-15). The plural da- is restated as Dist⁰ just above vP, agreeing with the closest argument it c-commands, with the inverse patient counted from the outer Spec,vP (H21a/b). Its place inside the verb word is one of the misfits recorded for the Merge-site account of prefix order (OPEN). The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Flat-S and other pre-VPISH notation below is labelled contrast. Status on the spine: see unified structure, §4, distributive row and §3.2. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

The deep-or-shallow page left T53d open: under the inverse bi-, can the plural prefix da- ever count the patient alone (many patients, one doer)? This page reads Willie 1991 pp. 83–86 again from the page images and answers it. In plain words, the answer is no. da- always needs more than one doer. But in the inverse, and only there, it can also make the patient plural, and that plural lets a bare noun stand in the subject slot without being definite. One prefix therefore has two hooks. One is fixed to the doer's role. The other is optional and reaches the patient only after the patient has moved up into the subject slot. That second hook is a small new piece of evidence that bi- really moves the patient (a weak fifth "deep inverse" test).

On this page: 1 Question · 2 Sources · 3 Data · 4 T53d answered · 5 Rules · 6 Two hooks: a derivation-based account · 7 Derivations · 8 A fifth deep-inverse test? · 9 Grant's ideas · 10 Bottom line · 11 Open

Notation (the site's own; no Syntax · Semantics · Interface bot conventions file has arrived, so the provisional notation is kept). Σ = role-schema step; A = word level (rule set A: the prefixal, left-headed spine, spelled out as one word W); B = clause level (rule set B: right-headed clause syntax, noun linking). W is the point where A hands over to B. ρA = agent slot, ρP = patient slot (labels ours). EXT = the slot realized as the external argument. #(ρ) = the number of the slot's referent (sg, du, pl = three or more). Pl = the head spelled out da-. DIST-1…5: number & scope §3. INV-4 (the inverse subject must be definite) and INV-6 (the noun next to the verb is internal): inverse page. RCC, D1–D4: deep or shallow inverse?

Reminder: Merge builds structure, and moving something already built is internal Merge (Move). Agree copies features such as person and number between a head and an argument, and nothing moves. Below, the patient's rise into the subject slot is Merge (Move). What da- does is Agree.

1. The question

  • T53d (from deep or shallow inverse? §11): can da- ever count the raised patient alone (plural patient, singular agent) under bi-? The 4:46am run read pp. 84–85 only in the text layer, which garbles Navajo, and so it left this open.
  • Correction to the record SOURCED: the data were already page-checked on the number & scope page (2026-10-09; DIST-1, DIST-3, DIST-5). The 4:46am page did not connect T53d to them. This page makes the connection, checks the images again, and adds the new analysis in §6–§8.

2. Sources opened this run (2026-10-23, ~12:46pm PT run)

  • Willie, MaryAnn 1991. Navajo Pronouns and Obviation. PhD diss., University of Arizona. Box PDF. Page images read at 110 dpi: printed pp. 83, 84, 85, 86 (PDF pp. 89–92). Every Navajo line below was re-read from these images. They match the transcriptions already on the number & scope page.
  • Sandoval, Merton & Eloise Jelinek 1989. "The bi-construction and pronominal arguments in Apachean." In the Jelinek collection PDF on the box, printed pp. 345–352 (PDF pp. 57–64), page images. These pages are about Jicarilla. They were re-opened for T53b (is RCC hard or soft?) and contain nothing on rank strength or on number. They are used here only for the remark on (35) in §8, and no Navajo form is taken from them.

Nothing else was opened. Young's survey and Platero 1982 are still not on the box (T53b, T53f).

3. The data: Willie 1991 pp. 83–85 SOURCED

The forms are transcribed from the page images. The gloss lines are Willie's as printed, and they already use Leipzig-style category labels (PL, INV, 3). Willie does not segment the verb words into morphemes, and neither do we. A ✓ or * is hers. Unglossed nouns are glossed in brackets with her glosses of the same words elsewhere on the page.

Willie 1991:83–85
(64a)dabiiłtsą́
PL.INV3.3.saw
‘They/he (was)were seen by them/*him.’ (p. 83)
(64b)shizhéʼéʼashiikédabiiłtsą́
1s.fatherboysPL.INV3.3.saw
‘My father was seen by the boys.’ (p. 83)
(65a)ʼashiikéʼatʼéékédabiiłtsą́
boysgirlsPL.INV3.3.saw
‘The boys were seen by the girls.’ (p. 83)
(65b)*ʼashiikéʼatʼééddabiiłtsą́
[boys][girl][PL.INV3.3.saw]
(The boys were seen by the girl.) (p. 83)
(65c)*ʼatʼéédʼashiikédabiiłtsą́
[girl][boys][PL.INV3.3.saw]
(The girl was seen by the boys.) (p. 83)
(66)ʼéidí ʼatʼéédʼashiikédabiiłtsą́
[DEM girl][boys][PL.INV3.3.saw]
‘That girl (over there) was seen by the boys.’ (p. 84)
(67a)mąʼiitłʼízídeishxash
coyotegoatPL-3-3-bit
‘The coyotes bit the goat(s).’ (p. 84)
(67b)tłʼízímąʼiidabishxash
[goat][coyote]INV3-3-bit
‘The goats were bitten by the coyotes.’ (p. 84)
(68)mąʼiitłʼízí yitahdaʼashxash
coyotegoat 3.amongPL.bit
‘The coyote (went among and) bit the goats.’ (p. 85)
(69)ʼéidí tłʼízímąʼiidabishxash
DEM. goatcoyoteINV3.3.bit
‘That goat there was bitten by the coyotes.’ (p. 85)

p. 83: "With common nouns, da- plus bi- requires that both internal and external arguments be plural." "However, if the external argument is made maximally referential with demonstratives, then the sentence becomes acceptable." p. 84: "I conclude that bi- requires the external argument to be definite, but when the Distributive da- is present, it can license a plural external argument, which can escape the requirement on definiteness. With dabi-, the internal argument must be plural, and the external argument may optionally be interpreted as plural." p. 85: "In (67a), the agent must be interpreted as plural. In (67b), both must be interpreted as plural. If we want to produce a reading in which the agent may be singular, then the following construction is used." [(68)] "We have seen that number can be associated with different arguments according to the voice alternation. This would not be possible if number were a property of the nominals, rather than the pronouns. In the Inverse voice, the external argument, the patient, must be either definite or plural if the sentence is to be acceptable." As printed: (67b)'s gloss has no PL although the word begins da-. (69)'s gloss uses dots, and (67)'s uses hyphens.

Also relevant, already on the site: (57) ʼatʼééké John biztsʼǫs ‘Those two girls were kissed by John’ (p. 79, no da-). Willie: "the external patient argument is interpreted as definite" (number & scope §2.2).

4. T53d answered: no patient-only reading SOURCED

Three independent pieces of data in Willie say that, under dabi-, the agent must be plural. A singular agent with a plural patient is therefore never available:

EvidencePlural patient, singular agent?Page
(64a), with no nouns at all‘They were seen by *him’: outp. 83
(65b), common nouns with plural forms*‘The boys were seen by the girl’p. 83
(67b), nouns with no plural formonly ‘the goats … by the coyotes’; the agent "must be interpreted as plural"pp. 84–85
(68), the way around it‘one coyote, many goats’ is said with a different construction: the goats are the object of the postposition yitah ‘among’, and the verb has no bi-p. 85

So T53d: closed, negative. In the inverse transitive frame, da- never counts the patient alone. This is DIST-1, now confirmed directly for the question as posed. What da- counts in (68) is still T10d (DIST-5: Willie glosses the verb only PL.bit).

5. Rules, tightened

DIST-1   Pl present → #(ρ_A) = pl, in both voices.                              (as before)
DIST-1c  Corollary.  No yi-/bi- derivation lets Pl be satisfied by ρ_P alone.     NEW
DIST-6   Second hook (inverse only).  Under bi- + Pl, #(EXT = ρ_P) may be pl.
         When it is pl, EXT is exempt from INV-4 (definiteness).                   NEW
DIST-6′  Without Pl, a plural (paucal) EXT under bi- does not escape INV-4:
         it is read definite ((57)).                                               NEW (weak)
INV-4/DIST-3 restated:  under bi-, EXT must be [+def] ∨ (Pl ∧ #(EXT) = pl).
RuleStatusBasis
DIST-1cSOURCED(64a), (65b), (67b); p. 84 "the internal argument must be plural"; (68) as the alternative.
DIST-6SOURCED as Willie's generalization (p. 84 "can license a plural external argument, which can escape the requirement on definiteness"; p. 85 "either definite or plural")(65a) ✓ vs (65c) *; (67b) where both "must be" plural; (66) and (69), where a demonstrative lets the EXT be singular.
DIST-6′HYPOTHESIS (weak)(57) is paucal (two), not plural (three or more), so the contrast with (65a) mixes two variables: whether da- is present, and how many there are. Whether a bare EXT of three or more can appear under bi- without da- is not shown (T54a).
Disjunctive INV-4SOURCED (Willie p. 85 wording) + HYPOTHESIS (that the plural disjunct needs Pl)p. 84 ties the plural escape to "when the Distributive da- is present".

6. Two hooks on one prefix HYPOTHESIS

Put DIST-1 and DIST-6 side by side:

Direct (yi-)Inverse (bi-)
Agent must be pluralyes (63d *, 67a)yes (64a, 65b, 67b)
Patient plural licensed by da-no visible effect: patient number "irrelevant" (p. 32; DIST-2)yes, optionally, and it lifts definiteness (DIST-6)

A single Agree relation cannot produce this pattern. If Pl always agreed with the closest argument, it would agree with the raised patient in the inverse, and the agent would be free there. That is wrong ((64a), (65b); this is the row that number & scope §5.1 already excludes). If Pl agreed only with the agent's role, it would never touch the patient, and DIST-6 would be unexplained. The proposal is that Pl has two probes:

PL-2H  Pl carries two probes.
       (i)  [#, ρ_A]  role-restricted, obligatory: Agree with the slot bearing the most
            Proto-Agent entailments (C5), wherever it sits.         → DIST-1, DIST-1c
       (ii) [#]       unrestricted, optional: Agree with the CLOSEST φ-bearing goal
            c-commanded by Pl.                                       → DIST-6
       Pl is merged above the landing site of the inverse Move (outer Spec,v; D-2 on the
       deep-or-shallow page).

What follows:

  • Direct. Nothing has moved, so the closest goal is ρA itself. Probe (ii) is either idle or redundant with (i). Result: the patient's number is untouched (DIST-2) ✓.
  • Inverse. ρP has moved over ρA, so it is now the closest goal. Probe (i) still has to find ρA, so the agent must be plural ✓. Probe (ii) may agree with the raised ρP, so the patient can be plural, and a plural EXT is licensed against INV-4 ✓ (DIST-6). When probe (ii) does not apply, EXT must be definite on its own: (66), (69), (64b) ✓.
  • The patient alone. This would need probe (i) to fail and probe (ii) to apply, but (i) is obligatory. So it is impossible ✓ (DIST-1c, T53d).

This is the "da- targets the agent's role feature" option (e2) on the inverse-as-Agree page, K5, with a second, ordinary locality probe added. It also deals with the CYC ordering problem recorded there: Pl sits above the moved patient, so da- before bi- (dabiiłtsą́) is the expected order under rule set A, where higher heads go further left. OPEN whether Pl really is that high (T3, T30c).

Alternatives, not excluded. (a) The plural escape is semantic: a plural bare noun can be read as a kind or a group, which counts as "definite enough", with no help from Pl. DIST-6′ ((57)) argues weakly against this, because there a plural without da- is still read definite. (b) W&J's constant gloss (agent = 3NOM in both voices), where Pl just piggybacks on Mode/Subject. That would give DIST-1 but not DIST-6. Neither alternative is contradicted by the opened data, and PL-2H is the only one of the three that derives both rules.

7. Derivations, step by step

The steps are labelled by rule set: Σ role schema, A word (W), B clause. Each step names its operation. VPISH: both slots are first-merged inside the verb's phrase, ρA in Spec,v and ρP as the complement of V. The spine is the site's Qual > M/S > v > V, with Voice above, as on the deep-or-shallow page, and Pl above the outer Spec,v (PL-2H). No flat-S tree is used, and the verb words are not segmented beyond Willie's gloss line.

G-1. (67b) tłʼízí mąʼii dabishxash ‘The goats were bitten by the coyotes.’

  1. Σ Σ(√BITE) = {ρA (biter), ρP (bitten)}. The roles are fixed here and at external Merge.
  2. A Merge V + ρP. Merge v[EPP]. Merge ρA in Spec,v. Rank is equal (animal/animal, both third person), so the voice is a free choice and discourse picks EPP.
  3. A Move (internal Merge) ρP to outer Spec,v, over ρA. RCC: equal rank, allowed.
  4. A Merge Pl. Agree, probe (i): Pl ⟶ ρA, so #(ρA) = pl. Agree, probe (ii), chosen: Pl ⟶ ρP, the closest goal, so #(ρP) = pl.
  5. A Merge M/S, Qual, Voice (spelled out bi-). INV-4 check on EXT = ρP: there is no demonstrative, but the slot is pl and Pl is present, so the plural disjunct is satisfied ✓. Seal → W = dabishxash.
  6. B Link (INV-6): mąʼii, next to W, links to the internal slot ρA. tłʼízí links to EXT = ρP. Neither noun has a plural form, so the number is read off the slots: ‘coyotes’ and ‘goats’. Willie: "both must be interpreted as plural" ✓.
  7. A Counterfactual: if probe (ii) is skipped at step 4, then #(ρP) = sg and EXT is bare. INV-4 fails at step 5, so there is no ‘the goat was bitten by the coyotes’ reading without a demonstrative. That matches (69), which has one.

G-2. (69) ʼéidí tłʼízí mąʼii dabishxash ‘That goat there was bitten by the coyotes.’

  1. ΣA Steps 1–3 as in G-1.
  2. A Merge Pl. Probe (i): Pl ⟶ ρA, pl ✓. Probe (ii) is not applied, so #(ρP) = sg.
  3. A INV-4: EXT is singular, so it must be [+def]. B This is met by the clause-level demonstrative ʼéidí. HYPOTHESIS The word-level check must therefore be able to see a property supplied in rule set B, or INV-4 is checked after W is sealed. This is the same timing issue as T7 and the strata question in §9.
  4. B Link: mąʼii → ρA (‘coyotes’, by the slot's number), ʼéidí tłʼízí → EXT = ρP (‘that goat’) ✓.

G-3. (65b) *ʼashiiké ʼatʼééd dabiiłtsą́ (the patient alone)

  1. Σ Σ(√SEE) = {ρA (perceiver), ρP}. Intended: ρA = girl (sg), ρP = boys (pl).
  2. A Merge and Move as in G-1, steps 2–3.
  3. A Merge Pl. Probe (ii) could value ρP as pl, but probe (i) is obligatory: Pl ⟶ ρA, which is sg ✗. The derivation crashes inside W, before rule set B. The same star appears with no nouns: (64a) ‘by *him’.
  4. B The only way to say it is to change the frame, as (68) shows for ‘bite’: make the patient the object of a postposition (yitah) and drop bi-.

8. A fifth deep-inverse test? HYPOTHESIS (weak)

If PL-2H is right, probe (ii) can reach the patient only because the patient has moved above the agent. A shallow inverse, which only changes the marking, would leave ρA as the closest goal, and DIST-6 would have nothing to grab. So DIST-6 adds a fifth item to the four tests on the deep-or-shallow page:

TestDeep predictsShallow predictsNavajo
D5 Optional locality agreementA plain closest-goal probe can reach the patient in the inverse onlyIt never reaches the patientPatient plural is licensed in the inverse only (DIST-6 vs DIST-2): deep, weakly

It is weak for two reasons. First, it relies on PL-2H, which is our analysis and not Willie's. Second, alternative (a) in §6 could produce the same contrast without any movement. Sandoval & Jelinek's Jicarilla pair (35a/b) shows only the change in which noun is indexed ("the coindexing … has been reversed", p. 349), which is test D3 again. They say nothing about number.

9. Grant's ideas

  • SOURCED Willie's own argument (p. 85): number moves between arguments with the voice, and "this would not be possible if number were a property of the nominals, rather than the pronouns." This is the source's reasoning, and it points the same way as GRANT'S CONJECTURE that the verb's arguments are feature bundles (role + person + number), not nouns. Nothing on pp. 83–86 contradicts the conjecture.
  • HYPOTHESIS Strata. PL-2H gives one prefix a role-level hook (probe (i), read off the theta/proto-role layer, which is fixed first) and a position-level hook (probe (ii), read off the structure after Move). The definiteness check (INV-4) needs information from noun selection (ʼéidí), the latest layer. The order the derivations need is: roles → Move → number agreement → definiteness, with the noun visible. That fits Grant's order (theta roles first, then event-structure roles, then noun selection last), with one caveat. Event structure plays no visible part here, so this data neither supports nor tests the middle stratum.
  • OPEN Intra-word Merge/Agree sites. All of G-1 to G-3 apart from linking happens before W is sealed: one Move and two Agree relations, all inside the word. If PL-2H holds, the word contains an Agree site with two goals. That is one more candidate for Grant's word-internal Agree sites (intra-word Merge/Agree).
  • Chemistry: no correlate is claimed.

10. Bottom line

SOURCED In Navajo inverse sentences, the plural prefix da- always requires more than one doer. It never counts only the patient (Willie (64a), (65b), (67b); (68) is the workaround). HYPOTHESIS The prefix has a second, optional hook that can also make the patient plural, but only once the patient has moved up into the subject slot. That counts as weak extra evidence that bi- really moves the patient.

11. Open

IDQuestionWhat would settle it
T54aCan a bare EXT of three or more stand under bi- without da-? (This tests DIST-6′ and alternative (a) of §6.)A source with a ≥3 patient-subject and no da-; a speaker. OPEN
T54bIs Pl merged above the inverse landing site (needed by PL-2H)? (= T3, T30c)Scope of da- relative to bi-; Hale 2003 on position III. Update 2026-10-23 ~8:46pm PT: answered in part on how high is da-? (Hale 2003: QualP or vP, SOURCED; above outer Spec,v, HYPOTHESIS). HYPOTHESIS
T54cWith an equal-rank pair, does probe (ii) ever apply in the direct when the patient is fronted for some other reason (e.g. focus)?Willie Ch. 8 focus data with da-. OPEN
T54dIs a 1st-person-possessed kin term (64b shizhéʼé) definite by itself, or licensed some other way?Other possessed EXT nouns under bi-. OPEN
T53dCan da- count the raised patient alone?Closed 2026-10-23 ~12:46pm PT: no (§4). SOURCED

How high is the plural prefix da-? Hale's three placements and the inverse (T54b)

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

Verbs on this page → dictionary: yishxash ‘bite’

Restated on the unified structure (2026-11-15). The plural da- is restated as Dist⁰ just above vP, agreeing with the closest argument it c-commands, with the inverse patient counted from the outer Spec,vP (H21a/b). Its place inside the verb word is one of the misfits recorded for the Merge-site account of prefix order (OPEN). The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). Flat-S and other pre-VPISH notation below is labelled contrast. Status on the spine: see unified structure, §4, distributive row and §3.2. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

2026-11-11 audit pass 2. Wording that used VPISH as the name of a spine, tree, skeleton or set of slots named a structure after the hypothesis. VPISH is a hypothesis name, never a node; the structure is a vP/TP spine with the subject first-merged in left Spec,vP. Only the wording changed.

2026-11-11 audit pass 3. Status badges now use only the label set SOURCED, HYPOTHESIS, SPECULATIVE, OPEN, NONE and GRANT'S CONJECTURE; verdict or task words that had been styled as badges (for example Opened, PRE-VPISH, DONE, SUPPORTS, PARTLY ANSWERED) are now plain text or a label with a qualifier. See the pass 3 report.

2026-11-11 audit pass 5. In the gloss line of ʼÁdabiʼniilyaa (3), the morpheme ʼnii was typed with a right quotation mark (U+2019) instead of the glottal-stop letter ʼ (U+02BC). Character fix only; the form is unchanged. See the pass 5 report.

The da-reach page gave the plural prefix da- two hooks (rule PL-2H). The second hook only works if da- sits above the spot the patient moves to in the inverse. That was left open as T54b. This page checks where Ken Hale actually put da- in the tree. In plain words: Hale put it in three different places over the years. In his two earlier works it sits low, right next to the verb stem. In his latest work (the 2003 manuscript) it sits high, at the top of the word's structure, above the subject piece. The latest placement is the one PL-2H needs. The earlier ones would leave the doer out of its reach and empty the "fifth deep-inverse test". Hale's latest view therefore fits our analysis, but he never discusses the inverse, so whether da- really sits above the patient's landing spot is still our inference, not his.

On this page: 1 Question · 2 Sources · 3 Data · 4 Hale's three placements · 5 Restated with VPISH · 6 Rules · 7 Derivations · 8 T54b: what is settled · 9 Grant's ideas · 10 Bottom line · 11 Open

Notation (the site's own; no Syntax · Semantics · Interface bot conventions file has arrived, so the provisional notation is kept). Σ = role-schema step; A = word level (rule set A: the prefixal, left-headed spine, spelled out as one word W); B = clause level (rule set B: right-headed clause syntax, noun linking). W is where A hands over to B. ρA = agent slot, ρP = patient slot (labels ours). EXT = the slot realized as external argument. Pl = the head or adjunct spelled out da- (template position III). Pl(H) = Pl attached high, Pl(L) = Pl attached at vP, Pl(VP) = Pl inside VP below v. DIST-1…6, PL-2H: da-reach §5–6; RCC, D1–D5: deep or shallow inverse?; IA-1*, T30g: the inverse as Agree and EPP.

Reminder: Merge builds structure. Attaching da- to the tree is Merge (an adjunction), and moving the patient up is internal Merge (Move). Agree copies features such as number between a head and an argument, and nothing moves. The height question is about Merge (where da- is attached). PL-2H is about Agree (what da- can reach from there). Agree can only look down the tree, so the Merge site limits the Agree targets.

1. The question

  • T54b (from da-reach §11; = T3, T30c): is Pl merged above the inverse landing site? PL-2H probe (ii) ("Agree with the closest φ-bearing goal") reaches the patient in the inverse only if Pl c-commands the moved patient and the moved patient is closer to Pl than the agent.
  • The da-reach derivations simply assumed Pl is merged just after the inverse Move to outer Spec,v (G-1 step 4). The candidate source named for checking it was "Hale 2003 on position III".

2. Sources opened this run (2026-10-23, ~8:46pm PT run)

  • Hale, Ken 2003 ms. "On the significance of Eloise Jelinek's Pronominal Argument Hypothesis" (box copy halejel.pdf). Page images at 150 dpi: printed pp. 17 and 38 (PDF pp. 17, 38), fn. 14 on p. 71 (PDF p. 71). Text layer: pp. 14–20, 39–40, fn. 4–5 pp. 63–64.
  • Hale, Ken 2001. "Navajo verb stem position and the bipartite structure of the Navajo conjunct sector." Linguistic Inquiry 32: 678–693 (box copy hale2001.pdf). Page images at 150 dpi: pp. 686–688 (PDF pp. 9–11). Text layer: p. 689.
  • Hale, Ken. "The Phonological Aspect of the Navajo Conjunct Verb System." Handout, MIT (hale062). Page images at 150 dpi: handout pp. 9–10, (21)–(24).
  • Willie, MaryAnn 1991. Navajo Pronouns and Obviation. Box PDF, page images at 110 dpi: printed pp. 84–85 (PDF pp. 90–91), (66)–(69) re-read for the derivation in §7.

Nothing else was opened. Rice 2000 (the source of the scope principle all three Hale works lean on) is not on the box.

3. The data SOURCED

All forms were transcribed from the page images (the text layers garble ł, ɫ, tone marks and ’). Hale does not give interlinear glosses. His own position-by-position segmentation is shown, with his labels in small capitals where he uses category names. Hale prints the glottal stop as a straight apostrophe, which is rendered ’ here.

Hale 2003 (31a), p. 38 (also Hale 2001 (19), p. 688; Hale 2003 says (31) is "essentially the same as (9-11) above, omitting Position V, the deictic subject")
(1)Yisdánídashidoołtééł. (~ Yisdáádashidoołtééł.)
yisdánídashdwłtééł
to.safety (Pv)REVPL1SG.OBJQUALM/S (3, perfective)VOICEhandle
‘They will get me back to safety.’

Segmentation from Hale 2003 (32) and (10). Hale 2003 p. 40 describes doołtééł as "3 subject will handle animate object"; Hale 2001 (16b) labels the stem "handle inanimate object". The two works disagree; both are reported, neither is chosen.

Hale 2001 (15a), p. 686
(2)daacha
daØØcha
PLM/SVOICEcry
‘They cry.’
Handout (24a, b), p. 10 (Hale cites Y&M87:187)
(3)ʼÁdabiʼniilyaa.
ʼá-da-bi-ʼnii(´)-iid- Ø--laa
thusPLitINCHPERF1NSG VOICEmake
‘We started to make it.’
Willie 1991 (67a, b), p. 84
(4a)mąʼiitłʼízídeishxash
coyotegoatPL-3-3-bit
‘The coyotes bit the goat(s).’
(4b)tłʼízímąʼiidabishxash
goatcoyoteINV3-3-bit
‘The goats were bitten by the coyotes.’

Willie p. 84, as printed: "With dabi-, the internal argument must be plural, and the external argument may optionally be interpreted as plural." p. 85: "In (67b), both must be interpreted as plural." (In Willie's terms the internal argument of the inverse is the agent.) The gloss line of (67b) has no PL, as printed.

4. Hale's three placements of Pl SOURCED

WorkWhere Pl (da) is attachedRelative to the subject piece (M/S) and vWhat Hale says
Handout (23c), (24c), pp. 9–10Adjoined inside V′, next to the stem. In (23c) it is the lowest adjunct (da above V ghal, below náá and the P object). In (24c) da sits under ʼá, above the P object and V laaBelow M, below the subject D iid (Spec,vP), and below vNo comment on Pl's height
Hale 2001 (15b), (16b), (21), pp. 686–689Adjoined to VP, below Voice (v). In (16b) the order from top is Preverb, Iterative, Plural, then [Object V]Below Mode/Subj and VoiceThe adjunct substructure "is a flat structure, properly speaking and despite appearances" (p. 688). Adjuncts "asymmetrically c-command the verb" but there are "no asymmetric c-command relations among the adjuncts themselves" (p. 689). Arguments and adjuncts "belong to distinct dimensions" (p. 689; three dimensions, (20) p. 688)
Hale 2003 ms. (11) p. 17, (31b) p. 38Adjoined to QualP, the maximal projection of the highest nuclear head. Order from top: Pv yisdá, Rev ní, Pl da, then Qual d, M/S w, v ł, [Obj sh V tééł]Above Qual, above M/S, above vThe adjuncts, i.e. "the preverbs and the quantifier elements (the event quantifiers It/Rev and the argument quantifier (Pl))", correspond to the disjunct sector, and "In (11), their high position corresponds straightforwardly to their leftward position in the template" (p. 17). But "the adjuncts are not uniformly high": Rice is right to assign them "sometimes at least" to "the lower position (which I will take to be adjunction to the verbal projection vP, comprising the stem (V) and the classifier (v)" (p. 17). Fn. 14 (p. 71): the quantifier elements can perhaps be adjoined "at VP as well as the higher positions (maximal projection of v or Qual) proposed for Navajo", in "the relevant languages" (other Athabaskan)
  • SOURCED Hale's latest stated position for Navajo is that Pl is adjoined at the maximal projection of v or of Qual (fn. 14), and his worked trees (11) and (31b) use the high option (QualP). The VP-internal placement is mentioned only for other Athabaskan languages.
  • SOURCED Hale's diagnostic for high vs low is semantic scope over the subject, stated for preverbs: in (9) the preverb "has both the subject (in M/S) and the object in its semantic scope … I take this to mean that the preverb is adjoined to one or another of the phrasal projections defined by the functional heads Qualifier and M/S" (p. 16). In (13)–(14) yisdá in ‘I rescue it’ "does not have the subject in its scope", so it is adjoined at vP (p. 19). The text layer of p. 16 and p. 19 was read; the trees were not page-checked.
  • HYPOTHESIS Hale does not say why Pl moved up between 2001 and 2003. Applying his own scope diagnostic gives a reason: in (1) and (2) da counts the subject (‘they’), so it must have the subject in its scope, and the subject is in M/S in his trees. A Pl below Voice and Mode/Subj (2001) would count a subject outside its scope. This motivation is ours.
  • SOURCED The order inside the disjunct sector is set by scope "to the left" (Rice, as reported by Hale 2001:689), and Iterative vs Plural can vary across Athabaskan because "Iterative is relevant to event structure, while (Distributive) Plural is relevant to argument structure" (p. 689). Hale 2003 p. 40: the nonnuclear elements are prefixed "working leftward (upwards in the structure)".

SOURCED (older author's pre-VPISH analysis; contrast only) None of the three trees has the subject in Spec,v. The 2003 trees put the subject in the M/S head (the pronominal argument), and the handout puts the subject D in Spec,vP above v only for (23)–(24). None of them is a flat-S tree, but they must be restated before they can test PL-2H, which is stated with vP/TP slots. That is §5.

5. Restated with VPISH

VPISH: ρA is first-merged in Spec,v and ρP as the complement of V. Under the deep inverse (D-2 on the deep-or-shallow page), ρP moves to an outer Spec,v. The site's own record has two other candidate landing sites: Spec,M/SP (IA-1*) and Spec,QualP (Hale 1997's EP; T30g). The three Hale placements become:

Hale's placement, restatedDoes Pl c-command ρA (Spec,v)?Landing at outer Spec,vLanding at Spec,M/SPLanding at Spec,QualP
Pl(VP): handout, 2001NoPl below ✗Pl below ✗Pl below ✗
Pl(L): 2003 low option, adjoined to vP after the inverse MoveYesPl above ✓Pl below ✗Pl below ✗
Pl(H): 2003 worked trees, adjoined to QualPYesPl above ✓Pl above ✓Depends on order: ✓ if Pl adjoins after Spec,QualP is filled, ✗ if before (OPEN)
  • HYPOTHESIS The restatement keeps Hale's attachment sites and moves only the subject into Spec,v. The ✓/✗ cells are ours: Hale discusses neither the inverse nor its landing site.
  • HYPOTHESIS Pl(VP) is not compatible with PL-2H. Probe (i) could not reach ρA by looking down (DIST-1 would need upward Agree), and probe (ii) would find ρP (the complement of V) as its closest goal in both voices. Then the patient could be counted in the direct too, and D5 (the fifth deep-inverse test) would lose its point, because the patient would be reachable without moving. This is not a contradiction of Hale 2001 itself, whose three-dimension geometry does not define c-command between adjuncts and arguments.
  • HYPOTHESIS Pl(L) is exactly what the da-reach derivations assumed, and it supports PL-2H if the landing site is outer Spec,v.
  • HYPOTHESIS Pl(H) puts Pl above the landing site for two of the three candidate sites. But it puts the M/S head between Pl and both slots. If the M/S head's φ-features count as a goal for probe (ii), then what (ii) finds in the inverse depends on which slot M/S agrees with there, which is the old open question T10b/T30b. If M/S agrees with EXT = ρP, DIST-6 still follows. If it agrees with ρA (W&J's constant 3NOM), probe (ii) is captured by ρA's features and DIST-6 is lost, unless agreeing heads are not goals.
  • OPEN Hale's scope diagnostic no longer separates Pl(H) from Pl(L) once the agent is in Spec,v, because both placements c-command it. A different test is needed (T55a). Update 2026-10-24: this holds only if the subject stays in Spec,v; see pl-move-reason §7 (PL-R4, T56b).

6. Rules

PL-H0  (Hale 2003)  Pl is an adjunct ("argument quantifier"), not a nuclear head.        SOURCED
       Navajo attachment: the maximal projection of the highest nuclear head (QualP, else
       M/SP), or vP.  Worked trees (11), (31b): QualP.
PL-H1  (restated)  With ρ_A in Spec,v, both Navajo options of PL-H0 c-command ρ_A.        HYPOTHESIS
PL-H2  PL-2H needs Pl ∈ {Pl(L), Pl(H)} and not Pl(VP).                                     HYPOTHESIS
PL-H3  Pl(L) + landing at outer Spec,v: DIST-6 follows with no further assumption.          HYPOTHESIS
       Pl(H): DIST-6 follows only if M/S is not an intervener for probe (ii), or M/S
       agrees with EXT in the inverse (T10b, T30b).
PL-H4  Linear check: da- precedes bi- in dabishxash (4b) and in ’Ádabi’niilyaa (3).        SOURCED (order)
       By Hale 2003 p. 40 (nonnuclear prefixes added "working leftward (upwards in the        HYPOTHESIS (reading)
       structure)"), da- is higher than the position-IV material in both.

HYPOTHESIS PL-H4 is weak. In (3) bi- is an oblique P object, not the inverse bi-, and the site has elsewhere treated inverse bi- as a spelled-out head rather than a nonnuclear object; the procliticization order is Hale's for nonnuclear elements only.

7. Derivations, step by step

Labels: Σ role schema, A word (W), B clause. Spine: Qual > M/S > v > V, as on the da-reach page. VPISH throughout. Verb words are segmented only as Hale segments them; Willie's are not segmented.

P-1. Hale (31a) Yisdánídashidoołtééł ‘They will get me back to safety’, with Pl(H)

  1. Σ Σ(yisdá#√HANDLE) = {ρA, ρP}. Preverb and stem form one lexical unit (Hale 2001:678, 688).
  2. A Merge V tééł + ρP (1sg; Hale's Obj sh is its sister). Merge v ł. Merge ρA in Spec,v (VPISH; Hale has it in M/S).
  3. A Merge M/S w. Agree M/S ⟶ ρA: person = 3 (Hale: "third person perfective … Mode/Subj prefix ([w] in this context)", 2001:687). Merge Qual d.
  4. A Merge (adjoin) Pl da to QualP (Hale (31b)). Agree, probe (i): Pl ⟶ ρA, # = pl, hence ‘they’ ✓ (DIST-1). Probe (ii) is not applied (the patient is 1sg; no inverse).
  5. A Merge (adjoin) Rev ní, then Pv yisdá, to QualP. Hale: the preverb has "both the subject … and the object in its semantic scope" (p. 16), so high.
  6. A Spell-out, nuclear plane (Hale 2003 pp. 39–40): ł takes the coda slot, w the onset, epenthetic vowel; Qual d prefixed with epenthesis → diwiłtééł (Hale (33); text layer, read with the same glyph mapping as (31a)) → rounding and glide loss → doołtééł.
  7. A Spell-out, nonnuclear plane: procliticize "working leftward (upwards in the structure)": sh(i), then da, ní, yisdá → yisdánídashidoołtééł. Seal W.
  8. B No overt nouns; the clause is W alone.

Steps 2–3 move the subject into Spec,v, which Hale's trees do not have; the rest follows Hale 2003's derivation of (31) as printed. Ruleset A covers steps 2–7.

P-2. Willie (67b) tłʼízí mąʼii dabishxash, with Pl(H) instead of Pl(L)

  1. Σ Σ(√BITE) = {ρA (biter), ρP (bitten)}.
  2. A Merge V + ρP; Merge v[EPP]; Merge ρA in Spec,v. Move ρP to outer Spec,v (equal rank, RCC allows it).
  3. A Merge M/S. Agree M/S ⟶ ? Here the derivation forks (T10b): (a) M/S ⟶ ρP (closest, after Move), or (b) M/S ⟶ ρA (W&J's constant 3NOM). Merge Voice (spelled bi-), Qual if any.
  4. A Merge (adjoin) Pl to the highest projection. Probe (i): Pl ⟶ ρA, # = pl ✓ (DIST-1; Willie: the internal argument "must be plural").
  5. A Probe (ii), chosen. Fork (a): the closest φ-bearer below Pl is M/S, valued by ρP, or ρP itself in outer Spec,v; either way # (ρP) = pl ✓, and Willie's "both must be interpreted as plural" is derived. Fork (b): the closest φ-bearer is M/S valued by ρA, so probe (ii) only repeats probe (i) and ρP stays unvalued. Then the bare singular EXT tłʼízí fails INV-4, contrary to Willie's ✓ with the plural reading. HYPOTHESIS Fork (b) works only if agreeing heads do not count as goals for (ii).
  6. A Spell-out and Seal → W = dabishxash; da- left of bi-, matching Pl attached higher than the material spelled out as bi- (PL-H4, weak).
  7. B Link (INV-6): mąʼii, next to W, → ρA; tłʼízí → EXT = ρP. Both read plural from the slots ✓.

P-3. Counterfactual: the same sentence with Pl(VP) (Hale 2001, handout)

  1. A Merge V + ρP; Merge (adjoin) Pl to VP; Merge v; Merge ρA in Spec,v; Move ρP to outer Spec,v.
  2. A Probe (i): ρA is above Pl, so downward Agree fails. DIST-1 is underived unless Agree may look up. Probe (ii): the only goal below Pl is the lower copy of ρP, and it would be just as available in the direct (4a), where Willie says the number of the patient is free ("goat(s)") and da- must count the agent ("the agent must be interpreted as plural", p. 85, of (67a)).
  3. Result: Pl(VP) gets the facts backwards for a downward-probing Pl. HYPOTHESIS (within the site's single-tree Agree; Hale 2001's own three-dimension geometry is not refuted by this).

8. T54b: what is settled and what is not

PartAnswerStatus
Where does Hale put Navajo Pl?Latest view (2003): high, adjoined to the top projection of the word (QualP or M/SP), or at vP; worked trees use the high option. Earlier (2001, handout): low, inside VP.SOURCED
Is that above the inverse landing site?Yes for outer Spec,v under both 2003 options, and for Spec,M/SP under the high option. Hale himself does not discuss the inverse.HYPOTHESIS
Does PL-2H survive?Yes with Pl(L), and with Pl(H) if M/S does not capture probe (ii). It fails with the earlier Pl(VP).HYPOTHESIS
Pl(H) or Pl(L) in Navajo inverse verbs?Not decided. Hale's scope diagnostic does not separate them under VPISH.OPEN (T55a)
Earlier site claim (T30g): "CYC orders [the landing site] below Pl either way"True under Pl(H) for outer Spec,v and Spec,M/SP. Not true under Pl(L) if the landing site is Spec,M/SP or Spec,QualP.HYPOTHESIS (correction to the record)

9. Grant's ideas

  • GRANT'S CONJECTURE Verbs take feature bundles, not nouns. Nothing here contradicts it. Hale calls Pl an "argument quantifier" attached inside the word (p. 17), and in every derivation above it counts slots, never nouns. Willie (p. 85) says number belongs to "the pronouns", not the nominals. That is compatible with the conjecture but does not prove it, since a pronominal-argument analysis predicts the same.
  • GRANT'S CONJECTURE Strata (theta roles first, then event structure, then late noun selection). Hale 2001's three "dimensions" (argument structure, adjunction structure, nuclear structure, (20)) look like layers, but Hale presents them as simultaneous, not ordered, so they neither support nor contradict an ordering. Rice's split (Iterative is event structure, Plural is argument structure, Hale 2001:689) does put Pl in the argument (role) layer. The middle stratum is still untested here.
  • OPEN Intra-word Merge/Agree sites. Pl's attachment is word-internal Merge (adjunction) whose height Hale himself revised. Under Pl(H) it is a Merge site at the very top of W, just below the A/B handover, and it hosts the two Agree relations of PL-2H. That is one more candidate for the intra-word Merge/Agree list.
  • Chemistry: no correlate is claimed.

10. Bottom line

Hale's last word on Navajo puts da- high, above the subject piece of the verb (SOURCED). That is high enough for the two-hook analysis of da- to work in the inverse (HYPOTHESIS). His earlier, lower placement would not work with it. Which of his two final options is right, and whether the subject agreement piece gets in the way, is still open.

11. Open

IDQuestionWhat would settle it
T55aPl(H) or Pl(L)? A diagnostic that separates QualP from vP attachment once the agent is in Spec,v.Update 2026-10-24: prefix order ruled out and a causative test proposed on pl-move-reason §7 (T56a). Also: scope of da- against a Qualifier or Mode meaning (e.g. distributive over a future vs a future of a group); Rice 2000 on distributive scope; speakers. OPEN
T55bDoes M/S agree with EXT or with ρA in the inverse (= T10b, T30b)? Decides whether Pl(H) keeps DIST-6.Probably not decidable from M/S spell-out: the inverse needs two third persons (Willie), and third-person M/S has no distinct number, so the two forks spell out alike. Needs an indirect test, e.g. the binding/scope judgments of T53a. OPEN
T55cRice 2000:80–81 (incorporates between the quantifier elements) and Rice's scope principles, read first-hand.Rice 2000, not on the box. OPEN
T55dDoes Hale give a reason anywhere for moving Pl from VP (2001) to QualP (2003)?Answered 2026-10-24 on why did Hale move da- up?: no Pl-specific reason; Pl moves with the adjunct class. SOURCED
T54bIs Pl merged above the inverse landing site?Answered in part 2026-10-23 ~8:46pm PT (§8): SOURCED for Hale's placement, HYPOTHESIS for the inverse. HYPOTHESIS

Sources: Hale 2003 ms. pp. 14–20, 38–40, fn. 4–5, 14 (page images pp. 17, 38, 71); Hale 2001 pp. 686–689 (page images pp. 686–688); Hale handout pp. 9–10; Willie 1991 pp. 84–85. Cross-links: can da- count the patient?, number & scope, deep or shallow inverse?, the inverse as Agree and EPP, headedness, open questions, sources.

Why did Hale move da- up? His reasons, and a causative test for high vs low (T55d, T55a)

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

Restated on the unified structure (2026-11-15). The plural da- is restated as Dist⁰ just above vP, agreeing with the closest argument it c-commands, with the inverse patient counted from the outer Spec,vP (H21a/b). Its place inside the verb word is one of the misfits recorded for the Merge-site account of prefix order (OPEN). The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). Status on the spine: see unified structure, §4, distributive row and §3.2. 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. A rule block written with Hale 2003’s M/S head was not marked as older notation; it is now labeled as a historical record. See the pass 3 report.

Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference.

The pl-height page found that Ken Hale put the plural prefix da- low (inside the verb phrase) in 2001 but high (at the top of the word) in his 2003 manuscript, and asked whether he ever says why (T55d). It also asked for a test that tells "high" from "low" once the doer sits inside the verb phrase (T55a). This page re-reads both Hale works from the page images. In plain words: Hale never gives a reason that is about da- itself. He argues for the high position using preverbs (words like ‘to safety’), then applies the result to all the elements of the front part of the verb, da- included. He also says that the order of the pieces in the word is the same whichever height they sit at, so the order of the prefixes cannot settle the question. His causative examples suggest a test that can: in a "make someone do X" verb there are two doers, and a low da- could count only the one who is made to act. Navajo data with da- in such a verb was not on the box, so that test has not been run yet.

On this page: 1 Question · 2 Sources · 3 Data · 4 What Hale says, 2001 vs 2003 · 5 T55d answered · 6 A scope tension in the 2003 tree · 7 Toward T55a: Hale's height tests carried over to Pl · 8 Rules · 9 Derivations · 10 Grant's ideas · 11 Bottom line · 12 Open

Notation (the site's own; no Syntax · Semantics · Interface bot conventions file has arrived, so the provisional notation is kept). Σ = role-schema step; A = word level (rule set A: the prefixal, left-headed spine, spelled out as one word W); B = clause level (rule set B: right-headed clause syntax, noun linking). W is where A hands over to B. ρA = agent slot (labels ours); in a causative, ρA1 = causer, ρA2 = causee. Pl = the adjunct spelled out da- (template position III). Pl(H) = Pl adjoined to the top projection (QualP, else M/SP); Pl(L) = Pl adjoined to vP; in a causative with two v's, Pl(L1) = adjoined to vP1 (causer's vP), Pl(L2) = adjoined to vP2 (causee's vP). Pl(VP) = Pl inside VP below v (Hale 2001). PL-2H, DIST-1…6: da-reach; PL-H0–H4: pl-height §6.

Reminder: Merge builds structure; attaching da- or a preverb is Merge (adjunction), and moving something already built is internal Merge (Move). Agree values features such as person and number between a head and an argument without moving anything. This page is about where Merge attaches Pl, which decides what Pl's Agree probes can see (Agree looks only down the tree).

1. The question

  • T55d (from pl-height §11): does Hale give a reason for moving Pl from VP-adjoined (Hale 2001 (15b), (16b)) to QualP-adjoined (Hale 2003 ms. (11), (31b))? The pl-height page proposed a reason of our own (Pl counts the subject, so it should have the subject in its scope) and marked it HYPOTHESIS. The named next step was "Hale 2003 pp. 41–62 re-read".
  • T55a: a diagnostic that separates Pl(H) from Pl(L) once ρA is in Spec,v (VPISH), where the plain subject-scope test no longer separates them.

2. Sources opened this run (2026-10-24, ~4:46am PT run)

  • Hale, Ken 2003 ms. "On the significance of Eloise Jelinek's Pronominal Argument Hypothesis" (box copy [local working copy, not in the repository]; PDF page = printed page). Page images at 130–150 dpi: pp. 15, 16, 19, 23, 24, 25, 30, 38, 42, 43, 70. Text layer: all 72 pages searched for Pl, da, plural, quantif, adjunct, dimension, flat, equidistant; pp. 14–30 and 41–44 read in full.
  • Hale, Ken 2001. "Navajo verb stem position and the bipartite structure of the Navajo conjunct sector." Linguistic Inquiry 32: 678–693 (box copy hale2001.pdf). Page images at 130 dpi: pp. 686, 689 (PDF pp. 9, 12). Page image also rendered for p. 688. Text layer: pp. 678–680, 686–689, 691–692.

Nothing else was opened. Rice 2000 and Young & Morgan 1987 (the source of Hale's causative examples) are not on the box.

3. The data SOURCED

All forms were transcribed from the page images (the text layer garbles ł, tone marks and ’). Hale gives no interlinear gloss lines. His own segmentation line is quoted as printed, and the category line below it is taken from his tree for the same example; it is labelled as his. Ø is his.

Hale 2003 (18a), p. 24 (Hale cites Young and Morgan 1987:373); tree (18c), p. 24
(1)habiishyeed
ha#b-y-sh-ł-l-gheed  (Hale's segmentation, as printed)
habyshłlgheed
Pv ‘up out’Obj (obl)QualM/Sv1v2V
‘I run him/her/it up out.’ (idiomatically) ‘I start it (car).’

Category line from Hale's tree (18c): Qual y > M/S sh > v1 ł > [vP2 Pv ha [vP2 Obj(obl) b [v2 l V yeed]]]. The tree prints the stem as yeed; the segmentation line as gheed (p. 24: "y = fronted gamma"). The gloss ‘up out’ for ha is Hale's (p. 43).

Hale 2003 (19), p. 25
(2)nabiishłá
na#b-y-sh-ł-Ø-á  (Hale's segmentation, as printed)
nabyshłØá
Pv ‘round and about’(no tree given; the category line of (1) would apply by analogy, which is ours)
‘I walk him/her (baby) about.’
Hale 2003 (31a), p. 38; tree (31b), p. 38
(3)Yisdánídashidoołtééł. (~ Yisdáádashidoołtééł.)
yisdánídashdwłtééł
PvRevPlObjQualM/SvV
‘They will get me back to safety.’

Category line from Hale's (31b)/(32). Tree (31b), top down: Pv yisdá, then Rev ní, then Pl da, each adjoined to QualP, above Qual d > M/S w > v ł > [Obj sh V tééł]. Already used on pl-height (1); re-checked against the p. 38 image this run.

Hale 2001 (15a, b), p. 686
(4)daacha
daØØCVCchaC
Plural (adjunct plane)Mode/SubjVoiceV
‘They cry.’

Tree (15b): Mode/Subj > Voice > [VP da Plural [VP V]]. Already used on pl-height (2); re-checked against the p. 686 image.

4. What Hale says, 2001 vs 2003 SOURCED

PointHale 2001 (LI 32)Hale 2003 ms.
Where the adjuncts (Pv, It/Rev, Pl) attachVP, below Voice and Mode/Subj ((15b), (16b), (21))"variable": high (the maximal projection of the highest functional head, Qual or M/S) or low (vP) (pp. 16–17); fn. 14: "maximal projection of v or Qual … proposed for Navajo"
What "attach" means geometricallyThe adjunct substructure "is a flat structure, properly speaking and despite appearances" (p. 688). "While the adjuncts, as a group and individually, asymmetrically c-command the verb, there are no asymmetric c-command relations among the adjuncts themselves." "One cannot speak of relative distance from the verb here because adjuncts and arguments belong to distinct dimensions" (p. 689)No such statement. The words dimension, flat and equidistant do not occur in the text layer. Fn. 13, p. 70: "Heretorfore [sic], however, I have regarded the planes as expositional devices only."
Why the preverb attaches where it doesPreverb and verb stem "form a unit in the lexicon, a fact well argued in Rice 2000" (p. 678); the preverb is a sister-like adjunct of V (as Hale 2003:15 summarizes it)p. 15: "In Hale (2001), I adopted this view for all preverbs … However, Sharon Hargus and Siri Tuttle (p.c.) question the correctness of this for preverbs in general, relegating this sister-like relation to Preverb-Verb combinations which are in some sense "truly lexical," unique combinations and idioms." p. 16: "I believe now that productive preverbs are variable, in keeping with their generally adverbial nature. In (9), the most prominent interpretation of the preverb has both the subject (in M/S) and the object in its semantic scope … I take this to mean that the preverb is adjoined to one or another of the phrasal projections defined by the functional heads Qualifier and M/S." "Since there is no evidence for one of these over the other, I will assume that the high adjunction site is the projection defined by the highest functional head present"
Pl specifically"the (distributive) plural [da]" is one of "two additional adjunct categories" (p. 686). Plural "is relevant to argument structure", Iterative "to event structure", and the two "do not ‘‘interact’’ in relation to semantic scope" (p. 689, after Rice)"The adjuncts — i.e., the preverbs and the quantifier elements (the event quantifiers It/Rev and the argument quantifier (Pl) — correspond to the disjunct sector … In (11), their high position corresponds straightforwardly to their leftward position in the template" (p. 17). No separate argument for Pl
MovementNo head movement; surface order by spell-out (pp. 680–688)p. 30: "there is no motivation at all for movement of adjuncts (the elements of the disjunct sector) — these must be alternatively base generated in the high or low position in any event."
Does height show in the word?Order among adjuncts is set by "scope alone", right to left (p. 689)p. 43: "Notice that the surface position of adjuncts does not vary according to whether they are attached high (above the subject) or low (below the subject). In either case, adjuncts precede the overt nuclear elements in the verbal projection"
Tests for low attachment (preverbs)—(a) subject outside the preverb's scope, (12)–(14), p. 19: "It is the coat alone that "comes to be in safety""; (b) idiomatic/"lexical" Pv+V, (15)–(16), with "thin ice" caveat (p. 21); (c) causatives, (17)–(18): "the preverb … must be lower than the surface subject (sh) within the derived causative configuration" (p. 23); but (19) has "both the agent (grammatical subject) and the "causee" (derived Obj) within the scope of the preverb" (p. 25); (d) labile transitives, (20)–(21), p. 26

5. T55d answered

SOURCED Hale gives no reason specific to Pl. The 2003 argument for the high site is made for preverbs: the preverb in (9) = (3) has the subject in its scope (p. 16), and Hargus and Tuttle's objection means the 2001 "sister of V" analysis holds only for lexical Pv+V units (pp. 15–16). On p. 17 the result is stated for "the adjuncts" as a class, which includes "the argument quantifier (Pl)". The only Pl-specific statements in the manuscript are that label (p. 17), the trees (11) and (31b), the planar display (32), and fn. 14 (Pl and It may also adjoin at VP in other Athabaskan languages). This is a text-layer search plus page images of the key pages; a garbled passage could in principle have been missed.

HYPOTHESIS Why Pl moved anyway: three ingredients, all Hale's, put together by us.

  1. Class membership. Pl belongs to the disjunct sector with the preverbs (2001:686; 2003:17). Once Hale makes the attachment of "the adjuncts" variable, Pl moves with the class.
  2. Height became a real claim. In 2001 the VP-adjoined position was inside a separate "adjunction structure" dimension, in which distance between adjuncts and arguments is not defined (p. 689). So "Pl below Mode/Subj" in (15b) made no scope claim about the subject. The 2003 manuscript drops the dimensions (none mentioned) and reasons from ordinary c-command and scope ("a projection which properly includes the subject", p. 16). Under that reading, a VP-internal Pl would be a real claim, and a false one for da- counting the subject, as in (4) and (3).
  3. No movement. Adjuncts are "base generated in the high or low position" (p. 30), and spell-out puts them in front whatever their height (p. 43). So the leftward surface position neither requires nor rules out high attachment, and the height has to be chosen at Merge on interpretive grounds.

Ingredient 2 is the step that turns the pl-height page's own motivation (Pl counts the subject) into a likely reason. It is still ours: Hale does not connect the dropped dimensions to Pl.

6. A scope tension in the 2003 tree

SOURCED Hale 2001 p. 689 (after Rice): "semantic scope is to the left, so that the iterative [ná ~ ní] and the plural (her distributive plural) [da] have the preverb within their scope"; "an element with wide scope is to the right of elements appearing within its scope." Hale 2003 p. 17 keeps this: the order among the adjuncts, "a side issue here, is determined by Rice's principles of semantic scope (Rice 2000; cf. Hale 2001)." But in trees (11) and (31b) the preverb yisdá is adjoined outermost, with Rev ní below it and Pl da closest to Qual.

HYPOTHESIS If (31b) is read as one ordinary tree, the way Hale reads it on p. 16 for the preverb and the subject, then Pv c-commands Pl and should take Pl in its scope. That is the reverse of the Rice scope statement Hale still cites. The two can be reconciled only if the height of the adjunct group relative to the nuclear heads and the argument slots is real, while the order inside the group is a scope-driven linearization that is not read off c-command. That is the 2001 flat structure surviving inside the 2003 tree. We state this as rule PL-R2. Hale does not flag the tension; this is our reading.

HYPOTHESIS PL-R2 does not hurt PL-2H. PL-2H needs only Pl's c-command over ρA and the inverse landing site, which is the group's height, not its internal order.

7. Toward T55a: Hale's height tests carried over to Pl

SOURCED Template position III is not evidence for Pl's height, because adjunct surface position "does not vary according to whether they are attached high … or low" (p. 43). Only meaning-based tests can decide. This sets aside the linear reading in pl-height PL-H4 as evidence about height (PL-H4 was already marked weak).

HYPOTHESIS A VPISH caveat on Hale's low sites, and a correction to the record. Hale's low site is vP, and it counts as "below the subject" only because his subject is in M/S ((14), (21)). Under VPISH ρA is first-merged in Spec,v, inside vP. Then a vP-adjoined preverb c-commands the subject's first position, yet (13) has the subject outside the preverb's scope. Two repairs: (i) the subject raises to Spec,M/SP (where Hale's M/S agreement sits) and scope is read from that copy; or (ii) the low site is below Spec,v (v′ or VP). Hale himself gives the subject an underlying Spec,v position and says it ordinarily enters "the usual portmanteau relation with Mode" (p. 42, SOURCED), which is closer to (i), but he does not say where scope is read. The pl-height page said the scope test cannot separate Pl(H) from Pl(L) under VPISH (its §5). That holds only if the subject stays in Spec,v. Under repair (i) the test separates them again: Pl(L) would be below the raised subject. This is T56b.

HYPOTHESIS The causative test. Hale's causatives contain two agent-like slots: the causer (his M/S sh; VPISH Spec,v1) and the causee (his Obj(obl) b in Spec,v2, (18c)). He says the causee "cannot fuse with Mode" and is "trapped within the v-projection" (p. 43), so it stays low under either repair. Hale shows that a preverb can attach inside vP2 ((18), scope over the causee only) or high ((19), scope over both). Carry this over to Pl with PL-2H probe (i) (Agree, downward, with the slot bearing the most Proto-Agent entailments):

Pl attached atSlots Pl c-commandsProbe (i) values # onPredicted reading of a da- causative
Pl(H): QualPρA1, ρA2, objectρA1 (causer: more Proto-Agent entailments)causers plural
Pl(L1): vP1ρA1 (first position), ρA2ρA1causers plural
Pl(L2): vP2 (Hale's site for ha in (18c))ρA2 onlyρA2causees plural, causer free
Pl(VP): inside VP2 (Hale 2001's site)no agent slotnothing (DIST-1 fails)no reading via probe (i)

HYPOTHESIS So the causative test separates Pl(L2) from {Pl(H), Pl(L1)}. It is the only test found so far that a da- verb can be put through without speaker judgments of scope over a Mode meaning: it asks who must be more than one, which grammars and dictionaries often record. It does not separate Pl(H) from Pl(L1); that still needs a Qual or Mode scope contrast (T55a's original suggestion) or repair (i) above.

OPEN No Navajo causative with da- was opened. Hale gives none, and Young & Morgan 1987 (his source for (17)–(19)) is not on the box. No form is supplied here (T56a).

8. Rules

OPEN Older notation: the block below uses Hale 2003’s M/S (Mode/Subject) head, which is not a node in our current trees. It is kept as a historical record, not as one of our analyses.

PL-R0  Hale 2003 argues for high adjunction on preverbs (subject scope, (9), p. 16; Hargus &  SOURCED
       Tuttle p.c., pp. 15–16) and states the result for "the adjuncts", Pl included (p. 17).
       There is no Pl-specific argument in the manuscript.
PL-R1  Adjuncts are base-generated high or low, never moved (p. 30); their surface position   SOURCED
       is the same at either height (p. 43).
       ⇒ template position III is no evidence for Pl's height.                              HYPOTHESIS (corollary)
PL-R2  The adjunct group's height is a real c-command relation to nuclear heads and slots;    HYPOTHESIS
       the order inside the group is scope-ordered linearization (Rice; Hale 2001:689), not
       c-command.  (Reconciles (11)/(31b) with Hale 2003:17.)
PL-R3  In a causative [Qual [M/S [vP1 ρA1 v1 [vP2 ρA2 v2 VP]]]], Pl(L2) values # on ρA2    HYPOTHESIS
       only; Pl(L1) and Pl(H) value # on ρA1 (PL-2H (i), downward).
PL-R4  Under VPISH, Hale's narrow-scope low adjuncts ((13), (20b)) need either subject        HYPOTHESIS
       raising to Spec,M/SP with scope read there, or a low site below Spec,v.
       With raising, subject scope separates Pl(H) from Pl(L) again (correction to
       pl-height §5 OPEN note).

9. Derivations, step by step

Labels: Σ role schema, A word (W), B clause. VPISH throughout: each agent-like slot is first-merged in its own Spec,v. Hale's subject sits in M/S; we put it in Spec,v1 and let M/S Agree with it. Only Hale's segmentations are used; the spell-out steps for (1) are his (p. 43), as printed.

Q-1. Hale (18a) habiishyeed ‘I run him/her/it up out’: low preverb in a causative

  1. Σ Σ(CAUS + ha#√RUN) = {ρA1 (causer), ρA2 (runner, causee)}. The embedded verb is unergative (Hale p. 23: "the complete (unergative) verbal projection (vP2, headed by v2 l, the inner classifier)").
  2. A Merge V yeed with v2 l. Merge ρA2 in Spec,v2 (Hale's Obj(obl) b is already there in (18c); p. 42: "The basic or underlying position of the subject of a v-projection is Specifier in that projection").
  3. A Merge (adjoin) Pv ha to vP2 (Hale (18c)). Its scope is vP2: "it is the car … which produces the "crying out"" (p. 23, said of (17); "the same argument is applicable" to (18)).
  4. A Merge v1 ł (the causative light verb). Merge ρA1 (1sg) in Spec,v1 (VPISH).
  5. A Merge M/S. Agree M/S ⟶ ρA1: person 1sg, spelled sh. ρA2 gets no Mode agreement and is oblique (Hale p. 42: "the subject of v will not enter into the usual portmanteau relation with Mode"; p. 43: it "cannot fuse with Mode"). Merge Qual y (the causative qualifier, p. 22).
  6. A Spell-out, nuclear plane (p. 43): after the inner and outer classifiers and M/S are placed in the skeleton (p. 42), "there is just the causative qualifier y in the nuclear plane; hypothetically this assumes the onset position, giving now the intermediate sequence *yishyeed".
  7. A Spell-out, nonnuclear plane: "prefixation of Obj(obl) gives hypothetical *biyishyeed which, by intervocalic glide deletion, surfaces as actual biishyeed. To this is prefixed the preverb ha" → habiishyeed. Seal W.
  8. B No overt nouns; the clause is W alone.

Ruleset A does steps 2–7; ruleset B has nothing to do. Step 7 shows PL-R1 at work: ha is attached in vP2, below the causer, yet it comes out leftmost.

Q-2. Hale (19) nabiishłá ‘I walk him/her (baby) about’: high preverb in a causative

  1. Σ Σ(CAUS + na#√WALK) = {ρA1 (causer), ρA2 (walker, causee)}.
  2. A As Q-1 steps 2, 4, 5, with inner v2 Ø and stem á (Hale's segmentation), and no adjunction inside vP2.
  3. A Merge (adjoin) Pv na to QualP. Hale: "both the agent (grammatical subject) and the "causee" (derived Obj) within the scope of the preverb. The agent manipulates the causee at every step and thus "moves round and about"" (p. 25).
  4. A Spell-out and Seal → nabiishłá (surface form as printed; Hale gives no step-by-step for (19)). Same linear slot for na as for ha in Q-1.

Q-1 and Q-2 together are Hale's evidence that one adjunct category can attach at either height in the same causative frame, with only the meaning to tell them apart.

Q-3. Schematic: a causative with Pl at three heights (no Navajo form)

Schematic word: [PL + CAUS + V], written DA-CAUS-V. No such Navajo word was opened, and none is supplied.

  1. Σ {ρA1 causer, ρA2 causee}.
  2. A Build vP2 (ρA2 in Spec,v2), then vP1 (ρA1 in Spec,v1), then M/S (Agree with ρA1), then Qual, as in Q-1.
  3. A Fork on where Pl is Merged: (L2) adjoined to vP2, right after step 2's vP2; (L1) adjoined to vP1; (H) adjoined to QualP.
  4. A Agree, PL-2H probe (i), looking down from Pl: (L2) the only agent slot in reach is ρA2 → causee plural; (L1), (H) ρA1 and ρA2 are both in reach, and ρA1 has more Proto-Agent entailments (it causes) → causer plural.
  5. A Spell-out: da- in position III in all three (PL-R1). Seal W.
  6. B Any nouns link to the slots after W is sealed. The plural reading comes from the slot, so a noun's own number cannot decide among the forks; the reading of who must be several can.

10. Grant's ideas

GRANT'S CONJECTURE Verbs take feature bundles, not nouns. Nothing here contradicts it. In Q-1 the causee is a pronominal element that "cannot fuse with Mode" (Hale p. 43), so its case depends on which head its bundle can combine with, not on any noun. In Q-3 Pl reads the slots' role entailments, never a noun. This is compatible with the conjecture; a pronominal-argument analysis predicts the same, so it is not proof.

GRANT'S CONJECTURE Strata (theta roles first, then event structure, then noun selection). Q-3 needs the role schema (Σ) fixed before Pl is merged, because probe (i) reads Proto-Agent entailments, and nouns enter only at B. That order fits the conjecture. The middle stratum gets no test: Hale 2001:689 (after Rice) says It/Rev (event structure) and Pl (argument structure) "do not ‘‘interact’’ in relation to semantic scope", so their relative order says nothing about which layer comes first. Neutral, not contradicting.

HYPOTHESIS The word as the A/B changeover. PL-R1 says rule set A linearizes adjuncts the same way at any height: nuclear heads fill the skeleton first, then nonnuclear material is prefixed (pp. 30, 43). Height matters for meaning (Agree, scope) but not for order inside W. On the site's terms, A's word-internal order is plane-driven, while B's clause order is not computed from the word's structure at all.

OPEN Intra-word Merge/Agree sites: vP2 adjunction in causatives (Q-1, Q-3 (L2)) is a candidate Merge site below an agent slot inside W. If da- can use it, it would be a word-internal Agree site that counts the causee alone. Chemistry: no correlate is claimed.

11. Bottom line

Hale never says why da- in particular moved up; he moved the whole class of front-of-verb elements on the evidence of preverbs (SOURCED). The likely reason it had to go with them is that his 2003 trees treat height as real scope, which his 2001 flat "adjunct dimension" did not (HYPOTHESIS). Prefix order cannot tell high from low (SOURCED, p. 43). A causative with da- could, by showing whether the causer or only the causee has to be plural (HYPOTHESIS; data not yet found).

12. Open

IDQuestionWhat would settle it
T55dHale's reason for moving Pl from VP (2001) to QualP (2003).Answered 2026-10-24 ~4:46am PT (§5): no Pl-specific reason (SOURCED); inherited from the preverb argument and the dropped dimensions (HYPOTHESIS). SOURCED
T55aPl(H) or Pl(L)?Narrowed (§7): prefix order ruled out (PL-R1); the causative test separates Pl(L2) from {Pl(H), Pl(L1)} (PL-R3); Pl(H) vs Pl(L1) still needs a Qual/Mode scope contrast or T56b. OPEN
T56aA Navajo causative with da-: must the causer be plural, or can the causees alone be plural?Young & Morgan 1987 causative entries (not on the box); Willie or Hale examples with da- + causative; speakers. OPEN
T56bUnder VPISH, do Hale's narrow-scope low adjuncts ((13), (20b)) require subject raising to Spec,M/SP (PL-R4 (i)) or a site below Spec,v (ii)?A Navajo test of the subject's height (binding, T53a), or Hale's own discussion of where scope is read. OPEN
T56cDoes Hale (or Rice) address the clash between (31b)'s Pv-outermost tree and leftward scope (PL-R2)?Rice 2000 (not on the box); other Hale manuscripts. OPEN

Sources: Hale 2003 ms. pp. 15–30, 38, 41–44, 70 (page images pp. 15, 16, 19, 23, 24, 25, 30, 38, 42, 43, 70); Hale 2001 pp. 678–689 (page images pp. 686, 689). Cross-links: how high is da-?, can da- count the patient?, plain ł-causer, spell-out rules, headedness, open questions, sources.

Number and quantifiers across the inverse: da- follows the agent

2026-11-11 audit pass 2. Apostrophe fixes only. Navajo glottal stops typed as ' or ’ (for example in prefix names like ʼa- and ʼeʼe-, or words like Awééʼ) now print as ʼ (U+02BC), and a build step no longer turns the closing quote of an English translation inside a Navajo row into a glottal stop (‘…the girls.’, not ‘…the girls.ʼ). No form, gloss or translation wording changed.

Working note. Kept as written (URL, data and status labels unchanged). Filed under The roles-vs-words thesis; the consolidated argument is the essay roles-not-words. Rule IDs: rules index.

Verbs on this page → dictionary: yiztsʼǫs ‘kiss’ · yishxash ‘bite’ · yiyiisxį́ ‘kill (one)’ · dabiiłtsą́ ‘see’ · yíghááh ‘join; with -tah ‘among’: go among’

Restated on the unified structure (2026-11-15). The plural da- is restated as Dist⁰ just above vP, agreeing with the closest argument it c-commands, with the inverse patient counted from the outer Spec,vP (H21a/b). Its place inside the verb word is one of the misfits recorded for the Merge-site account of prefix order (OPEN). The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). Status on the spine: see unified structure, §4, distributive row and §3.2. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

The inverse page left one test open (T5c): if VPISH raising is real, the inverse should move the patient slot above the agent slot, and that should show up in scope. This page reads the part of Willie 1991 §3.3 that the inverse page had not page-checked, pp. 78–86: plural nominals, a universal quantifier, and the distributive prefix da- in both voices. The main result is about da-. In both voices it counts the agent. It does not count the subject. The quantifier pair does not settle T5c, but it narrows it to one question for a speaker. New open questions are T10a–T10f.

On this page: 1 Question · 2 Data · 3 Rules · 4 Role, not position · 5 What it means for VPISH · 6 Derivations N1–N4 · 7 Verdict · 8 Open

1. The question

GRANT'S CONJECTURE In Navajo the units that are ranked and moved are theta roles, not lexical items. Proto-role entailments are fixed by the event. The inverse changes only which argument is realized where.

Two questions follow from the inverse page.

  • T5c. Under bi-, does the external patient take scope over the internal agent, when it does not do so under yi-? V-P (VPISH raising) predicts that it does. V-H (Hale's in-situ licensing) predicts no difference.
  • T5a / T7. What does each piece of the verb word track: a position (EXT, the external argument or subject) or a role (ρA, the slot with the most Proto-Agent entailments)? The inverse is the one place where the two come apart, because under bi- EXT ≠ ρA.

Willie's own framing (p. 80): "In Navajo, we see number marked obligatorily on the verb; this is where the pronouns are." Number marking is therefore a probe that sits inside the word. We can ask which slot it reads in each voice.

2. Data SOURCED

Source opened this run (2026-10-09). Willie, MaryAnn 1991, Navajo Pronouns and Obviation, PhD dissertation, University of Arizona (UMI 9210298; the scan already on the box). We read page images of printed pp. 78–86 (PDF pp. 84–92) at 150 dpi. Every Navajo line of (54)–(69) was checked against 300 dpi crops. We also read printed pp. 28 and 31–32 (PDF pp. 33, 36–37) for her first statement about da-. The OCR text layer is garbled for Navajo and was used only to find pages.

Transcription. The typescript prints the glottal stop as a straight apostrophe. It is given here as ’, as on the other pages of this site. Nasal hooks and accents are as printed. The barred ł is clear on every line used here. Glosses are copied exactly as printed (PL.3.3.saw, PL.INV3.3.saw, PL-3-3-bit, INV3-3-bit, 3.among PL.bit, DEM.). Willie's typescript is not consistent between dots and hyphens, and it is not normalized here. Where Willie prints no gloss line, none is added except word glosses that Willie herself prints for the same word on a nearby page, and these are labelled. No verb is segmented beyond Willie's gloss. W&J 2000's gloss for the same prefixes, PL-3ACC-3NOM-saw for dabiiłtsą́ (W&J (70), already on the fourth-person page), is cited only for comparison.

2.1 Willie's first statement about da- (Chapter 2)

Willie 1991:28, 32, (9)
(9)dayiiłtsą́
PL.3.3.saw
‘They saw him/them.’ *‘He saw them.’

p. 28: "Arguments are marked singular, dual and plural in number. If the number of the subject is three or more, the Distributive prefix da- is added to the dual number of the verb." p. 32: "In direct constructions, when da- appears, the subject must be plural (three or more), and the number of the object is irrelevant. Da- is not an argument, but marks a semantic feature of the subject argument." Note the qualifier "in direct constructions". Chapter 3 (below) supplies the inverse.

2.2 Plural nominals without da-: either argument, either voice (54)–(57)

Willie 1991:78–79, (54)–(57)
(54)Johnʼatʼéékéyiztsʼǫs‘John kissed the girls.’
(55)Johnʼatʼéékébiztsʼǫs‘John was kissed by two girls.’
(56)ʼatʼéékéJohnyiztsʼǫs‘Two girls kissed John.’
(57)ʼatʼéékéJohnbiztsʼǫs‘Those two girls were kissed by John.’
No gloss lines in the source. Willie glosses ʼatʼééké as ‘girls’ in (63b), p. 82.

pp. 78–79, after (55): "(With more than two, the distributive must be used.)" After (57): "Note that the external patient argument is interpreted as definite. The paucal plural (less than three) can appear with either argument in either the Direct or Inverse." Linking check (INV-6, the noun next to the verb is internal): in (55) the girls are next to biztsʼǫs and so are the internal agent; in (57) they are outer, so they are EXT = patient, and only (57) gets the definite ‘Those’.

2.3 A universal quantifier on the external argument (58)–(60)

Willie 1991:79–80, (58)–(60)
(58)tʼááʼałtsoʼatʼéékéʼashkiiłaʼbiztsʼǫs
‘All the girls were kissed by a boy.’
(59)tʼááʼałtsoʼatʼéékéʼashkiiłaʼyiztsʼǫs
‘Each of the girls kissed a boy.’
(60)tʼááʼałtsoʼatʼéékéʼashkiiłaʼdeiztsʼǫs
a. ‘Every girl kissed a boy. (one boy)’ b. ‘Every girl kissed a boy. (many boys)’
No gloss lines in the source. Word glosses found elsewhere: tʼáá ʼałtso ‘just all’ (W&J 2000:276 (70), gloss as printed there); ʼashkii ‘boy’ (Willie (61)); łaʼ is translated ‘a’ here and is the indefinite that (74c) bars from the inverse subject (inverse §2.4).

p. 79: "When the universal quantifier has scope over the external argument, the voice alternation provides different interpretations of the quantifier". (59) is introduced as "With the Direct". p. 80, after (60): "With greater plurals, the distributive da- must be used." Willie does not segment or gloss deiztsʼǫs. That it is the da- form is our reading, from her next sentence and from (67a) deishxash, which she glosses PL-3-3-bit. What she does not say: she does not say whether (58) allows different boys for different girls, or whether (59) allows one boy. Only (60) is given both readings.

2.4 da- with both voices: ‘see’ (61)–(66)

Willie 1991:81–84, (61)–(66)
(61)ʼashkiiʼatʼéédyooʼį́
boygirl3.3.sees
‘The boy sees the girl.’
(62)ʼatʼéédʼashkiibooʼį́
INV3.3.sees
‘The girl is seen by the boy.’
(63a)dayiiłtsą́
PL.3.3.saw
‘They saw them/it/him/her.’
(63b)ʼashiikéʼatʼéékédayiiłtsą́
boysgirlsPL.3.3.saw
‘The boys saw the girls.’
(63c)ʼashiikéʼatʼééddayiiłtsą́
boysgirlPL.3.3.saw
‘The boys saw the girl.’
(63d)*ʼashkiiʼatʼéékédayiiłtsą́
(The boy saw the girls.)
(64a)dabiiłtsą́
PL.INV3.3.saw
‘They/he (was)were seen by them/*him.’
(64b)shizhéʼéʼashiikédabiiłtsą́
1s.fatherboysPL.INV3.3.saw
‘My father was seen by the boys.’
(65a)ʼashiikéʼatʼéékédabiiłtsą́
boysgirlsPL.INV3.3.saw
‘The boys were seen by the girls.’
(65b)*ʼashiikéʼatʼééddabiiłtsą́
(The boys were seen by the girl.)
(65c)*ʼatʼéédʼashiikédabiiłtsą́
(The girl was seen by the boys.)
(66)ʼéidíʼatʼéédʼashiikédabiiłtsą́
‘That girl (over there) was seen by the boys.’

p. 81: "If we add the distributive plural daa/da, then certain sentence types are excluded." p. 82, on (63): "the distributive marking is acceptable when the noun modifying the agent pronoun is plural, but excluded when it is not, as in (63d). It appears that the agent falls under the scope of the distributive marker, and it does not not seem to matter whether the internal argument, the patient in the Direct construction) is plural or singular (63b & c)." [sic, both the doubled "not" and the unmatched parenthesis.] p. 82: "In the Inverse, the agent is the internal argument." p. 83: "With common nouns, da- plus bi- requires that both internal and external arguments be plural." "However, if the external argument is made maximally referential with demonstratives, then the sentence becomes acceptable." p. 84: "I conclude that bi- requires the external argument to be definite, but when the Distributive da- is present, it can license a plural external argument, which can escape the requirement on definiteness. With dabi-, the internal argument must be plural, and the external argument may optionally be interpreted as plural."

2.5 Nouns with no plural form: ‘bite’ (67)–(69)

Willie 1991:84–85, (67)–(69)
(67a)mąʼiitłʼízídeishxash
coyotegoatPL-3-3-bit
‘The coyotes bit the goat(s).’
(67b)tłʼízímąʼiidabishxash
INV3-3-bit
‘The goats were bitten by the coyotes.’
(68)mąʼiitłʼízíyitahdaʼashxash
coyotegoat3.amongPL.bit
‘The coyote (went among and) bit the goats.’
(69)ʼéidítłʼízímąʼiidabishxash
DEM.goatcoyoteINV3.3.bit
‘That goat there was bitten by the coyotes.’

p. 84: "Where there are no plural forms for the noun, these same constraints operate on the permitted interpretations of sentences." p. 85: "In (67a), the agent must be interpreted as plural. In (67b), both must be interpreted as plural. If we want to produce a reading in which the agent may be singular, then the following construction is used." [(68)] "We have seen that number can be associated with different arguments according to the voice alternation. This would not be possible if number were a property of the nominals, rather than the pronouns. In the Inverse voice, the external argument, the patient, must be either definite or plural if the sentence is to be acceptable." As printed: (67b)'s gloss line has no PL even though the form begins da-. The gloss of (67a) uses hyphens and the gloss of (69) uses dots.

3. The rules

Notation as on the inverse page. Σ = {ρA, ρX}. EXT is the slot realized as the external argument. #(ρ) is the number of the slot's referent: sg, du (two, "paucal"), or pl (three or more). Pl is the head spelled out as da- (Hale 2003's template position III, "(Distributive) Plural"; see headedness §2.2).

DIST-1  Agent number.   Pl present  →  #(ρ_A) = pl,   in BOTH voices.
DIST-2  Free patient (direct).   Under yi- + Pl, #(ρ_P) is unconstrained (sg or pl).
DIST-3  Inverse subject.   Under bi- + Pl, EXT (= ρ_X) is pl, or else "maximally referential"
        (demonstrative; 1s-possessed kin term passes).  A bare singular common noun fails.
DIST-4  Paucal.   #(ρ) = du can be expressed without Pl, on either slot, in either voice.
        #(ρ_A) ≥ 3 requires Pl ("With more than two, the distributive must be used").
DIST-5  Domain.   DIST-1 holds of the yi-/bi- transitive frame.  In (68) Pl co-occurs with a
        singular agent when the second participant is a postpositional object (yitah).
Q-1     Universal on EXT.   t’áá ’ałtso on the outer (EXT) nominal is possible in both voices;
        an internal ła’ nominal is possible in both voices.
RuleStatusBasis
DIST-1SOURCEDDirect: p. 32 (9) *‘He saw them’; (63d); (67a) "the agent must be interpreted as plural". Inverse: (64a) ‘by them/*him’; (65b); p. 84 "With dabi-, the internal argument must be plural"; (67b).
DIST-2SOURCEDp. 32 "the number of the object is irrelevant"; (63b) vs (63c); (67a) ‘goat(s)’.
DIST-3SOURCED for plural and demonstrative; HYPOTHESIS for why (64b) passes(65a) ✓, (65c) *, (66) ✓, (69) ✓; p. 84; p. 85 "either definite or plural". (64b) shizhéʼé ✓ with a singular EXT. Willie's restriction is stated "With common nouns", and we take a 1st-person-possessed kin term to fall outside it. That is our reading.
DIST-4SOURCEDpp. 78–80, (54)–(57); "(With more than two, the distributive must be used.)"; "With greater plurals, the distributive da- must be used."
DIST-5SOURCED (fact); what da- counts in (68) is OPEN(68), p. 85. Willie glosses the verb only as PL.bit, and the second prefix is not glossed. T10d.
Q-1SOURCED (strings and translations); its scope readings are OPEN(58), (59), (60). Only (60) has both readings stated.

Two kinds of requirement inside one word HYPOTHESIS. Compare DIST-1 with INV-4 (the inverse subject must be definite, Willie pp. 76–77):

RequirementWhich slot it reads under yi-Which slot it reads under bi-So it tracks
INV-4 definiteness (and DIST-3)(none: either may be indefinite, p. 76)EXT = patientposition (the external argument)
DIST-1 da- pluralρA = EXTρA = internalrole (the proto-agent slot)

The voice alternation moves EXT and leaves da- on the agent. If only position mattered, da- would follow the subject into the patient under bi-, and (64a) would mean ‘he was seen by him’ with a plural ‘he’. Willie stars that reading's counterpart (‘by *him’).

Update 2026-10-23 (~12:46pm PT): DIST-1 and DIST-3 are combined on can da- count the patient? into DIST-1c (never the patient alone; T53d), DIST-6 (an optional patient plural in the inverse only) and rule PL-2H (two probes on Pl; HYPOTHESIS).

4. Role, not position: what da- shows

  1. SOURCED Willie states the Chapter 2 generalization over the "subject", and limits it to "direct constructions" (p. 32). In Chapter 3 she restates it over the agent ("the agent falls under the scope of the distributive marker", p. 82) and shows that it holds when the agent is the internal argument (p. 84).
  2. HYPOTHESIS In the site's notation, da- places a condition on ρA, the slot picked out by Dowty's Proto-Agent count (C5). It does not place one on EXT. That is a generalization stated over a role slot, and grammatical relation is the wrong variable. This is Grant's picture, in which the verb's arguments are proto-role slots and not lexical items. The nominals here are adjuncts, and their number only has to match (Willie p. 85: number is "a property of … the pronouns").
  3. HYPOTHESIS Bearing on T7 (is there a generalization over entailments that pronouns alone cannot state?). It depends on T5a.
    • If Willie's case analysis is right (inverse agent = internal, Ergative, p. 72), then the agent pronoun has a different case in each voice. In that case "da- pluralizes the agent" cannot be stated over one case or one position. It has to be stated over the role. That would make da- the site's first T7 case.
    • If W&J's constant gloss is right (PL-3ACC-3NOM for dabiiłtsą́, with the agent as 3NOM in both voices), then "da- pluralizes the NOM pronoun" says the same thing as DIST-1. In that case da- is not a T7 case.
    So da- does not settle T5a. It does tie T5a to T7: whichever analysis of the inverse wins also decides whether da- is a role-level fact.
  4. SOURCED + HYPOTHESIS Rice (2000), as summarized on the headedness page, puts the distributive plural in the argument system, not the event system (Hale 2001:689; Bird 2001). DIST-1 is compatible with that, and it says which argument: the proto-agent. That is the same slot in both voices.

5. What it means for VPISH

5.1 da- as a locality test HYPOTHESIS

Treat Pl (da-) as a probe that values number by Agree with the closest φ-bearing slot it c-commands. Where is Pl in the tree? Under rule set A, functional heads precede their complements. Position III (da-) is left of the object position IV and of Mode/Subject (VII–VIII), so A reads Pl as structurally above M/S. We do not know that Pl is there. Rice's scope ordering would read the left edge as lower scope (T3). So the argument below is conditional.

Direct (yi-)Inverse (bi-)Does "closest goal = ρA" hold?
V-H (no A-movement; Hale 2003:29)ρA in Spec,v c-commands ρP in Comp,Vthe same: nothing movesYes, in both voices. DIST-1 follows from first-Merge positions.
V-P (VPISH raising), Pl above M/SρA raises to Spec,M/SP and is closestρX raises to Spec,M/SP past ⟨ρA⟩ and is closestNo under bi-. This version predicts that da- pluralizes the patient. That is wrong for (64a) and (65b).
V-P, Pl inside vP (merged before ρX raises)Pl Agrees with ρAPl Agrees with ρA, and then ρX raisesYes. The order of operations is Agree(Pl, ρA) before Move(ρX). da- then reaches the left edge by a non-structural rule, as Hale's skeleton filling does for the object prefix.
V-P, Pl role-indexedPl selects the slot with the most Proto-Agent entailments (a Σ-level condition), not the closest goalYes, by stipulation. This is the T7 reading of §4.

HYPOTHESIS So da- does not refute VPISH. It does remove one version of it, namely raising of the inverse patient to a position below a high Pl that probes for the closest goal. Any V-P analysis has to say which of the last two rows it takes. V-H gets DIST-1 for free. Note that derivation I1 step 6 on the inverse page ("Agree M/S ⟶ EXT") is one more place where this matters: if da- piggybacks on Mode/Subject agreement, M/S must agree with ρA, not with EXT, in the inverse. That matches W&J's 3NOM gloss (§4, item 3) and conflicts with I1 as written. OPEN (T10b).

Update 2026-10-16 (8:46pm): the rows above are re-run with the site's spell-out rule CYC on the inverse as Agree and EPP, K5. The “Pl inside vP” escape (row 3) gets the agreement right, but then CYC puts the raised object prefix before da- (*bi-da-), against dabiiłtsą́, unless the lower copy is pronounced. Raising therefore needs da- to target the agent's role feature (e2), bi- to be an agreement head as in Hale 1997 (e3), or da- to copy M/S's number (e4). Under Bruening's Move-before-Agree order, M/S agrees with ρA in both voices, which resolves the I1 conflict noted here.

5.2 The quantifier pair and T5c OPEN

(58) and (59) are a minimal pair for voice. The quantified nominal tʼáá ʼałtso ʼatʼééké is the outer, EXT nominal in both. The indefinite ʼashkii łaʼ is next to the verb, and so it is internal in both. Only the roles swap. The girls are patients in (58) and agents in (59).

Reading of (58) … biztsʼǫsV-P predictsV-H predicts
R1 ∀girl ∃boy (boys can differ)surface scope: EXT ρP c-commands ⟨ρA⟩ → availableonly as inverse scope (ρA in Spec,v is higher) → marked or absent
R2 ∃boy ∀girl (one boy)needs reconstruction, or wide scope for the indefinitesurface scope → available
  • SOURCED Willie gives (58) one translation, ‘All the girls were kissed by a boy’, and (59) one, ‘Each of the girls kissed a boy’. She says the voice "provides different interpretations of the quantifier" (p. 79). Only (60), the da- form, is marked for both readings.
  • HYPOTHESIS The contrast in the translations is all vs each. That is a contrast between collective and distributive readings. It is not stated as a contrast in scope. If it is real, the distributive reading goes with the girls when they are the agent (59). That is the same pattern as DIST-1: distributivity tracks ρA. If R1 turned out to be absent or marked for (58), that would favour V-H over V-P. If it is present, the test is still weak, because indefinites like łaʼ can take wide scope freely in many languages.
  • OPEN T5c is not settled by (58)–(60). What would settle it is one judgment per form, from speakers: in (58), can each girl have been kissed by a different boy? In (59), can one boy have been kissed by all the girls? And then the same questions for a pair where the universal is on the internal nominal (schematic only: [’ashkii ła’] [t’áá ’ałtso ’at’ééké] yi-/bi-V). No source prints that string, and none is invented here. (T10a.)
  • SOURCED + HYPOTHESIS A side result for INV-4: (58) has an indefinite (łaʼ) internal agent under bi-, and it is fine. (74c) * iiʼniʼ łííʼ łaʼ biisxį́ has łaʼ on the external nominal, and it is out. Together they confirm that INV-4 is about position (EXT only). A universally quantified EXT, ‘all the girls’, satisfies it. (That a universal counts as "definite" is our reading.)

6. Derivations N1–N4, step by step

Conventions as on the derivations page. Σ schema step, A word-internal linearization, B clause level. V-H = in-situ licensing; V-P = VPISH raising. Pl = the head spelled out da-. Its height is not known (§5.1), so we give it the inside-vP placement that keeps V-P consistent, and we flag where the high placement would fail. Entailments for ‘see’ and ‘kiss’ are as in the verb dictionary (ours, HYPOTHESIS). No verb is segmented beyond Willie's gloss.

N1. (63b) ʼashiiké ʼatʼééké dayiiłtsą́ ‘The boys saw the girls.’ and (63d) *ʼashkii ʼatʼééké dayiiłtsą́ (direct)

  1. Σ Σ(√SEE) = {ρA, ρP}. EρA ⊇ {sent}, and EρP has no clear Proto-Agent entailment. C5: ρA = perceiver. posρA = Spec,v, and posρP = Comp,V.
  2. Merge V with ρP. A ρP⌢V. Merge v. A v⌢VP. Merge ρA in Spec,v.
  3. Σ Number. In (63b), #(ρA) = pl (boys) and #(ρP) = pl (girls). In (63d), #(ρA) = sg (boy).
  4. Merge Pl (inside-vP placement, §5.1). Agree Pl ⟶ ρA. ρA is the closest goal, and DIST-1 is checked. (63b): pl ✓. (63d): sg ✗. The derivation of (63d) fails here, inside the word. With no nominals at all the same star appears: (9) *‘He saw them’.
  5. Merge M/S. Σ π = 3 for both slots, so INV-0 holds. Rank is equal (human/human), so INV-3 leaves EXT free and discourse chooses EXT = ρA. INV-2 gives yi-.
  6. Agree M/S ⟶ ρA. Move: under V-P, ρA goes from Spec,v to Spec,M/SP and nothing crosses. Under V-H, nothing moves.
  7. Spell-out A → dayiiłtsą́, glossed PL.3.3.saw. da- is leftmost.
  8. B Linking (INV-6). ʼatʼééké, next to the verb, links to the internal slot, ρP. ʼashiiké links to EXT = ρA. The number of each nominal matches its slot. (63c) differs only in #(ρP) = sg, and DIST-2 lets it through.

N2. (64b) shizhéʼé ʼashiiké dabiiłtsą́ ‘My father was seen by the boys.’ (inverse)

  1. Σ Same Σ and entailments as N1. ρA = the boys (#pl), and ρP = my father (#sg).
  2. Merge V, v, ρA as in N1.
  3. Merge Pl. Agree Pl ⟶ ρA, which is pl ✓ (DIST-1). HYPOTHESIS This Agree has to happen before step 6. If Pl were merged above M/S, then after step 6 its closest goal would be ρP (sg), which would wrongly require a plural father, or wrongly allow (65b).
  4. Merge M/S. Σ INV-0 holds. Rank is equal, so discourse chooses EXT = ρP, and INV-2 gives bi-.
  5. Σ DIST-3: EXT is singular, so it must be "maximally referential". shizhéʼé (1s.father) passes (our reading, §3).
  6. Move. Under V-P, ρP goes from Comp,V to Spec,M/SP, crossing ⟨ρA⟩. Under V-H, nothing moves.
  7. Spell-out A → dabiiłtsą́, glossed PL.INV3.3.saw (Willie) or PL-3ACC-3NOM-saw (W&J).
  8. B INV-6. ʼashiiké, next to the verb, links to the internal slot, which under bi- is ρA. shizhéʼé links to EXT = ρP. ✓

N3. (65b) *, (65c) *, and (66) ✓: the two different ways to fail with dabi-

  1. (65b) *ʼashiiké ʼatʼééd dabiiłtsą́. Under bi-, ʼatʼééd is next to the verb, so it is the internal agent, and it is sg. N2 step 3 fails: DIST-1, decided inside the word. Compare (64a) dabiiłtsą́ ‘by them/*him’, which has the same star with no nominals.
  2. (65c) *ʼatʼééd ʼashiiké dabiiłtsą́. The internal agent ʼashiiké is pl, so DIST-1 ✓. EXT is ʼatʼééd, a bare singular common noun. DIST-3 ✗.
  3. (66) ʼéidí ʼatʼééd ʼashiiké dabiiłtsą́. The same EXT, now with the demonstrative ʼéidí. DIST-3 ✓.
  4. HYPOTHESIS The two stars come from different places. The DIST-1 star is made by Σ and Agree inside the word: it survives with no nominals (64a), so it is rule set A territory. The DIST-3 star depends on the form of the linked nominal (bare vs demonstrative), so it is checked after linking, in rule set B. The bare word (64a) allows a singular ‘he’ as EXT, and a pronoun is definite. This repeats the split that the inverse page found for voice (decided inside the word) versus fronting (rule set B).
  5. OPEN Without da-, a bare singular EXT is fine: (45) ʼashkii ʼatʼééd biztsʼǫs. Why da- raises the bar from "definite" to "maximally referential" is not explained by Willie (T10c).

N4. (58) tʼáá ʼałtso ʼatʼééké ʼashkii łaʼ biztsʼǫs vs (59) … yiztsʼǫs: where the scope readings would come from

  1. Σ Σ(√KISS) = {ρA, ρP}. EρA ⊇ {vol, sent, mov}, EρP ⊇ {stat} (ours, as in the dictionary). The referents are girls (∀) and a boy (indefinite). Rank is equal, so both voices are allowed (INV-3), and Willie prints both.
  2. Merge V, v, ρA, M/S as in N1. There is no Pl. DIST-4: if the girls are two or fewer, no da- is needed. (Willie's "With greater plurals, the distributive da- must be used" follows (60). Whether (58)–(59) are read as paucal is OPEN, T10e.)
  3. Σ (59): EXT = ρA (girls), so the verb is yi-. (58): EXT = ρP (girls), so the verb is bi-. INV-4 holds in both. The EXT is universally quantified, and so definite in our reading. The indefinite is internal, which INV-4 allows.
  4. Move. Under V-P, (59) raises ρA. In (58), ρP raises past ⟨ρA⟩, so the girls' slot c-commands the boy's slot. Under V-H nothing moves, and the boy's slot (Spec,v) c-commands the girls' slot in both.
  5. B INV-6. ʼashkii łaʼ, next to the verb, links to the internal slot: ρP in (59) and ρA in (58). tʼáá ʼałtso ʼatʼééké links to EXT. HYPOTHESIS Under the Pronominal Argument view the nominals are adjuncts, so a quantifier's scope is computed through the slot that its nominal links to. That is why the height of the slot (step 4) is what the readings test.
  6. Result. V-P predicts that R1 (boys vary) is a surface reading of (58). V-H predicts that R1 needs inverse scope. Willie's translations do not tell the two apart (§5.2). OPEN T5c / T10a.

What N1–N4 show HYPOTHESIS

  1. da- counts the agent in both voices. It reads a role slot (ρA), not the subject position. The definiteness condition reads the subject position. One word carries both kinds of requirement.
  2. The agent-number star is decided inside the word. It holds with no nominals ((9), (64a)). The "maximally referential" condition on the inverse subject needs the nominal, so it belongs to rule set B.
  3. VPISH raising in the inverse survives only if Pl Agrees with the agent before the patient raises, or if Pl is role-indexed. A high Pl that probes for the closest goal would get (64a) and (65b) wrong. V-H gets the pattern with no extra assumption.
  4. The quantifier test (T5c) is still open, but it is now one concrete question per sentence for a speaker.

7. Verdict

  • Main finding. SOURCED Willie 1991:32, 82–85: the distributive da- requires a plural agent in the direct and in the inverse, and does not care about the patient in the direct. HYPOTHESIS So the prefix follows the proto-agent role across the voice flip, while definiteness follows the subject position. That is Grant's "theta roles, not lexical items, are the arguments" stated over a real prefix.
  • For VPISH. HYPOTHESIS da- narrows the options. V-P is consistent only if Agree(Pl, ρA) comes before the inverse patient raises, or if Pl is role-indexed. V-H needs nothing extra. Neither variant is refuted.
  • T5c. OPEN Willie's quantifier pair (58)–(59) is now page-checked and on the site. It does not settle T5c, because the needed readings are not stated. The question is reduced to two yes/no judgments (§5.2).
  • T5a / T7. OPEN Linked: if the inverse agent changes case (Willie), da- is a generalization that must be stated over the role (a T7 case). If the agent stays 3NOM (W&J), it is not.

8. Open questions this page adds

#QuestionWhat would settle it
T10a(= T5c, sharpened) In (58) tʼáá ʼałtso ʼatʼééké ʼashkii łaʼ biztsʼǫs, can different girls have been kissed by different boys? In (59) … yiztsʼǫs, can one boy have been kissed by all the girls? And with the universal on the internal nominal (schematic only)?Speaker judgments. Perkins, E. 1978, "The Role of Word Order and Scope in the Interpretation of Navajo Sentences" (PhD dissertation, University of Arizona), is cited in Willie's bibliography (p. 234) and is the obvious next source; not opened. Added 2026-10-09 (afternoon): Bruening 2005 on the Passamaquoddy inverse turns this into a three-way prediction for (58) (ambiguous under V-P with reconstruction; see fourth person compared §6, T11c). OPEN
T10bWhere is Pl (da-, position III) in the tree: above M/S (rule set A's reading of the left edge), or inside vP? Does Mode/Subject agreement in the inverse go to ρA (W&J's 3NOM) or to EXT (as inverse I1 step 6 assumes)?Young & Morgan 1987 on the distributive with intransitives and with the fourth person; Hale 2003's analysis of position III, re-read; Rice 2000 ch. on the plural. OPEN
T10cWhy does dabi- require a plural or "maximally referential" EXT when plain bi- accepts a bare definite singular ((45) vs (65c))? And is (64b) shizhéʼé acceptable because kin terms with a 1st-person possessor are maximally referential?More dabi- examples with possessed and proper-name subjects; speakers. OPEN
T10dIn (68) mąʼii tłʼízí yitah daʼashxash, a singular coyote co-occurs with da-. What does da- count there (the goats via yitah, or the biting events), and what is the unglossed second prefix?Young & Morgan 1987 entry for the verb; Willie Ch. 2 on da-; a speaker. OPEN
T10eAre (58)–(59), which have no da-, read as paucal (two girls)? Willie's "With greater plurals, the distributive must be used" suggests that they are, but the translation ‘All the girls’ does not say.Speakers. OPEN
T10fWith a 4th-person agent (ji-), does da- still count the agent? (W&J (77) dahoołtsą́ has da- with a 4th-person object and a 3rd-person subject, ‘all … children’.) This would test DIST-1 outside the yi-/bi- domain.Willie Chs. 4–5 (T8b); Young & Morgan paradigms. OPEN

Builds on the inverse page (T5) and the fourth-person page (T8). Dictionary entries updated with these examples: see, kiss, bite.

Where might subject d sit? Testing the older template-based alternatives (T92a, T92c)

Working note. Kept as written (URL, data and status labels unchanged). Filed under The word as a spell-out domain; the consolidated argument is the essay word-edge. Rule IDs: rules index.

Restated on the unified structure (2026-11-15). Prefix order is restated as Grant’s hypothesis that headedness is fixed locally by Merge site, so the sublexical domain is linearized opposite to the phrasal domain and the prefixes are successive word-internal Merge and Agree sites (GRANT’S HYPOTHESIS, tested on the sourced template in §3.2; fit/misfit OPEN). A possible sublexical V-to-T is a separate, untested GRANT’S HYPOTHESIS (§3.3). Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). Derivations below in which the verb or subject does not raise are labelled contrast; the working default raises V→v→T and the subject to Spec,TP. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §3. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

2026-11-11 audit pass 1. The older title and SD rules claimed a surviving d displacement under one template-filling procedure. They do not eliminate V→v→T, PF merger or a different realization rule. HYPOTHESIS: subject DP first-merges in left Spec,vP. Nominal d versus verbal v, extraction of a head from a subject DP and its landing site are OPEN; the template is a rival, not our premise. Older mixed-label brackets below are historical sketches for pass 2.

2026-11-11 audit pass 3. dP (the light-d phrase) on this page is a HYPOTHESIS label, not an established category. See the pass 3 report.

Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference.

2026-11-13 audit pass 7. An earlier dated note on this page treats V→v→T and subject raising as Grant’s conjecture or a comparator to test against PF/template rivals. Since 2026-11-13 they are the working default in tensed clauses: VPISH frames where the arguments start so that syntax and semantics line up; it is not a word-order account. PF/template accounts are kept only as labelled contrast, and the open question is what else has to move. The earlier note is kept as written. See the 2026-11-13 note and the pass 7 report.

The verb as a derivation left open whether the subject's light determiner d is (a) the head of a dP in Spec,Voice or (b) a Voice-like head in the verbal spine (T92a), and whether the word is built by head movement or after syntax (T92c). This page re-opened Hale 2001 (Linguistic Inquiry 32: 678–693) and runs every option through a derivation, with the rules stated first. Two printed forms constrain that particular template comparison, not every possible derivation: Hale's (22b) shiʼdooʼniid 'I was told (lit. someone told me)' and Young & Morgan's biʼdi- (212/3). In both, the indefinite subject ʼ- comes after the object prefix and before every spine head. Neither (a) nor (b) produces that order. What does is a third option, already used on subject on the head (SH-4): the subject dP merges in Spec,vP, and only its head d moves to the top of the spine. That is analysis D's "only the head d joins the word". This page restates it in /derivation/'s terms. HYPOTHESIS The same derivation answers part of T92c: the spine itself needs no head movement (Hale's claim, SOURCED as his proposal), but the deictic subject d does need one displacement, out of a specifier. Ordinary head movement cannot do that.

On this page: 1 Question · 2 Sources · 3 Structure · 4 Rules first · 5 Data · 6 Derivations · 7 T92c: does anything move? · 8 What this does to D and to /derivation/ · 9 Limits · 10 Open questions

Notation follows pos-labels and derivation. A = word level (left-headed prefix string), B = clause level (right-headed). d is Grant's light determiner (GRANT'S CONJECTURE, used here as HYPOTHESIS). "Nuclear" means a head in the extended projection of V (Hale's term, p. 681: V, Voice, Mode/Subj, Qualifier). "Nonnuclear" means everything else that ends up in the word: complements, adjuncts, preverbs (Hale p. 682). M/S = Hale's Mode/Subj portmanteau. 3i is Young & Morgan's indefinite ʼa-/ʼ-. Position labels (IV, V) are cited only as data.

Historical note: the old "roles, not words" idea is not pursued here (Grant has dropped it).

1. Question

/derivation/ §1 names two ways the subject d could be built, with a split as its working choice: the object d takes a DP, and the subject d is a Voice-like spine head (option b). §4 then says that the object-before-subject order "cannot come straight from the order of Merge" and states a template instead. This page asks whether an explicit spell-out procedure, the one Hale 2001 gives, can (i) decide between (a) and (b) and (ii) derive the order rather than list it. Focus 1 (the word as the changeover from ruleset B to ruleset A) and Focus 2 (how a prefix relates to the argument it stands for) are both at stake, because the 3i is a subject prefix that is pronounced in the word while its phrase, under D, starts in the clause structure.

2. Sources

Re-opened this run. Hale, Ken 2001. "Navajo verb stem position and the bipartite structure of the Navajo conjunct sector." Linguistic Inquiry 32(4): 678–693. Box text layer [local working copy, not in the repository], read in full (pp. 678–693). Used: Rice's head-movement derivation (4)–(5), pp. 679–680; Hale's tree (6) and its heads, p. 681; "There is no head movement in the traditional sense", p. 681; the skeleton spell-out (8)–(12), pp. 682–683; the object as complement after Jelinek, p. 682; nonnuclear proclisis (18)–(19), p. 688; §4 deictic subjects (22)–(24), pp. 689–692; n. 7 on Lowering, p. 692.

Not re-opened, used as transcribed earlier on this site. The Navajo forms of (22a, b), (23), (24a) as read from page images at 300 dpi on subject on the head §3.1. The OCR text layer drops ł and some tone marks, so its segment lines were compared with those transcriptions: the segments and category labels agree. Young & Morgan 1987 212/3 biʼdi-, as quoted on pos-labels LB-1.

Cited by Hale, not opened: Rice 2000; Speas 1984, 1990; McDonough 2000; Hale 2000; Embick & Noyer 2001. Nothing is attributed to them beyond what Hale reports.

3. Structure

SOURCED (Hale's proposal, p. 681, tree (6)). The nuclear heads, lowest to highest, are V, Voice, M/S and Qualifier. Each selects the next one down and takes it on its right ("the functional projections of Navajo are taken to be left-headed"). V takes its object complement on its left. Preverb, iterative and plural are adjuncts to VP. The object prefix "is the true object argument" and "appears within the verbal projection as a complement" (p. 682, following Jelinek 1984). Hale's tree has no Spec,Voice subject. The subject's person is part of M/S (pp. 690–691).

HYPOTHESIS (ours). Following VPISH and /derivation/, the external argument is first-merged in Spec,vP, with Voice = v (the "classifier", Hale p. 680: Voice "is a light verb that selects the verbal category VP"). Under D the object is [dP d (DP)], the sister of V. The subject is a dP whose base position is to be decided (§6).

[QualP Qual [M/SP M/S [vP ( dP_subj ) [v' Voice [VP dP_obj V ]]]]]      (spine head-initial; V head-final)
 nuclear:    Qual  M/S  Voice  V
 nonnuclear: dP_obj (complement of V), adjuncts to VP, and dP_subj if it stays a phrase
status: hypothesis (Hale's tree (6) plus a VPISH subject; Hale draws no subject position)

4. Rules first

SD-1  NUCLEAR SPELL-OUT. The nuclear heads are spelled into the stem's            SOURCED (Hale pp. 681–683)
      skeleton bottom-up, "successive-cyclically", so a higher nuclear head
      ends up further left. No head movement: "the process involved in defining
      the conjunct sector is not head movement, implicating traces" (p. 683).
SD-2  NONNUCLEAR PROCLISIS. After the nuclear heads are spelled, "the               SOURCED (Hale p. 688)
      nonnuclear elements are successively procliticized to the left of the
      conjunct sector", working upward, so every nonnuclear element precedes
      every nuclear one, and a higher nonnuclear element precedes a lower one.
SD-3  PL (theorem). From SD-1 + SD-2: (a) nonnuclear before nuclear;             HYPOTHESIS (stated as PL on
      (b) within one plane, the c-commanding element comes first.                 /spell-out-rules/#ab)
SD-4  SUBJECT PERSON ON M/S. A 3rd-person subject has its own exponent inside      SOURCED (Hale pp. 690–691,
      M/S: the rounded gamma-perfective allomorph w "appears only in the third     (24a, b))
      person". So subject phi is spelled on a nuclear head above Voice.
SD-5  DEICTIC SUBJECTS ARE NOT M/S. The 3a (4th person) j-, the 3i '- and the      SOURCED (Hale pp. 689–691)
      areal ho- are subject markers "construed with the third person in
      Mode/Subj". Their surface place is "to the right of Object and to the
      left of Qualifier". Hale calls them clause "adjuncts", "infixed" and spelled out as
      "prefixes to the conjunct sector of the verb (or, perhaps more narrowly,
      as prefixes to the Qualifier element ...)". They "belong grammatically
      to the same category of elements as full DP (or NP) arguments" (p. 691).
SD-6  VPISH. An agent is first-merged in Spec,vP, above the object.                 HYPOTHESIS (site frame)
SD-7  NO HEAD FROM A SPECIFIER BY HEAD MOVEMENT. Head movement takes a head to      HYPOTHESIS (standard
      the head that selects it. A head inside a specifier cannot reach the spine    assumption, not sourced
      this way.                                                                     here)

SD-1, SD-2, SD-4 and SD-5 are Hale's, in his words. The numbering and the use of PL as a filter on /derivation/'s options are ours.

5. Data

Forms and translations are Hale's (or Young & Morgan's). The Leipzig lines relabel Hale's category lines as on subject on the head. INDF.SBJ = 3i, 4.SBJ = 3a.

S1. Hale 2001 (22b), p. 690. Segments sh ʼ d w d niid; categories Object 3i Qualifier Mode/Subj Voice V.
shiʼdooʼniid
1SG.OBJ-INDF.SBJ-QUAL-M/S-VOICE-tell
'I was told (lit. someone told me).'

S2. Hale 2001 (22a), p. 689. Segments sh j Ø ł ghozh; categories Object 3a Mode/Subj Voice V.
shijiłhozh
1SG.OBJ-4.SBJ-M/S-VOICE-tickle
'He/She/They/One is tickling me.'

S3. Hale 2001 (24a), p. 691 (control: ordinary 3rd-person subject). Segments sh d w ł ghosh; categories Object Qualifier 3rd:Perf Voice V.
shidoołhosh
1SG.OBJ-QUAL-3.PFV-VOICE-tickle
'He/She will tickle me.'

S4. Young & Morgan 1987 212/3, prefix entry (as quoted on pos-labels): "Pos. IV, obj. pron., him + '(a)di-, Pos. V, 3i subj. pron. = biʼdi-, forms Passive B". Our segmentation bi-ʼ-di-:
biʼdi-
3.OBJ-INDF.SBJ-QUAL(?)
(prefix complex, no verb). Whether di here is the qualifier or part of the 3i entry is not shown in the quote. OPEN

SOURCED The facts the derivation must produce. (F1) With an ordinary 3rd subject (S3), the subject exponent w is to the right of the qualifier d. (F2) With a 3i subject (S1, S4), the 3i is to the right of the object prefix and to the left of the qualifier. (F3) With a 3a subject (S2), j is to the right of the object prefix. Here there is no overt qualifier, so S2 shows only the object side.

6. Derivations

Every option uses the same lower steps. Only how the subject is built differs. Each line names the operation and the ruleset (B = syntax/clause, A = word spell-out). The target is S1 shiʼdooʼniid.

COMMON STEPS (S1)
1  Merge   [VP dP_obj V]; V = √niid 'tell' (stem shape from Hale's segment line);
           dP_obj = [d 1SG] (no overt DP)                                         B
2  Merge   Voice = d (Hale: the agentive passive "requires the d-classifier", n. 6) B
3  Merge   M/S (gamma perfective; exponent w in this context, p. 691)             B
4  Merge   Qual = d (Hale's "Qualifier")                                           B
5  Agree   M/S probes and finds the subject: phi = 3 (SD-4)                        B
6  Spell   nuclear: Voice d → coda; M/S w → onset; Qual d prefixed (SD-1)          A
           nuclear string so far:  d – w – d – niid
7  Spell   nonnuclear: dP_obj sh procliticized (SD-2)                              A

OPTION (a): subject = dP in Spec,vP, stays a phrase
3a Merge   dP_subj = [d 3i] in Spec,vP (SD-6)                                      B
7a Spell   nonnuclear plane = {dP_subj, dP_obj}; dP_subj c-commands dP_obj,
           so by SD-3(b) 3i comes first:   ' – sh – d – w – d – niid               A
           predicted *'ishidoo'niid-type order (3i before object).  ✗ F2
           (S4 bi'di- shows the same: the object comes first.)

OPTION (a-low): the subject dP is lower than the object
           This gives sh – ' ✓, but only if the agent is first-merged below the
           object. That contradicts SD-6 for an agent: S1 means 'someone told me'.
           ✗ SD-6 (for an agentive 3i; see §9 for theme 3i)

OPTION (b-low): subject d = a Voice-like spine head next to Voice (/derivation/'s choice)
3b Merge   d_subj as a nuclear head between Voice and M/S                          B
6b Spell   nuclear, bottom-up (SD-1):  d(Qual) – w – ' – d(Voice) – niid           A
7b Spell   sh procliticized:  sh – d – w – ' – d – niid
           3i after M/S.  ✗ F2 (the 3i must precede the qualifier)
           Same failure for any spine slot below Qual (between Qual and M/S:
           d – ' – w ...  ✗ F2).

OPTION (b-high): subject d = a spine head above Qual
3c Merge   d_subj as the highest nuclear head                                      B
6c Spell   ' – d – w – d – niid; 7c sh procliticized → sh – ' – d – w – d – niid   A
           ✓ F2. Cost: the head that is supposed to introduce the agent is merged
           above aspect/mode (M/S) and above Qual, far outside vP. That clashes
           with SD-6, which is what made it "Voice-like".  (HYPOTHESIS: cost)

OPTION (c): dP in Spec,vP; its head d moves to the top of the spine (SH-4)
3d Merge   dP_subj = [d 3i] in Spec,vP (SD-6: agent above object)                  B
5d Agree   M/S finds dP_subj: phi = 3 → spelled as w on M/S (SD-4)                 B
5e Move    d_subj → adjoins to the highest nuclear head, Qual                 B → A
           (not head movement by SD-7: it leaves a specifier. Hale: "infixed")
6d Spell   nuclear: [' + d(Qual)] – w – d(Voice) – niid                            A
7d Spell   nonnuclear: sh procliticized:  sh – ' – d – w – d – niid  ✓ F2           A
           surface (Hale's epenthesis, glide rounding, w drop, pp. 687–688;
           not derived here): shi'doo'niid

HYPOTHESIS Result. Given Hale's two spell-out steps (SD-1, SD-2) and VPISH (SD-6), only (b-high) and (c) produce F2. (b-high) does it by giving up the reason for calling the subject d "Voice-like". (c) does it by splitting the subject in two. Its phi is spelled on M/S through Agree (step 5d). Its d is pronounced at the top of the nuclear spine (step 5e). The subject phrase itself stays in Spec,vP (Hale p. 691 says only "adjuncts to the clause"; the Spec,vP site is ours), which is where Grant's D puts the noun of a full subject DP (Hale's (23) Ashkii shiłhozh 'The boy is tickling me', boy 1SG.OBJ-M/S-VOICE-tickle, p. 690).

Update 2026-11-05 evening PT: qualifier-d-order shows that in six hodii- future cells the 4th-person zh follows a listed thematic di (ʼáhodizhdoo- and kin). An exceptionless reading of step 5e as “above every Qualifier-di” is too strong; QD-5 places the exponent at the left edge of future Mode, which may sit below thematic di. HYPOTHESIS

HYPOTHESIS Check with S3 (F1). With an ordinary 3rd subject there is no deictic d, so step 5e has nothing to move. The subject is pronounced only through Agree on M/S: d(Qual) – w(M/S, 3) – ł – hosh, and the object comes first: shidoołhosh ✓. Option (b-high) has to say that ordinary subjects lack the high head and deictic ones have it. Option (c) says the same thing in other words (only deictic subjects have an overt d). So F1 does not choose between them.

HYPOTHESIS Check with S2 (F3). sh – j – Ø(M/S) – ł – hozh. Option (a) predicts j before sh ✗. (b-low) between Voice and M/S gives sh – Ø – j – ł, which can't be told apart from F3 because M/S is silent. (c) ✓. So S2 rules out (a) only. S1 is the form that decides the rest.

7. T92c: does anything move?

SOURCED Two analyses in print. (i) Rice's, as rendered by Hale (pp. 679–680, (4)–(5)): V "raises and right-adjoins" to Voice, and Voice+V then right-adjoins to M/S. That is head movement, and it is rightward adjunction. This is how it avoids /derivation/'s Problem 2 (leftward adjunction would make the word suffixing). (ii) Hale's own (p. 681): left-headed functional heads with no head movement. "The processes involved are properly speaking phonological."

SOURCED Where Hale himself says his account may need movement (n. 7, p. 692). In the perfective, the M/S paradigms "display two sets of allomorphs depending upon the identity of the Voice element": one with the ł and zero classifiers, the other with d and l. If the right M/S form cannot be chosen in base position, "it might be necessary to apply a rule of Lowering, prior to Vocabulary Insertion", and then the "nonmovement" claim is "considerably weakened". Rice (as Hale reports her, 2000:169, 178) says the perfective marker is simply absent next to d and l. Hale replies that the environment for that still has to be identified.

HYPOTHESIS Our answer to T92c, in part.

  • The spine (V, Voice, M/S, Qual) needs no head movement under Hale's procedure. /derivation/'s Problem 2 is not a problem for it, because the order comes from SD-1, not from adjunction. In Hale's tree, M/S selects VoiceP, so Voice is M/S's complement head. If Vocabulary Insertion runs bottom-up (as SD-1's "successive-cyclically" says), Voice is already spelled when M/S is reached. On that reading the n. 7 allomorphy can be decided in situ, and Lowering is not needed. Whether that is right depends on the DM locality conditions for contextual allomorphy, which were not opened here. OPEN (T93a)
  • /derivation/'s Problem 1 (the object is farther from the stem than the subject, although it merges lower) is not a problem either. The object is nonnuclear (a complement) and the subject's person is nuclear (on M/S), so SD-3(a) puts the object first. The Mirror Principle is not violated within either plane: within the nuclear plane, higher is further left (SD-1). The order is derived, not listed.
  • The deictic subject d is the one thing that has to be displaced (§6 step 5e). By SD-7 this cannot be ordinary head movement, because the d starts inside a specifier. It is either a post-syntactic operation (Hale's "infixation", after Speas 1990 and Hale 2000 as he reports them, p. 691) or movement of a head out of a specifier. So "nothing moves" is wrong for words with a deictic subject. Only this one step is in question.

8. What this does to D and to /derivation/

  1. HYPOTHESIS T92a, in part. /derivation/'s working choice, that the subject d is a Voice-like spine head (b-low), fails on S1. The plain specifier option (a) fails on S1, S2 and S4. Option (c) keeps the subject's dP where VPISH puts it and puts only its head d into the word, at the top. That is analysis D applied to subjects: "only head d joins the verb word".
  2. HYPOTHESIS Object and subject d join the word differently. The object d is spelled as a nonnuclear proclitic, in place (SD-2). The deictic subject d is moved onto the nuclear spine (step 5e). This gives a derivational version of the Syntax bot's split hypothesis (object d takes a DP, subject d is "in the spine"). The subject d is not merged in the spine. It is pronounced there.
  3. HYPOTHESIS The changeover (Focus 1). Step 5e is the one point where a B-level object, a phrase in Spec,vP, contributes a piece to A-level spell-out away from its base position. Every other piece of the word is spelled where it was merged. If the word is the changeover from ruleset B to ruleset A, the deictic d is where the two meet.
  4. HYPOTHESIS Grant's light-verb idea (subjects generated at different sites, Japanese-style). It survives only as (b-high): a high subject-introducing head, above aspect. No opened source supports a Navajo agent introduced above M/S, so (b-high) is kept but not preferred. OPEN (T93d)
  5. HYPOTHESIS Stem listing. listed-d found the 3i listed in the stem's prefix string for some transitive verbs. Under (c), a listed 3i is a stem that selects a dP_subj headed by 3i d. Listing restricts which d is merged. It does not decide where that d is pronounced, so it is compatible with (c). (/derivation/ took the listing as evidence for (b). On this page it is neutral.)
  6. Multiple wh. Not used. Fronting orders phrases (DPs), not prefixes (Syntax bot), and S1–S4 contain no wh. Nothing in the wh-order notes bears on step 5e.

9. Limits

  • All four forms have an agentive or human deictic subject (S1 'someone told me', S2 'one is tickling me'). The passenger and vehicle 3i of driver-3i and pos-labels is a theme, and option (a-low) is not excluded for it. So this page says nothing about T87a (unaccusativity of passenger 3i). By step 5e, the 3i ends up at the top of the nuclear spine whatever its first-Merge site, which is LB-10's point again. OPEN
  • S4 is a prefix-entry quote with no verb, and its di may not be the qualifier. It supports F2 only on the object side (object before 3i).
  • The Navajo forms were not re-read from page images this run. The text layer agrees with the earlier 300 dpi transcriptions in segments and categories, but it drops ł and some tones.
  • Surface phonology (epenthesis, w rounding and drop, glottal absorption) is Hale's (pp. 682–688) and is not re-derived here.
  • SD-7 is a standard assumption about head movement. It is not taken from an opened source.

10. Open questions

IDQuestionWhat would settle it
T93aCan the perfective M/S allomorphy that depends on Voice (Hale n. 7) be chosen in situ under bottom-up insertion, or does it need Lowering?An opened DM source on locality for contextual allomorphy (e.g. Embick & Noyer 2001, cited by Hale). A Young & Morgan perfective paradigm pair (ł/Ø vs d/l classifier) for a schematic test. Answered in part on voice-allomorphy (2026-11-04 evening).
T93bIs step 5e (deictic d to the top of the spine) post-syntactic (Hale's "infixation") or syntactic? Does it feed anything syntactic, such as binding by the 4th person or the yi-/bi- choice?Hale 2000 "Remarks on the syntax of the Navajo verb" (not opened). A 4th-person binding fact where the d's surface position matters.
T93cWhen a deictic d is the object (3i object ʼa-, 4th-person object ha-/ho-), does it stay nonnuclear (SD-2), as (c) predicts, so that it precedes a deictic subject d?A sourced form with both a deictic object and a deictic subject, e.g. from Willie 1991 or Young & Morgan. Update 2026-11-05: answered in part on deictic-order (object first; 4th-person object case still OPEN).
T93dIs there any source-backed reason to merge a Navajo agent above M/S (option b-high, Grant's light-verb idea)?Sources on Athabaskan "classifier" vs subject introduction (Rice 2000 is not opened); scope of the 3i subject relative to aspect.

Answered in part here: T92a (options (a) and (b-low) excluded for agentive deictic subjects; (c) or (b-high) remain) and T92c (spine: no head movement needed; deictic d: one displacement that is not head movement). T92b and T92d are untouched.

Voice and perfective allomorphy: an in-place hypothesis, not a proof against movement (T93a)

Working note. Kept as written (URL, data and status labels unchanged). Filed under The word as a spell-out domain; the consolidated argument is the essay word-edge. Rule IDs: rules index.

Restated on the unified structure (2026-11-15). Prefix order is restated as Grant’s hypothesis that headedness is fixed locally by Merge site, so the sublexical domain is linearized opposite to the phrasal domain and the prefixes are successive word-internal Merge and Agree sites (GRANT’S HYPOTHESIS, tested on the sourced template in §3.2; fit/misfit OPEN). A possible sublexical V-to-T is a separate, untested GRANT’S HYPOTHESIS (§3.3). Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). Derivations below in which the verb or subject does not raise are labelled contrast; the working default raises V→v→T and the subject to Spec,TP. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §3. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

2026-11-11 audit pass 1. Local allomorphy is data, not proof that the stem or T remained stationary. The “without Lowering” argument assumes a particular locality and vocabulary-insertion theory. Our primary V→v→T comparator and PF/template rivals must predict the same attested allomorphs before either wins. The M/S-to-T and classifier-to-v identifications, and the full DM rule derivations, remain HYPOTHESIS / OPEN.

2026-11-13 audit pass 7. An earlier dated note on this page treats V→v→T and subject raising as Grant’s conjecture or a comparator to test against PF/template rivals. Since 2026-11-13 they are the working default in tensed clauses: VPISH frames where the arguments start so that syntax and semantics line up; it is not a word-order account. PF/template accounts are kept only as labelled contrast, and the open question is what else has to move. The earlier note is kept as written. See the 2026-11-13 note and the pass 7 report.

Hale 2001 (n. 7) worried that one fact might force his "no movement" Navajo verb to use a Lowering rule. The fact is that the perfective Mode/Subject prefix has one set of forms next to the zero and ł classifiers and another next to d and l. This page checks that worry against three DM locality sources and against Young & Morgan's paradigm tables. The result: Mode/Subject and Voice are a head and the head of its complement, and nothing is ever spelled between them. Under cyclic, inside-out insertion the right form can be chosen where Mode/Subject is merged. Lowering would relate the same two heads and adds nothing. One condition stays OPEN: whether contextual allomorphy needs a single complex head.

1. The question

T93a (from subject-d-site): can the perfective Mode/Subject (M/S) allomorphy that depends on Voice (Hale 2001, n. 7) be chosen in situ under bottom-up insertion, or does it need Lowering? Hale's spine has no head movement (SD-1). If the choice needs M/S and Voice in one complex head, M/S has to Lower onto Voice before Vocabulary Insertion (VI), and Hale says his "nonmovement" claim would then be "considerably weakened".

2. Sources opened this run

  • SOURCED Hale, Ken. 2001. Navajo verb stem position and the bipartite structure of the Navajo conjunct sector. Linguistic Inquiry 32: 678–693. Footnote 7, p. 692 (re-read from the text layer used on subject-d-site).
  • SOURCED Embick, David & Rolf Noyer. 2001. Movement operations after syntax. Linguistic Inquiry 32: 555–595. Lowering (6), pp. 561–562; Late Linearization Hypothesis (8); Figure 1, p. 566; Local Dislocation Hypothesis (16), p. 566; Bulgarian definite article allomorphy, p. 573, and n. 22.
  • SOURCED Bobaljik, Jonathan David. 2000. The ins and outs of contextual allomorphy. University of Maryland Working Papers in Linguistics 10: 35–71. Assumptions (3) (separation, cyclicity, rewriting) and the conclusion on inward vs outward sensitivity.
  • SOURCED Merchant, Jason. 2015. How much context is enough? Two cases of span-conditioned stem allomorphy. Linguistic Inquiry 46: 273–303. Node Adjacency Hypothesis, p. 273; his report of Bobaljik 2012 and Embick 2010/2012, pp. 275–276 and n. 4; spans (28)–(29), p. 288; p. 289; n. 18, p. 287; Span Adjacency Hypothesis (38), p. 294.
  • SOURCED Young, Robert W. & William Morgan Sr. 1987. The Navajo Language: A Grammar and Colloquial Dictionary, revised edition. Verb entries yishą́ (p. 779) and yishdlą́ (p. 781), and the prefix paradigm tables. All read through the CU Boulder Y&M interface (verbs.colorado.edu/~mahu0110/ym/, show_paradigm.php), not from print.

Not opened: Embick 2010, Bobaljik 2012, Rice 2000, Svenonius 2012. Where they matter, this page cites them only as Merchant or Hale report them.

3. Structure assumed

[QualP Qual [M/SP M/S [vP ( dP_subj ) [v' Voice [VP dP_obj V ]]]]]      (as on /subject-d-site/#structure)
 nuclear:    Qual  M/S  Voice  V          (spelled into the stem skeleton, SD-1)
 nonnuclear: dP_obj, adjuncts, dP_subj    (proclitic, SD-2; deictic subject d to the top, SD-5)
 M/S selects VoiceP (here written vP), so Voice is the head of M/S's complement.
status: hypothesis (Hale's tree plus VPISH, unchanged from subject-d-site)

4. Rules first

VA-1  TWO PERFECTIVE SETS. In the perfective, the M/S paradigms "display two       SOURCED (Hale 2001 n. 7,
      sets of allomorphs depending upon the identity of the Voice element":         p. 692)
      one with ł and zero, one with d and l. Choosing in base position needs
      "Mode/Subj properly situated relative to Voice". Otherwise "it might be
      necessary to apply a rule of Lowering, prior to Vocabulary Insertion".
VA-2  LOWERING. A head lowers to the head of its complement, before VI.             SOURCED (Embick & Noyer 2001
      It is stated over hierarchy, not over the string, and it skips              (6), pp. 561–562)
      specifiers and adjuncts (English T to v).
VA-3  INSIDE-OUT VI. Insertion "proceeds from the inside-out" (A1). "Contextual    SOURCED (Merchant 2015 p. 276,
      allomorphy requires concatenation (linear adjacency)" (A2).                   quoting Embick 2012; n. 4 Embick
                                                                                   2010 pp. 15, 42)
VA-4  INWARD SENSITIVITY IS TO FORM. Because insertion is cyclic and              SOURCED (Bobaljik 2000, (3)
      morphosyntactic features are used up when an inner node is spelled, an      and conclusion)
      outer morpheme can look inward only at the inner morpheme's
      morphophonological or diacritic properties, not at its syntactic features.
VA-5  SPAN ADJACENCY. "Allomorphy is conditioned only by an adjacent span."         SOURCED (Merchant 2015 (38),
      A span is a sequence of heads in complement relation in one extended         p. 294; (28), p. 288; p. 289;
      projection. Spans "do not require head movement": "head movement may feed    n. 18, p. 287)
      Vocabulary Insertion, but it need not". "Contiguity is a relation that is
      defined over strings of elements, not just inside complex heads."
VA-6  M/S VOCABULARY (schematic).                                                  HYPOTHESIS (ours, written over
        M/S[PFV, 1SG] ↔ ...sh (si-type: ...sís) / __ Voice{d, l}                   Y&M's two columns, §5)
        M/S[PFV, 1SG] ↔ ...(no sh)  elsewhere
      Parallel pairs for the 3rd persons and 2DU. No D-L entry is needed for 2SG
      (yi-/ni- types) or 1DU (F2).
      The context names Voice's exponent class {d, l}, not a feature [passive].
VA-7  NO INTERVENER. Nothing is ever spelled between M/S and Voice. Nonnuclear     HYPOTHESIS (theorem from SD-1,
      material is procliticized to the left of the whole conjunct sector (SD-2),  SD-2, SD-5 and step 5e of
      and a deictic subject d is pronounced at the top, before Qual (SD-5). So     /subject-d-site/)
      in the nuclear string M/S and Voice are always concatenated.
VA-8  IN-SITU CHOICE (theorem). From VA-3, VA-4, VA-6, VA-7: when M/S is reached,  HYPOTHESIS
      Voice is already spelled, adjacent, and visible by form, so VA-6 applies
      with M/S in its base position. Lowering (VA-2) would relate the same pair
      of heads and adds nothing the choice needs.

5. Data

Forms and translations are Young & Morgan's, copied from the CU Boulder interface. The Leipzig segmentations are ours (HYPOTHESIS). We do not mark a 3rd-person object prefix because these forms have none on the surface.

V1. Y&M 1987 p. 779, entry yishą́ 'to eat it' (classifier Ø).
Shimá sání bighandi naayízí tʼóó báhádzigo łikango yíyą́ą́ʼ ʼadą́ą́dą́ą́ʼ
'I ate a really delicious squash yesterday at my grandmother's.'
yíyą́ą́ʼ: yí-Ø-yą́ą́ʼ PFV.1SG.SBJ-VOICE(Ø)-eat.PFV

V2. Y&M 1987 p. 781, entry yishdlą́ 'to drink it, to quaff it, to eat it (liquid, as soup)' (classifier d; Y&M: "*d-dlą́ = dlą́").
ʼAhbínídą́ą́ʼ shimá sání bighandi tóshchiin yishdlą́ą́ʼ
'I ate (drank) some gruel yesterday at my grandmother's.'
yishdlą́ą́ʼ: yi-sh-(d)-dlą́ą́ʼ PFV-1SG.SBJ-VOICE(d)-drink.PFV

V3. Y&M 1987 p. 779, the same entry's perfective row: "(P) yíyą́ą́ʼ, yíníyą́ą́ʼ, yiyíiyą́ą́ʼ, jííyą́ą́ʼ, yiidą́ą́ʼ, wooyą́ą́ʼ (yidą́ą́ʼ/biʼdoodą́ą́ʼ)", with the note "Stem-initial y- = d- in the presence of D-Classifier, and in the 1st person dpl." The forms in parentheses are Y&M's passive forms. No sentences are given for them.

SOURCED The pair. V1 and V2 are the same frame from the same pages: transitive, yi-perfective, 1st singular subject, object noun outside the verb. The perfective prefix with the Ø classifier is yí- with no sh. With the d classifier it is yish-. V2's d is the verb's own listed classifier. 'Drink it' is active and transitive, not passive.

The paradigm tables

Y&M's prefix paradigm tables have two PERFECTIVE columns, headed "Ø - Ł" and "D - L". This run read every table the interface serves (table ids 1–410; the last non-empty one is 385). 140 distinct tables (by page and title) have both perfective columns and parsed cleanly. The others have no D - L column, or a header our script could not align. Counts below are over cells where both perfective columns are non-blank. The third column compares D - L with the IMPERFECTIVE cell, because Rice says (as Hale reports her, 2000: 169, 178) that M/S next to d/l is realized "as if imperfective".

Cell (Y&M row)Tables with both cellsD - L ≠ Ø - ŁD - L = IMPERFECTIVE
1sg13212760
2sg13218 (24 counting accent and apostrophe differences)0
311678 (87)46
3a134105 (106)46
3i2917 (18)9
1du13316 (17)85
2du13312324
3, distributive-plural block11087 (90)19

1sg endings: the D - L form ends in sh in 101 of 132 tables. Most of the rest are si-perfective tables where Ø - Ł -sé stands against D - L -sís (e.g. p. 154 béésé / béésís; p. 688 si-: se / sis). The Ø - Ł 1sg ends in sh in 1 of 132 (p. 332 di- ((2,5)): deesh / díí, perhaps with the columns swapped). 3rd person: Ø - Ł -z against D - L -s in 33 of 116.

Example, Y&M p. 50, ʼahééʼa- (cells copied as shown; IMPERFECTIVE / PERFECTIVE Ø - Ł / PERFECTIVE D - L): 1sg ʼahééʼnísh- / ʼahééʼní- / ʼahééʼnísh-; 2sg ʼahééʼí- / ʼahééʼííní- / ʼahééʼííní-; 3 ʼahééʼí- / ʼahééʼní- / ʼahééʼí-; 2du ʼahééʼnóh- / ʼahééʼnoo- / ʼahééʼnooh-. Y&M p. 63, ʼahi- ((1)): 1sg ʼahish- / ʼahíí- / ʼaheesh- (D - L differs from the imperfective). Y&M p. 48, ʼahééhi-: 3 ʼahééheez- (Ø - Ł) vs ʼahééhees- (D - L).

SOURCED The facts the derivation must produce (counts from the table above). (F1) With the D - L classifiers the perfective M/S has its own forms in the 1st singular (127/132), mostly ending in sh (101/132), and mostly also in the 3rd persons and the 2nd dual. (F2) Two cells mostly do not change: the 2nd singular (same in 114/132) and the 1st dual (same in 117/133). Most 2sg differences are in si-perfective tables (p. 155 bííní / bísíní), or in rows that look misaligned (p. 121 shows 2du-like ʼanoo / ʼanooh in the 2sg row). (F3) The D - L form equals the imperfective in only 60/132 1sg cells, 46/116 3rd cells and 0/132 2sg cells. So Rice's "as if imperfective" holds literally for under half. (F4) A verb with a lexical d classifier and active meaning (V2) takes the D - L form.

HYPOTHESIS Why the 1st dual doesn't change. Y&M p. 779 say the stem-initial consonant shows the d-classifier effect "in the presence of D-Classifier, and in the 1st person dpl" (V3: 1dpl yiidą́ą́ʼ with d although the verb is Ø-class). If the 1dpl already brings the d-effect on its own, a D - L Voice adds nothing, and the two columns coincide (F2). That would make the 1dpl d a second source of the same context (T94d).

6. Derivations

Same notation as subject-d-site: B = syntax/clause steps, A = word spell-out steps. Targets: V1 yíyą́ą́ʼ and V2 yishdlą́ą́ʼ.

DERIVATION I: IN SITU (no Lowering)
1  Merge   [VP DP_obj V]; V = √yą́ą́' 'eat' (V1) or √dlą́ą́' 'drink' (V2)          B
           DP_obj = naayízí 'squash' / tóshchiin 'gruel', a full DP (analysis D)
2  Merge   Voice: Ø (V1); d (V2). The stem is listed with its classifier          B
           (Y&M p. 781 "(*d-dlą́ = dlą́)"); this is selection, not a passive.
3  Merge   dP_subj (1SG pro) in Spec,vP (SD-6)                                    B
4  Merge   M/S [PFV(yi), uφ]                                                      B
5  Agree   M/S probes, finds dP_subj, gets φ = 1SG                                  B
6  Merge   Qual (none overt here); clause continues                               B
   ---- the word is handed to spell-out: B → A ----
7  VI      V first (A1): yą́ą́' / dlą́ą́'                                             A
8  VI      Voice: Ø / d  (d + dl → dl: Y&M's "*d-dlą́ = dlą́", not derived here)   A
9  VI      M/S[PFV, 1SG]. Voice is spelled (A1), concatenated with M/S (VA-7,     A
           A2), and visible by form (VA-4). VA-6 picks:
             V1: elsewhere → yí-      V2: / __ Voice{d} → yish-
10 Spell   nuclear string: yí–Ø–yą́ą́' ;  yi–sh–(d)–dlą́ą́'                           A
11 Spell   nonnuclear: no overt object prefix; DP_obj pronounced as a phrase      A/B
           before the word: naayízí ... yíyą́ą́' ;  tóshchiin yishdlą́ą́'   ✓ V1, V2

DERIVATION II: WITH LOWERING (Hale's fallback, VA-2)
1–6        as in I                                                               B
7' Lower   M/S → Voice (head of its complement), before VI: [Voice Voice M/S]     B→A
           (skips Spec,vP: VA-2)
8' VI      V, then Voice, then M/S inside the complex head; VA-6 picks as in 9    A
9' Linear  the complex must be linearized M/S before Voice (prefix), against      A
           the usual right-adjunction for lowering T to v (English -ed)
10'–11'    as in I                                                               A

HYPOTHESIS Comparison. I and II pick the same allomorph from the same two heads. Lowering goes to "the head of its complement" (VA-2), and that head is Voice. So Lowering can't reach anything that I can't already see. II costs two things. One is the Lowering step itself. The other is an extra ordering statement (9'), because Hale's spine puts M/S to the left of Voice, while English T lowers and is pronounced to the right. Step 9 of I needs no new order: SD-1 already gives it.

HYPOTHESIS Why the subject does not get in the way. In the tree, Spec,vP lies between M/S and Voice (the left-headed spine: M/S [vP dP_subj [Voice ...]]). If VI looked at the whole linearized clause, the subject phrase would separate the two heads, and A2 would fail without Lowering. That is the strongest case for Hale's worry. It does not arise, because Hale's spell-out builds the nuclear string from the nuclear heads only (SD-1). Phrases are added afterwards (SD-2). That separation is needed anyway to put the object and subject DPs outside the word.

7. Answer to T93a (in part)

  1. HYPOTHESIS In situ is enough. M/S and Voice are a head and the head of its complement. They are concatenated in the nuclear string (VA-7). Voice is spelled first (A1). So the Voice-dependent perfective can be chosen with M/S in its base position (VA-8). Hale's n. 7 condition, "Mode/Subj properly situated relative to Voice", is met by the head–complement relation itself.
  2. SOURCED The DM sources agree that head movement is not required for this. Embick's A2, as Merchant quotes it, asks for linear adjacency. Merchant says spans "do not require head movement" and that contiguity is "not just inside complex heads" (VA-5). Merchant reports Bobaljik 2012: 13 as suggesting that the strict no-maximal-projection-boundary condition "holds only of root allomorphy ... not of affixal allomorphy", and M/S is affixal.
  3. HYPOTHESIS The conditioning is by form. Hale groups ł with Ø and l with d. Y&M's columns are named by the classifiers themselves, and V2's lexical d triggers the D - L form with an active meaning (F4). All three point to the context being Voice's exponent {d, l}. That is what Bobaljik 2000 allows for inward sensitivity (VA-4). A context "[passive]" would be the wrong kind of feature for an outer morpheme to see, and it would misclassify V2.
  4. OPEN What is not settled. Embick & Noyer 2001's own allomorphy cases (Bulgarian Def, p. 573; English T, n. 22) occur after Lowering or adjunction. If Embick 2010 (not opened) requires the two pieces to be in one complex head (M-word), Lowering returns. Merchant's report of Embick (A2 as "linear adjacency") does not require it. T94a.
  5. HYPOTHESIS Rice's "as if imperfective". F3 refines it. The D - L perfective is a third set, not the imperfective set, in a little over half the tables (F3). The robust facts are F1 and F2: the D - L set differs from the Ø - Ł set in the 1sg, the 3rd persons and the 2du, and mostly not in the 2sg and the 1du.

8. Word and clause; prefixes and nouns

  • HYPOTHESIS Focus 1 (A vs B). The n. 7 allomorphy is entirely a ruleset A event, between two nuclear heads. Nothing from the right-headed clause (ruleset B) takes part, and no clause-level movement feeds it. Step 5e of /subject-d-site/ (the deictic d) stays the one place where B-level material is pronounced away from its base inside the word. VA-7 shows that it lands above Qual, so it can't interfere with M/S–Voice.
  • HYPOTHESIS Focus 2 (prefixes and nouns). In V1 and V2 the object nouns naayízí and tóshchiin are full phrases before the verb word. The word has no overt object prefix, and the subject's person shows up only through Agree on M/S. This matches analysis D: the DP stays outside the word, and only a d head (here a silent one) could join it. The classifier listing in step 2 is like listed-d's listed prefix strings: the stem selects its Voice head, as it can select its d.
  • Historical note: the "roles, not words" idea is dropped and was not used here.

9. Limits

  • All paradigm data are from the CU Boulder interface, not checked against print. Some rows look misaligned. In the p. 50 table the PASS. A row has a form in the Ø - Ł column and a blank D - L cell. Blank cells are left out of the counts. The interface's own "[Guess a paradigm]" output was not used.
  • The counts compare the first space- or parenthesis-delimited form in each cell. Variants in parentheses (for example -(s)-) are ignored.
  • V1 and V2 differ in stem as well as classifier. The same stem with both classifiers (V3's passive yidą́ą́ʼ) is listed only for the 3rd person, without the 1st-singular passive, so it can't show the 1sg sh directly.
  • Segmentations of yíyą́ą́ʼ and yishdlą́ą́ʼ are ours. Surface phonology (the d absorbed into dl, tone on yí-) is not derived.
  • Embick 2010, Bobaljik 2012 and Rice 2000 are cited only through Merchant 2015 and Hale 2001.

10. Open questions

IDQuestionWhat would settle it
T94aDoes contextual allomorphy need the conditioning head in the same complex head (M-word) as the target? If so, M/S has to Lower to Voice after all.Embick 2010, chapters 2–3 (the concatenation definition), opened first-hand.
T94bWhy is the 2nd singular perfective the same with both classifier sets (F2), except in some si-perfective tables (p. 155 bííní / bísíní)? Is 2SG underspecified, so that VA-6 has no D - L entry for it, or does one 2SG exponent win everywhere?Y&M print tables. An analysis of the 2sg perfective -íní- in Hale or Rice.
T94cAre the 3rd-person z (Ø - Ł) vs s (D - L) pairs (e.g. p. 48) the same allomorphy as the 1sg sh, or phonology (z + d → s)?A sourced statement of the Navajo d-effect (Y&M grammar section; Kari 1976).
T94dY&M p. 779 says stem-initial y becomes d "in the presence of D-Classifier, and in the 1st person dpl". Is the 1st-dual-plural d the same Voice-like d? The 1du column identity (F2, 117/133) suggests that it is. If so, a subject φ value conditions the same d-effect, and how that d is placed between M/S and V bears on VA-7.Hale or Rice on the 1dpl d-effect; Y&M grammar text.
T94eDoes the perfective type (yi-, si-, w-) depend on the stem V, across Voice? If so, M/S is conditioned by the span ⟨V, Voice⟩ (VA-5), not by Voice alone.Y&M stem entries listing perfective type, checked against classifier.

Answered in part here: T93a (in situ is enough under A1/A2 and spans; the complex-head version is OPEN as T94a). T93b–d are untouched.

Two deictic d's in one word: the indefinite object comes before the fourth-person subject, except where a sound rule swaps them (T93c)

2026-11-11 audit pass 2. Page titles only: Navajo glottal stops in titles now print as ʼ (U+02BC), and titles that called a structure a “VPISH spine” now say vP/TP (VPISH is a hypothesis, never a node). No claim changed.

Working note. Kept as written (URL, data and status labels unchanged). Filed under The word as a spell-out domain; the consolidated argument is the essay word-edge. Rule IDs: rules index.

Restated on the unified structure (2026-11-15). Prefix order is restated as Grant’s hypothesis that headedness is fixed locally by Merge site, so the sublexical domain is linearized opposite to the phrasal domain and the prefixes are successive word-internal Merge and Agree sites (GRANT’S HYPOTHESIS, tested on the sourced template in §3.2; fit/misfit OPEN). A possible sublexical V-to-T is a separate, untested GRANT’S HYPOTHESIS (§3.3). Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §3. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

2026-11-11 audit pass 3. A rule block written with Hale 2003’s M/S head was not marked as older notation; it is now labeled as a historical record. dP (the light-d phrase) on this page is a HYPOTHESIS label, not an established category. See the pass 3 report.

Subject-d-site asked (T93c) what happens when the object is itself a deictic d. Option (c) says an object d stays a nonnuclear proclitic in place (SD-2), while a deictic subject d is pronounced at the top of the nuclear spine, so the object should come first. Young & Morgan's table for the indefinite-object prefix ʼa- (1987, p. 123) has exactly the needed cells: the 4th-person (3a) row, with the indefinite object and the 4th-person subject in one prefix string. In 13 of its 14 cells the object ʼ comes first (ʼaji-, daʼji-). The one exception is the plural future dazhʼdoo-. Across all 387 saved prefix tables the exception is regular: a glottal stop is never written directly before zh (0 cases), and wherever the two meet the order is zhʼ (455 cells). HYPOTHESIS The syntax gives object before subject, as (c) predicts, and a late sound rule swaps the two consonants when they end up side by side.

1. The question

T93c (from subject-d-site): when a deictic d is the object (the 3i object ʼa-, or a 4th-person object), does it stay nonnuclear (SD-2), as option (c) predicts, so that it precedes a deictic subject d? The page asked for "a sourced form with both a deictic object and a deictic subject". This page finds one combination (3i object with 4th-person subject) and leaves the other (4th-person object with 3i subject) OPEN.

2. Sources opened this run

  • SOURCED Young, Robert W. & William Morgan Sr. 1987. The Navajo Language: A Grammar and Colloquial Dictionary, revised edition. Prefix paradigm tables, read through the CU Boulder Y&M interface (verbs.colorado.edu/~mahu0110/ym/, show_paradigm.php), not from print. Main table: p. 123, ʼa-((2)), which the interface titles "3i obj. -subj.pron., Pos.IV-V". Also p. 116 ʼa-((2)), and the 3a rows of 387 distinct saved tables (title plus page).
  • SOURCED Hale, Ken. 2001. Navajo verb stem position and the bipartite structure of the Navajo conjunct sector. Linguistic Inquiry 32: 678–693. Re-read from the box text layer: the distributive plural da and iterative ná as nonnuclear adjuncts, p. 686, (15); §4 deictic subjects, pp. 689–691.

Not opened: Y&M's grammar sections on prefix phonology, Kari 1976, McDonough 2003. No source opened this run states a glottal-stop and zh reordering rule. That rule is ours (DO-7).

3. Structure assumed

[ Qual [ M/S [vP dP_subj [ Voice [VP dP_obj V ]]]]]   + adjuncts (da, ná)       (as on /subject-d-site/#structure)
 nuclear:    Qual  M/S  Voice  V                       (spelled into the stem skeleton, SD-1)
 nonnuclear: dP_obj, adjuncts da / ná                  (proclitic, SD-2)
 deictic subject d: moved to the top of the nuclear spine (option (c), step 5e)
status: hypothesis (Hale's tree plus VPISH, unchanged from subject-d-site)

4. Rules first

DO-1  OBJECT d IS NONNUCLEAR. An object, deictic or not, is the complement of V   HYPOTHESIS (option (c); Hale
      and is procliticized in place (SD-2). So it precedes every nuclear head      p. 682, p. 688 for ordinary
      and every element moved onto the nuclear spine.                               objects)
DO-2  DEICTIC SUBJECT d ON TOP OF THE SPINE. The 4th-person j(i)- is placed "to    SOURCED (Hale p. 689–691, SD-5)
      the right of Object and to the left of Qualifier".                            as placement; HYPOTHESIS as
                                                                                    step 5e of option (c)
DO-3  PREDICTION. 3i object 'a + 4th subject j(i) surface as  ' ... j  .           HYPOTHESIS (theorem of DO-1,
                                                                                    DO-2)
DO-4  DA, NÁ ARE NONNUCLEAR ADJUNCTS, pronounced leftmost here.                    SOURCED (Hale p. 686, (15));
                                                                                    the order da before 'a is
                                                                                    Y&M's surface (da'a-)
DO-5  'A LOSES ITS VOWEL after da- / ná- before another prefix syllable            SOURCED as a surface pattern
      (p. 123: da'ji-, ń'jí-, da'doo-, ń'deesh- vs 'aji-, 'adoo-).                  (Y&M p. 123); mechanism OPEN
DO-6  J BEFORE A VOWEL, ZH BEFORE A CONSONANT (4th-person marker).                 SOURCED as a distribution
      3a cells, 387 tables: j + vowel 1771, zh + consonant 1936, other 2.          (counts below); mechanism not
                                                                                    derived
DO-7  ' ZH → ZH '. When the glottal stop would stand directly before zh, the       HYPOTHESIS (ours; fits 455/455,
      two are pronounced in the reverse order. Applies after DO-5 and DO-6, so     0 counterexamples; no source
      in the phonology, after ruleset A has fixed the order.                        opened states it)

DO-2's words are Hale's (as quoted in SD-5). DO-5 and DO-6 are only descriptions of the table strings. DO-7 is a proposal.

5. Data

Young & Morgan give prefix strings here, not whole verbs, and no sentences. Cells are copied as the interface shows them. The Leipzig segmentations are ours (HYPOTHESIS). Abbreviations: INDF.OBJ = 3i object, 4.SBJ = 4th-person (Y&M "3a") subject, DISTR = distributive plural, ITER = iterative, FUT = Y&M's future column (not segmented here).

D1. Y&M 1987 p. 123, ʼa-((2)), 3a row, singular (IMPERFECTIVE / ITERATIVE / PERFECTIVE YI Ø - Ł / PERFECTIVE YI D - L / PERFECTIVE SI / FUTURE / OPTATIVE):
ʼaji- / níʼjí- (ńʼjí-) / ʼajíí- / ʼajoo- / ʼajiz-(-s) / ʼazhdoo- / ʼajó-
ʼaji-: ʼa-ji- INDF.OBJ-4.SBJ-
ʼazhdoo-: ʼa-zh-doo- INDF.OBJ-4.SBJ-FUT-

D2. Same table, 3a row, distributive plural:
daʼji- / ńdaʼji- / daʼjíi- / daʼjoo- / daʼjiz-(-s) / dazhʼdoo- / daʼjó-
daʼji-: da-ʼ-ji- DISTR-INDF.OBJ-4.SBJ-
dazhʼdoo-: da-zh-ʼ-doo- DISTR-4.SBJ-INDF.OBJ-FUT- (surface order)

D3. Same table, 3rd-person rows, for comparison (no deictic subject): singular ʼa- / náʼá- / … / FUTURE ʼadoo-; distributive daʼa- / ńdaʼa- / … / FUTURE daʼdoo-. Reversionary row: náʼásh- (1sg imperfective), FUTURE ńʼdeesh-.

SOURCED The facts on p. 123. (F1) In 13 of the 14 3a cells, the object ʼ precedes the 4th-person j. (F2) The one exception, dazhʼdoo-, is the only cell where the object has no vowel and the 4th-person marker is zh. (F3) Its singular twin ʼazhdoo- keeps the vowel of ʼa and keeps the object first. (F4) The object loses its vowel after da- or ná- (D2, D3 daʼdoo-, ńʼdeesh-).

The same table type on p. 116 (ʼa-((2)), no gloss in the interface title) has the same split: 3a ʼají-, ńʼjí-, daʼjí- against future ʼazhdoo- and dazhʼdoo-. Its perfective ʼazhní- keeps the vowel, so ʼ and zh are not adjacent there.

3a cells in 387 distinct Y&M tables, written with ʼ and j/zh (2001 cells)CountExamples (page)
ʼ directly before j + vowel378ʼadaʼji- (10), ʼádeʼjoo- (13), daʼji- (123)
ʼ, then a vowel, then j/zh358ʼazhdoo- (14, 123), ʼazhdi- (22)
zh directly before ʼ (then d 314, n 137, y 4)455ʼadazhʼdoo- (10), ʼádizhʼdi- (17), bazhʼní- (147), hazhʼdii- (387)
ʼ directly before zh0none
j directly before ʼ; zh + vowel before ʼ0none

Many of these ʼ are not the 3i object. Some are a 3i subject (Y&M's biʼdi-, 212/3, with 3a bizhʼdi-), and some are part of a listed prefix string (haʼa-, ʼeʼe-). So the large counts bear on DO-6 and DO-7, which are about sounds. Only p. 123 (and p. 116, untitled) bears on DO-3, which is about syntax.

6. Derivations

OPEN Older notation: the block below uses Hale 2003’s M/S (Mode/Subject) head, which is not a node in our current trees. It is kept as a historical record, not as one of our analyses.

DERIVATION I: da'ji- and dazh'doo- (Y&M p. 123, 3a distributive), option (c)
1  Merge   [VP dP_obj V];  dP_obj = [dP 'a [DP pro]]  (3i object: a deictic d)  B
2  Merge   Voice (Ø in the table's column)                                      B
3  Merge   dP_subj = [dP ji [DP pro]] in Spec,vP (SD-6)                          B
4  Merge   M/S [IMPF] (I-a)  or  the future heads (I-b), with uφ                 B
5  Agree   M/S finds dP_subj; φ = 3rd, "construed with the third person" (SD-5) B
6  Merge   da (distributive adjunct, nonnuclear: Hale p. 686)                    B
   ---- the word is handed to spell-out: B → A ----
7  Move    d_subj ji to the top of the nuclear spine (step 5e, DO-2)          B → A
8  VI      nuclear string bottom-up (SD-1):  ji + [M/S: Ø | doo] + Voice + V    A
9  Procl.  object 'a, then adjunct da, to the left (SD-2, DO-1, DO-4)            A
           I-a:  da + 'a + ji + ...          I-b:  da + 'a + ji + doo + ...
   ---- phonology ----
10 DO-5    'a → ' after da:     I-a  da'ji-         I-b  da'jidoo-
11 DO-6    ji → zh before C:    I-a  (vowel follows: j stays)   I-b  da'zhdoo-
12 DO-7    ' zh → zh ':          I-a  (not adjacent to zh)       I-b  dazh'doo-
OUTPUT     I-a  da'ji- ✓ (Y&M p. 123)      I-b  dazh'doo- ✓ (Y&M p. 123)

DERIVATION II: 'azhdoo- (singular future, same table)
1–8        as I-b, without step 6
9  Procl.  'a + ji + doo + ...
10 DO-5    no da or ná: 'a keeps its vowel
11 DO-6    'azhdoo-
12 DO-7    ' and zh are not adjacent (a stands between them): no swap
OUTPUT     'azhdoo- ✓ (Y&M p. 123)

HYPOTHESIS Why the order of rules matters. DO-7 has to apply after DO-5 and DO-6, because the swap needs both the vowel loss and the zh form. Derivation II is the check. It has the same syntax as I-b and the same zh, but the vowel of ʼa survives, and the object stays first. A syntactic account that placed the 4th-person d above the object only in the plural future would have to state that just for that cell. The sound rule needs no extra statement.

OPEN Not derived. Why ʼa loses its vowel after da- and ná- (DO-5), why ji loses its vowel before d/n (DO-6), and the inner make-up of the future doo-.

7. Answer to T93c (in part)

  • SOURCED A form with both deictic d's exists. Y&M p. 123 3a cells combine the 3i object ʼa- with the 4th-person subject ji- (D1, D2).
  • HYPOTHESIS The object comes first, as (c) predicts. 13 of 14 cells put the object first (F1). This is what DO-1 and DO-2 give: the object d is spelled in place as a proclitic, and the subject d is moved onto the spine, which is pronounced after all proclitics.
  • HYPOTHESIS The exception is a sound rule. dazhʼdoo- is the one cell where ʼ would directly precede zh, and across 387 tables that sequence never occurs (0 cases against 455 of zhʼ). DO-7 is our statement of that. No source opened this run states it.
  • OPEN The other combination. A 4th-person object (ha-/ho-) with a 3i or 4th-person subject was not found in a sourced form. T95b.
  • HYPOTHESIS Bearing on T93b. At least one reordering of these prefixes happens after the order is fixed, in the phonology. That fits Hale's description of the deictic subjects as "infixed" (p. 691). It does not decide whether step 5e itself is syntactic or post-syntactic.
  • Update 2026-11-05 evening PT: qualifier-d-order refines DO-2 / step 5e against thematic di: in six hodii- futures zh follows listed di (QD-5). That is a placement fact, separate from DO-7's late swap.

8. Word and clause; prefixes and nouns

  • HYPOTHESIS Focus 1 (A vs B). Ruleset B (the clause) builds object below subject, with VPISH. Ruleset A (the word) gives the object-first order by proclisis. The zhʼ forms show a third layer after A: the phonology can swap two adjacent consonants. Surface prefix order therefore can't be read straight off as syntactic order. The rules must say at which layer each order is fixed.
  • HYPOTHESIS Focus 2 (prefixes and nouns). Under analysis D both deictic prefixes are heads d of full dPs whose nouns are silent (pro) here. The two d's join the word in different ways (proclisis in place for the object, movement to the spine for the subject), but neither noun is inside the word. The ʼ and zh swap touches only the d's. Nothing in it needs the nouns.
  • Historical note: the "roles, not words" idea is dropped and was not used here.

9. Limits

  • All paradigm data are from the CU Boulder interface, not checked against print. The counts take every 3a cell, including parenthesized variants, from 387 distinct (page, title) tables saved on the box. Two cells do not fit DO-6 (p. 236 biníkázhí-, p. 70 ʼahídajni-). They may be typos in the interface and are left out of the claim.
  • Y&M's tables give prefix strings, not words in sentences. No verb with these strings and a translation was opened this run.
  • The interface title "3i obj." is the only basis for calling the p. 123 ʼa- an object. The p. 116 table has no such gloss.
  • DO-7 is our generalization from written forms. Whether it is a metathesis, or whether the 4th-person marker is placed by the phonology in some other way, is not settled.

10. Open questions

IDQuestionWhat would settle it
T95aIs DO-7 (ʼ zh → zh ʼ) a stated rule of Navajo phonology? Does it hold outside the 3a rows, for example with other fricatives (sh, s, z)?Y&M grammar sections on prefix phonology; Kari 1976; McDonough 2003. A count over all rows, not only 3a.
T95bWhen a 4th-person object (ha-/ho-) combines with a 3i or 4th-person subject, does the object come first?A Y&M verb entry or a Willie 1991 example with that combination.
T95cIn biʼdi- (Y&M 212/3) the ʼ is a 3i subject, yet the table has a 3a row, bizhʼdi-. What does the 4th-person marker express when the 3i is already the subject?Y&M p. 212 text; a verb entry with a 3a passive form and its translation.
T95dWhy does ʼa lose its vowel after da- and ná- (DO-5)? Is it the same rule for other CV object prefixes?Y&M p. 123 context; other object-prefix tables (bi-, shi-) after da-.
T95eDoes any syntactic process (4th-person binding, the yi-/bi- choice) treat a zhʼ form differently from a ʼj form? DO-7 predicts that none does.Paired examples in Y&M or Willie 1991.

Answered in part here: T93c (3i object with 4th-person subject: object first, one cell reordered by DO-7). Touched: T93b (one late reordering exists; step 5e not decided). T93d is untouched.

The space d ho-: an object d that comes before the 4th-person subject, and one verb that bans its noun (T95b in part; T85a, T85b)

2026-11-11 audit pass 2. Apostrophe fixes only. Navajo glottal stops typed as ' or ’ (for example in prefix names like ʼa- and ʼeʼe-, or words like Awééʼ) now print as ʼ (U+02BC), and a build step no longer turns the closing quote of an English translation inside a Navajo row into a glottal stop (‘…the girls.’, not ‘…the girls.ʼ). No form, gloss or translation wording changed.

Working note. Kept as written (URL, data and status labels unchanged). Filed under The word as a spell-out domain; the consolidated argument is the essay word-edge. Rule IDs: rules index.

Verbs on this page → dictionary: yíghááh ‘join; with -tah ‘among’: go among’

Restated on the unified structure (2026-11-15). Prefix order is restated as Grant’s hypothesis that headedness is fixed locally by Merge site, so the sublexical domain is linearized opposite to the phrasal domain and the prefixes are successive word-internal Merge and Agree sites (GRANT’S HYPOTHESIS, tested on the sourced template in §3.2; fit/misfit OPEN). A possible sublexical V-to-T is a separate, untested GRANT’S HYPOTHESIS (§3.3). Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §3. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

2026-11-11 audit pass 3. dP (the light-d phrase) on this page is a HYPOTHESIS label, not an established category. See the pass 3 report.

Young & Morgan (1987) describe ha-/ho-/hw- as the 3s object-subject pronoun ("space, area, 'things', impersonal it"). They add that the same prefix "and its variants also serve as the 3a object pronoun, corresponding to the forms with the 3a subj. pron., ji-" (p. 458). So it is the object exponent that T95b asks about. In every 4th-person (3a) cell of the saved prefix tables where it meets the 4th-person subject, ho comes first: 150 of 150 cells in 15 tables (hoji-, hozhdoo-), with 0 reversed. The verb entries give the same order (hojiniih 'he (4th) knows it', p. 419). HYPOTHESIS This is what DO-1 and DO-2 predict: an object d is spelled where it is merged, and the subject d is moved onto the spine. The second finding is about nouns. Y&M say of hashbįʼ 'build a hogan' (p. 417) that "the object cannot be represented by a noun --- i.e. one may not say hooghan hashbįʼ". Other verbs with the same ho- object do occur with a bare noun (naanish hoséłį́į́ʼ 'I caused work', p. 418). HYPOTHESIS So the ban belongs to the verb root, not to the d.

1. The question

T95b (from deictic-order): when a 4th-person object (ha-/ho-) combines with a 3i or 4th-person subject, does the object come first? DO-1 (object d spelled in place, as a proclitic) and DO-2 (deictic subject d moved to the top of the nuclear spine) predict that it does. Two older items on the noun side are also touched. T85a asks whether a noun's features must ever match the prefix. T85b asks whether a noun is ever obligatory. The converse of T85b, whether a noun is ever banned, is answered here for one verb.

2. Sources opened this run

  • SOURCED Young, Robert W. & William Morgan Sr. 1987. The Navajo Language: A Grammar and Colloquial Dictionary, revised edition. Read through the CU Boulder Y&M interface (verbs.colorado.edu/~mahu0110/ym/; nsearch.php for entries, show_paradigm.php for tables), not from print. Entries: ho- p. 456 def. 1 and p. 458 def. 25; jóhonaaʼéí p. 490 def. 14. The 51 entries whose text says "(ha-((1)))", pp. 374–530, mainly pp. 411–419. Tables: ha-((1)) (ho-, hw-), p. 417, and the 3a rows of 388 distinct saved tables (page plus title).

Not opened: Y&M's grammar chapter on Position IV-V, Kari 1976, Willie 1991, and any source on Navajo sibilant harmony. No sentence in which a human 4th-person object and a deictic subject share one verb was found.

3. Structure assumed

[ Qual [ M/S [vP dP_subj [ Voice [VP dP_obj V ]]]]]   + adjuncts (da, ná, PPs)    (as on /deictic-order/#structure)
 dP_obj  = [dP ho [DP pro | noun]]        (analysis D: d takes a DP; only d joins the word)
 dP_subj = [dP ji [DP pro]]               (4th-person subject, merged in Spec,vP: VPISH)
 nuclear:    Qual  M/S  Voice  V          (stem skeleton, SD-1)
 nonnuclear: dP_obj, da, ná               (proclitic, SD-2)
status: hypothesis (unchanged from subject-d-site and deictic-order; analysis D is Grant's proposal)

4. Rules first

HO-1  ONE EXPONENT, TWO VALUES. ho-/ha-/hw- spells an object (or subject) d    SOURCED as Y&M's description
      valued 3s ('space, area, things') and also the 3a object d, the object    (p. 456 def. 1, p. 458 def. 25);
      matching the 3a subject ji-.                                              "d" is HYPOTHESIS
HO-2  OBJECT ho IS NONNUCLEAR. As an object it is the complement of V and       HYPOTHESIS (DO-1 applied to ho)
      is procliticized in place, so it precedes the moved subject d.
HO-3  PREDICTION. ho + 4th subject surface as  ho ... j/zh  .                   HYPOTHESIS (theorem of HO-2, DO-2)
HO-4  FACT. 3a cells with ho and j/zh: ho first in 150 of 150 (15 tables);     SOURCED (count, interface data)
      in the verb entries opened, ho first in every listed 3a form.
HO-5  NOUN BAN. For hashbį' 'build a hogan' the 3s object "cannot be            SOURCED (Y&M p. 417, their
      represented by a noun": *hooghan hashbį'. Y&M give hooghan 'áshłééh     statement and star-like judgment)
      instead.
HO-6  THE BAN IS LEXICAL. Other verbs with the same ho- object occur with a     HYPOTHESIS (construal of the
      bare noun read as the affected or created thing (hałchxǫǫh p. 411,        nouns is from Y&M's
      hashłeeh p. 418). So the ban is a property of the root *łbį', not of d.  translations only)
HO-7  WHERE THE BAN IS STATED. The root *łbį' selects a dP_obj whose DP is      HYPOTHESIS (ours)
      silent (pro); Merge of [dP ho [DP hooghan]] with *łbį' fails.
HO-8  AREA PHRASES ARE ADJUNCTS. Where the area is named in a sentence with     HYPOTHESIS (8 of 12 entries
      a ho- object verb, it is a postpositional or locative phrase              opened; see table)
      (shighan binaagóó 'around my home'), adjoined in the clause (ruleset B),
      not the DP complement of d.

HO-1, HO-4 and HO-5 restate the source. HO-2, HO-3 and HO-6 to HO-8 are proposals.

5. Data

Cells and sentences are copied as the interface shows them. The Leipzig segmentations are ours (HYPOTHESIS). Abbreviations: 3S = Y&M's 3s "space" pronoun (here the object), 4.SBJ = 4th-person (Y&M "3a") subject, DISTR = distributive plural, FUT = future column, PFV = perfective, OPT = optative, CLF = classifier. Portmanteau mode-and-subject syllables are glossed as one unit.

S1. Y&M 1987 p. 458, ho- def. 25: "Pos. IV-V, 3s obj.-subj. pron., space, area, 'things' impersonal it. Ho-, and its variants also serve as the 3a object pronoun, corresponding to the forms with the 3a subj. pron., ji-." Also p. 456 def. 1: "Ho- occurs frequently as a thematic element, and serves as the 3a object pronoun. (See ha-.)" The 3a-object use is illustrated on p. 490 (def. 14, in the entry jóhonaaʼéí): hanáshááh 'I'm walking around him, I'm walking around a person', against bináshááh.

S2. Y&M p. 417, table ha-((1)) (ho-, hw-), 3a rows (IMPERFECTIVE / ITERATIVE / PERFECTIVE YI Ø-Ł / YI D-L / SI Ø-Ł / SI D-L / FUTURE / OPTATIVE):
singular hoji- / náhoji- / hojíí- / hojoo- / hojiz-(-s) / hojis- / hozhdoo- / hojó-
distributive dahoji- / ńdahoji- / dahojíí-(dahojis-) / dahojoo- / dahojiz-(-s) / dahojis- / dahozhdoo- / dahojó-
hoji-: ho-ji- 3S-4.SBJ-
dahozhdoo-: da-ho-zh-doo- DISTR-3S-4.SBJ-FUT-
The 3rd-person row of the same table is ha- / … / FUTURE hodoo-. The table has no 3i row.

S3. Y&M p. 419, hashniih 'to know it, to know all about it' (def. 1), neuter imperfective: hashniih, hóniih, haniih, hojiniih, hwiiʼniih, hohniih. From baa hashniih 'to praise him' (def. 2): perfective 3a hojiznih, future 3a hozhdoonih, optative 3a hojónih.
hojiniih: ho-ji-niih 3S-4.SBJ-know 'he/she (4th) knows it'
hozhdoonih: ho-zh-doo-nih 3S-4.SBJ-FUT-tell

S4. Y&M p. 416, hasséés 'to burn (weeds on an area), to burn it off (an area)' (def. 17): imperfective 3a hojizéés; perfective 3a hojíízeez, distributive dahojíízeez; future 3a hozdoozis; optative 3a hojózéés. Note the future has z, not zh (see §9).

SOURCED Count over the saved tables. In the 3a rows of 388 distinct tables, 150 cells contain both ho/hó/hw and the 4th-person j/zh. In all 150 the h-syllable comes first. No cell has j/zh before ho. The 15 tables are on pp. 71 (two), 72, 73, 398, 417, 451, 454, 457, 463, 464, 528, 531, 541, 548 and 630 (for example náhoji- 531, nihojí- 630, hozhnii- 457, hahozhdiyoo- 398). In most of them the ho is part of a listed prefix string with the 3s meaning, which HO-1 says is the same exponent. The count therefore shows the order of this exponent, not of a human 4th-person object in particular.

S5. Y&M p. 417, hashbįʼ 'to build a hogan' (def. 7): "(Although transitive, the direct object of the verb is the 3s pronoun, and the object cannot be represented by a noun --- i.e. one may not say "hooghan hashbįʼ". One says "hooghan ʼáshłééh" or "hashbįʼ".)"
a. *hooghan hashbįʼ (Y&M: "one may not say")
   hooghan hash-bįʼ hogan 3S.1SG-build.dwelling
b. hooghan ʼáshłééh (Y&M's alternative; ʼáshłééh 'to make it' is not segmented here)
c. ʼAakʼeedą́ą́ʼ shimá sání bá dahosiilbįʼ. 'We built a hogan for our grandmother last fall.'
   ʼaakʼeedą́ą́ʼ shi-má sání b-á da-ho-sii-l-bįʼ
   last.fall 1SG.POSS-mother old 3-for DISTR-3S-PFV.1PL-CLF-build.dwelling

S6. Y&M p. 418, hashłeeh 'to cause, to bring into being' (def. 8): Doo ʼáháshyą́ą dago biniinaa naanish hoséłį́į́ʼ. 'I caused work because of my stupidity.'
   naanish ho-sé-ł-į́į́ʼ work 3S-PFV.1SG-CLF-cause.PFV (first clause not segmented)
Also on the same entry: baa saad hashłeeh 'to complain about it' (saad 'word'), bił ʼanaaʼ hashłeeh 'to declare war on them' (ʼanaaʼ 'war').

S7. Y&M p. 411, hałchxǫǫh 'to ruin it, to mess it up (as a road)' (def. 5): ʼAdą́ą́dą́ą́ʼ nahóółtą́ągo ʼatiin nihitsʼą́ą́ʼ hóółchxǫʼ lá. 'Yesterday's rain has ruined our road.'
   ʼatiin nihi-tsʼą́ą́ʼ hóó-ł-chxǫʼ road 1PL-from 3S.PFV-CLF-ruin.PFV (first clause not segmented)

S8. Y&M p. 416, hasséés (def. 17): ʼAdą́ą́dą́ą́ʼ shighan binaagóó hóózeez. 'I burned (weeds) around my home yesterday.'
   ʼadą́ą́dą́ą́ʼ shi-ghan bi-naa-góó hóó-zeez
   yesterday 1SG.POSS-home 3-around-toward 3S.PFV.1SG-burn.PFV

S9. Y&M p. 419, hashtaał 'to sing' (def. 5): Sin bee hashtaał 'to sing a song'; Naaʼahóóhaidi Nóódaʼí biyiin bee hóótáál 'I sang a Ute song at the Fair.'
   sin b-ee hash-taał song 3-with 3S.1SG-sing

Transitive ha-((1)) entry opened (Y&M page)What names the area or thing in Y&M's sentencesKind
hashbįʼ 'build a hogan' (417)nothing; a noun is ruled out (S5)ban stated
hałchxǫǫh 'ruin it' (411)ʼatiin 'road' (S7)bare noun
hashłeeh 'cause, bring into being' (418)naanish, saad, ʼanaaʼ (S6)bare noun
hashniih 'know it' (419)Dziłkʼijí 'the Mountain Chant', naʼalgizhjí 'surgery', ʼazeeʼ naagháagi 'medicine'-jí / -gi forms
hasséés 'burn it off' (416)shighan binaagóó, shidáʼákʼehdi (S8)PP / locative
hashdééh 'clear it off' (417)shighan binaagóó 'around my hogan'PP
hashdleesh 'paint it (an area)' (417)shikin góneʼ 'inside my house'PP
hashchʼid 'paw an area' (417)bitsiitaa 'among his hair'PP
hassééh 'limber up' (416)bijáátah, bitahPP
hashdlaad 'plow it under' (417)bił łeeh, shidáʼákʼehdiPP / locative
hashchí 'make a cache' (417)nooh ("into a storage cyst or place of concealment"), tsétahdi, cháshkʼeh góyaalocative
hashchįįh 'anger him' (417)bá, bikʼijįʼ (the person, not the "things")PP

SOURCED The sentences and translations. HYPOTHESIS The "Kind" column. It is read off the forms and Y&M's English. No source opened says which of these phrases is the argument of ho-. hashtłʼó 'tie it to it' (p. 421) is listed under ha-((1)) too. It is left out because its own paradigm has 3o yiyii- and 3a jii- with no ho, so its object is not ho- (T96e).

6. Derivations

DERIVATION I: hojiniih 'he/she (4th) knows it' (Y&M p. 419), option (c)
1  Merge   [VP dP_obj niih];  dP_obj = [dP ho [DP pro]]  (3S object d)           B
2  Merge   Voice (Ø classifier)                                                 B
3  Merge   dP_subj = [dP ji [DP pro]] in Spec,vP (VPISH; SD-6)                   B
4  Merge   M/S [neuter IMPF], with uφ                                           B
5  Agree   M/S finds dP_subj; φ valued (4th person, "3a")                        B
   ---- the word is handed to spell-out: B → A ----
6  Move    d_subj ji to the top of the nuclear spine (DO-2)                   B → A
7  VI      nuclear string bottom-up (SD-1):  ji + M/S Ø + Voice Ø + niih         A
8  Procl.  object ho to the left (SD-2, HO-2)                                   A
           ho + ji + niih
   ---- phonology ----
9  DO-6    ji: vowel follows, j stays
OUTPUT     hojiniih ✓ (Y&M p. 419)

DERIVATION II: hozhdoonih (future 3a, same entry)
1–3        as I, with stem nih
4  Merge   the future heads (doo), with uφ                                      B
5–6        as I
7  VI      ji + doo + nih                                                       A
8  Procl.  ho + ji + doo + nih                                                  A
9  DO-6    ji → zh before a consonant: hozhdoonih
           DO-7 does not apply (no ' next to zh)
OUTPUT     hozhdoonih ✓ (Y&M p. 419)

DERIVATION III: *hooghan hashbį' against hoséłį́į́' with naanish (S5, S6), analysis D
III-a  Merge  [dP ho [DP hooghan]]                                             B
       Merge  that dP with the root *łbį'                                       B
       HO-7: *łbį' selects a dP_obj with a silent DP. Merge fails.   ✗  (*hooghan hashbį', Y&M p. 417)
III-b  Merge  [dP ho [DP pro]] with *łbį'                                       B
       ... as I, 1sg subject: hash-bį'.                              ✓  (hashbį', Y&M p. 417)
III-c  Merge  [dP ho [DP naanish]] with *ł-leeh 'cause to become'               B
       no selection for a silent DP; spell-out B → A:
       the DP naanish stays in the clause, to the left of the word (head-final B);
       only d ho joins the word (A):  naanish  ho-sé-ł-į́į́'          ✓  (Y&M p. 418)

DERIVATION IV: shighan binaagóó hóózeez (S8)
1  Merge   [dP ho [DP pro]] with the root *zéés; Voice; dP_subj (1sg) in Spec,vP B
2  Merge   PP [ [DP shighan] bi-naa-góó ] adjoined to VP or vP (HO-8)            B
           (P after its object: head-final, ruleset B)
   ---- B → A: the word ----
3  VI      nuclear string; object ho proclitic: hóó-zeez                         A
OUTPUT     shighan binaagóó hóózeez ✓ (Y&M p. 416)

HYPOTHESIS Why HO-7 must name the root. III-a and III-c have the same d, the same Voice type (ł) and the same clause position for the noun. Only the root differs. A ban stated on d (for example "ho takes no overt DP") would wrongly rule out III-c and S7. A ban stated on the root fits all three. One alternative is that in hashbįʼ the d takes no DP at all (a pro-form, T85f). That too has to be listed with the root, so it is not a different claim about where the ban lives.

OPEN Not derived. Why the root *łbįʼ 'build a dwelling' bans the noun. One guess is that the root already names the dwelling, so hooghan would add nothing. No source opened says this. Also not derived: how the mode and subject syllables fuse (sé, hóó), and the z of hozdoozis.

7. Answers

  • SOURCED The object exponent meets the 4th-person subject. ho- is the 3s and the 3a object pronoun (S1). Its 3a cells combine it with ji- (S2–S4).
  • HYPOTHESIS T95b, 4th-person subject: the object comes first. 150 of 150 cells and every verb-entry form opened put ho before j/zh. HO-2 and DO-2 predict this. Together with the 3i object on deictic-order, both deictic-like objects precede the 4th-person subject.
  • OPEN T95b, 3i subject. No table opened has ho- as an object with a 3i subject. The p. 417 table has no 3i row. One more cell type needs checking: in ʼahoʼnii- (p. 73) the 3a row is ʼahozhʼnii-, with a ʼ after zh. What that ʼ is was not checked.
  • OPEN T95b, human 4th-person object. The counts are for the exponent. They are not about verbs where ho- means 'a person'. hanáshááh (S1) is such a use, but its subject is 1sg.
  • SOURCED T85b, converse: a noun can be banned. Y&M state that hashbįʼ cannot take a noun object (S5).
  • HYPOTHESIS T85a in part: the ban is lexical, not about the noun's features. hooghan is ruled out with hashbįʼ. With the same ho-, naanish and ʼatiin are fine (S6, S7). No feature of hooghan clashes with ho- that would not also clash for ʼatiin 'road', an area noun. So the restriction is on the root (HO-6, HO-7), not a feature match between noun and prefix.

8. Word and clause; prefixes and nouns

HYPOTHESIS Focus 1 (A vs B). The postpositional phrase shighan binaagóó is built by ruleset B: the head -naa follows its object, as clause heads do. Inside the word, ruleset A places the object d ho before the subject d and before the stem. The same contrast sits in one sentence (Derivation IV). The noun and the PP are outside the word and follow B. Only the d is inside the word and follows A.

HYPOTHESIS Focus 2 (prefixes and nouns). Under analysis D the d always has a DP complement, and only d joins the word. hashbįʼ shows that a root can require that DP to be silent. hashłeeh shows that the same d with another root allows an overt one. So whether a noun can "fill" the prefix's slot is decided by the root's selection of the d's complement, not by the prefix alone. In most of the entries opened the area is named by an adjoined PP instead (HO-8). A PP of that kind is not the d's complement, so it does not count as filling the slot. This matches the vehicle facts on named-unnamed, where a bee-phrase names what a 3i leaves unnamed. Here sin bee hashtaał 'sing a song' (S9) has the song in a bee-phrase too.

Historical note: the "roles, not words" idea is dropped and was not used here.

9. Limits

  • All data are from the CU Boulder interface, not checked against print. The 51 "(ha-((1)))" entries are those the interface search returns, which may not be all of them.
  • Which phrase in S6–S9 is the argument of ho- is our reading of the forms and Y&M's English. In S7 ʼatiin could be a topic or a possessor-like phrase linked to nihitsʼą́ą́ʼ, not the DP complement of d. If so, S6 alone carries HO-6.
  • The 150-cell count includes thematic ho in listed prefix strings. HO-1 treats it as the same exponent, following Y&M. A count limited to verbs with a human 3a object was not possible with the data opened.
  • hozdoozis (S4) has z where DO-6 predicts zh. The stem has z and s, so this may be sibilant agreement with the stem, but no source opened states such a rule. It is not counted against HO-4, which is about order only.
  • In some hodii- tables the future 3a cell puts zh after di (ʼáhodizhdoo- p. 72, náhodizhdoo- p. 541, hodizhdiyoo- p. 454), while the other cells have hozhdii-. This does not affect the ho-first order, but it bears on where DO-2 puts the subject d relative to di (T96d). Update 2026-11-05 evening: answered in part on qualifier-d-order (6 flip cells; QD-5).

10. Open questions

IDQuestionWhat would settle it
T96aIs there a sourced verb with a human 4th-person object ho-/ha- and a 3i or 4th-person subject? Does the object come first there too?Y&M entries glossed 'him (a person)' with 3a forms; Willie 1991; Y&M grammar on 3a objects.
T96bDo Y&M state the noun ban (HO-5) for any other verb? Is it always a root that names its own object (dwelling, cache)?A search of Y&M entry text for "cannot be represented by a noun" and similar notes; print pp. 411–423.
T96cIn S7 is ʼatiin the DP complement of ho-, a topic, or linked to nihitsʼą́ą́ʼ?More hałchxǫǫh sentences; Willie 1991 on topics; a speaker.
T96dWhy does the 4th-person zh follow di only in some future cells of hodii- tables (pp. 72, 454, 541)? And is the z of hozdoozis (p. 416) sibilant agreement?Y&M print tables; Y&M on the future di-; a source on Navajo sibilant harmony (McDonough 2003, Sapir & Hoijer 1967). Update 2026-11-05 evening: order half answered in part on qualifier-d-order; sibilant half still OPEN (T97e).
T96eWhy is hashtłʼó (p. 421) listed under ha-((1)) when its 3a forms (jii-, hizhdoo-) have no ho and it takes a noun object (shilį́į́ʼ)?Print p. 421; the ha-((1)) prefix entry.

Answered in part here: T95b (4th-person subject: object first, 150/150; 3i subject and human object OPEN), T85a (one restriction found, and it is lexical, not a feature match), T85b (converse: a noun can be banned). Touched: T85f (a pro-form d for hashbįʼ is an alternative, not tested).

Where 4th-person zh sits relative to thematic di: in six hodii- future cells it follows di, so “subject d above every Qualifier” needs a finer spell-out rule (T96d in part)

Working note. Kept as written (URL, data and status labels unchanged). Filed under The word as a spell-out domain; the consolidated argument is the essay word-edge. Rule IDs: rules index.

Restated on the unified structure (2026-11-15). Prefix order is restated as Grant’s hypothesis that headedness is fixed locally by Merge site, so the sublexical domain is linearized opposite to the phrasal domain and the prefixes are successive word-internal Merge and Agree sites (GRANT’S HYPOTHESIS, tested on the sourced template in §3.2; fit/misfit OPEN). A possible sublexical V-to-T is a separate, untested GRANT’S HYPOTHESIS (§3.3). Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §3. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

2026-11-11 audit pass 3. dP (the light-d phrase) on this page is a HYPOTHESIS label, not an established category. 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. Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference.

Space-d left open (T96d) why some future 3a cells of hodii- tables write zh after di (ʼáhodizhdoo-, hodizhdiyoo-, náhodizhdoo-) while the same tables’ other modes have hozhdii-. The testability note (§3.3) asked whether that order falsifies “subject d above every spine Qualifier” (subject-d-site step 5e). This page re-opens the three named tables plus the related hodi-, dii- and naʼáhodii- paradigms. Across every saved 3a cell that contains both ho and di with a 4th-person marker, zh precedes di in 40 cells and follows it in 6. All six are futures of long hodii- tables. Short hodi- futures keep zh before di (hozhdidoo-, p. 451), and plain dii- futures put j before di (jididoo-, p. 343). SOURCED The six cells do not reverse ho and zh. HYPOTHESIS They refine step 5e: when a listed thematic di and future Mode co-occur, the 4th-person exponent is spelled between them, not above every di-looking Qualifier. That refinement is a morphological placement fact. It fits both a moving-d account and a stationary-d account with an independent placement rule; the tables do not choose between them.

1. The question

T96d (from space-d): why does the 4th-person zh follow di only in some future cells of hodii- tables (pp. 72, 454, 541)? And is the z of hozdoozis (p. 416) sibilant agreement? The first half is the subject of this page. It bears on subject-d-site step 5e, which places the deictic subject d “at the top of the nuclear spine,” above Qualifier. If every di in these strings is Qualifier, the six future cells put the subject d below a Qualifier. The second half (sibilant agreement) stays OPEN: no source on Navajo sibilant harmony was opened this run.

2. Sources opened this run

  • SOURCED Young, Robert W. & William Morgan Sr. 1987. The Navajo Language: A Grammar and Colloquial Dictionary, revised edition. Prefix paradigm tables via the CU Boulder Y&M interface (show_paradigm.php), not print. Focal tables: p. 72 ʼáhodii-((1,2)) (saved [local working copy, not in the repository]); p. 454 hodii-((2)) (252.html); p. 541 náhodii- (285.html); also p. 548 naʼáhodii- (290.html), p. 451 hodi-((2,4,5,7)) (251.html), p. 343 dii-((2,3)) (205.html). Count over 3a rows of all 410 saved tables that contain both ho/hó/hw and di with a 4th-person j/zh.
  • SOURCED Hale, Ken. 2001. Navajo verb stem position and the bipartite structure of the Navajo conjunct sector. Linguistic Inquiry 32: 678–693. Re-read §4 (pp. 689–691) on deictic subjects “to the left of Qualifier”; tree (6) and Qualifier as the highest nuclear head (p. 681). Box text layer [local working copy, not in the repository].
  • SOURCED (as site context) The light-determiner testability report (2026-11-05), §3.3, which flagged these future strings and warned that “whether this di is a relevant Qualifier and how it is segmented must be established” before treating step 5e as exceptionless.

Not opened: Y&M’s grammar chapters on Position VI/VII and the future, Kari 1976, McDonough 2003, Sapir & Hoijer 1967. No print check of the interface tables. The z of hozdoozis is not re-derived here.

3. Structure assumed

[ Qual [ M/S [vP dP_subj [ Voice [VP dP_obj V ]]]]]   + adjuncts (da, ná)
 nuclear:    Qual  M/S  Voice  V          (SD-1; Hale p. 681)
 nonnuclear: object d, da, ná             (SD-2)
 deictic subject d: displaced onto the nuclear string (option (c) step 5e, or I-interface placement)
status: hypothesis (unchanged from /subject-d-site/ and /space-d/; “d” is GRANT'S CONJECTURE)

OPEN Whether the listed thematic di of hodii- is the same category as Hale’s Qualifier d-, a lower thematic head, or part of a listed prefix string, is not settled by Y&M’s table labels. The page treats surface order only, then states what each analysis would need.

4. Rules first

QD-1  HALE’S SURFACE PLACE. Deictic subject markers stand “to the right of Object        SOURCED (Hale pp. 689–691,
      and to the left of Qualifier”.                                                    SD-5); “Qualifier” as Hale’s
                                                                                        category
QD-2  HO BEFORE ZH/J. In every 3a cell with both ho and j/zh, ho comes first.           SOURCED (HO-4; reconfirmed:
      The six zh-after-di cells keep ho first ('áhodizhdoo-, not *'ázhhodidoo-).       6/6)
QD-3  FACT (COUNT). 3a cells with ho + di + j/zh in the saved tables: zh before di     SOURCED (interface count;
      in 40; zh after di in 6. All 6 are FUTURE cells of long hodii- tables            segmentation of “di” is
      (pp. 72, 454, 541, 548).                                                          ours: first d[ií] in the cell)
QD-4  SHORT HODI- KEEPS ZH BEFORE DI IN THE FUTURE. hodi- p. 451 3a future is          SOURCED (p. 451 table)
      hozhdidoo- / dahozhdidoo-, not *hodizhdoo-.
QD-5  REFINEMENT (spell-out). When a listed thematic di and future Mode co-occur,      HYPOTHESIS (fits QD-3;
      the 4th-person exponent is spelled between them:  ho-di-zh-Mode- . When only     does not choose moving-d
      one di-looking nuclear piece is present (short hodi- future, or plain dii-),     vs stationary-d)
      the 4th-person exponent precedes that di:  zh/j-di-Mode- .
QD-6  STEP 5e IS NOT “ABOVE EVERY di”. If every di counted as Qualifier, the six       HYPOTHESIS (construal of
      cells would falsify an exceptionless reading of step 5e. Under QD-5 they do      QD-3 against subject-d-site)
      not: the target of placement is the left edge of future Mode, which may sit
      below a listed thematic di.
QD-7  PLAIN DII- CONTROL. dii- p. 343 3a future is jididoo- / dazhdidoo- (j/zh         SOURCED (p. 343); shows
      before future di), matching QD-5’s “single di” case with no thematic ho.         the order is not “future
                                                                                        always flips zh past di”
QD-8  P.454 DISTRIBUTIVE ASYMMETRY. Same table: singular future hodizhdiyoo-           SOURCED (p. 454); reason
      (zh after first di) vs distributive dahozhdiyoo- (zh before diyoo). Not          OPEN (T97a)
      derived.

QD-1, QD-2, QD-3, QD-4, QD-7 and QD-8 restate the opened tables (or Hale). QD-5 and QD-6 are proposals.

5. Data

Cells are copied as the CU interface shows them. Leipzig segmentations are ours (HYPOTHESIS). 4.SBJ = Y&M’s 3a subject; TH.DI = the listed thematic di of the table title; FUT = future Mode column.

Q1. Y&M 1987 p. 72, ʼáhodii-((1,2)), 3a rows (IMPERFECTIVE / ITERATIVE / PERFECTIVE / FUTURE / OPTATIVE):
singular ʼáhozhdii- / ʼánáhozhdii- / ʼáhozhdii- / ʼáhodizhdoo- / ʼnhozhdoo- (optative as printed)
distributive ʼádahozhdii- / ʼáńdahozhdii- / ʼádahozhdii- / ʼádahodizhdoo- / ʼádahozhdoo-
3rd-person future for comparison: ʼáhodidoo- / ʼadahodidoo-.
ʼáhozhdii-: ʼá-ho-zh-dii- …-3S?-4.SBJ-TH.DI-
ʼáhodizhdoo-: ʼá-ho-di-zh-doo- …-TH.DI-4.SBJ-FUT-

Q2. Y&M p. 454, hodii-((2)), 3a:
singular imperfective hozhdii-; perfective hozhdiiz-(-s) / hozhdiyeesh-; future hodizhdiyoo-; optative hozhdiyó-
distributive imperfective dahozhdii-; future dahozhdiyoo- (zh before diyoo); optative dahozhdiyó-
3rd-person future: hodidiyoo- / dahodidiyoo- (two di-pieces visible).
hodizhdiyoo-: ho-di-zh-di-yoo- TH.DI-4.SBJ-FUT.DI-…
dahozhdiyoo-: da-ho-zh-di-yoo- DISTR-…-4.SBJ-… (surface; internal cut OPEN)

Q3. Y&M p. 541, náhodii-, 3a:
singular náhozhdii- / nínáhozhdii- / náhozhdii- / náhodizhdoo- / náhozhdoo-
distributive ńdahozhdii- / … / ńdahodizhdoo- / ńdahozhdoo-
3rd-person future: náhodidoo-.

Q4. Y&M p. 548, naʼáhodii-, 3a (same long pattern): imperfective naʼáhozhdii-; future naʼáhodizhdiyoo-; optative naʼáhozhdiyó-. 3rd-person future naʼáhodidiyoo-.

Q5. Contrast — short hodi-, Y&M p. 451, 3a:
imperfective hozhdi-; future hozhdidoo- / distributive dahozhdidoo-.
Here future keeps zh before di, matching the imperfective order, not the long hodii- future flip.

Q6. Contrast — plain dii- (no thematic ho in the 4th-person row), Y&M p. 343, 3a:
imperfective jidii-; future jididoo-; distributive dazhdii- / dazhdidoo-.
The 4th-person marker precedes the future di. The 3s row of the same table has hodidoo- (areal subject, not 4th).

Table (Y&M page)3a imperfective (zh vs di)3a future (zh vs di)Pattern
ʼáhodii- (72)ʼáhozhdii- (zh before)ʼáhodizhdoo-, ʼádahodizhdoo- (zh after)long; future flips
hodii-((2)) (454)hozhdii- (zh before)sg hodizhdiyoo- (after); distr dahozhdiyoo- (before)long; sg flips, distr not
náhodii- (541)náhozhdii- (zh before)náhodizhdoo-, ńdahodizhdoo- (zh after)long; future flips
naʼáhodii- (548)naʼáhozhdii- (zh before)naʼáhodizhdiyoo- (zh after)long; future flips
hodi- (451)hozhdi- (zh before)hozhdidoo-, dahozhdidoo- (zh before)short; no flip
dii- (343)jidii- (j before)jididoo-, dazhdidoo- (j/zh before)no thematic ho; no flip

SOURCED Count. In the 3a rows of the 410 saved tables, 46 cells contain ho/hó/hw, a di-piece, and a 4th-person j/zh. Of those, 40 have zh before the first di; 6 have zh after it. The six are exactly the future cells listed in Q1–Q4 (singular and distributive where both flip). No cell reverses ho and zh.

6. Derivations

DERIVATION I: 'áhozhdii- (p. 72, 3a imperfective) — zh before thematic di
1  Merge   [VP … V]; Voice; dP_subj = [dP ji [DP pro]] in Spec,vP (SD-6)            B
2  Merge   M/S [IMPF]; Agree values φ = 4th on M/S (SD-4/SD-5)                       B
3  Merge   listed thematic di (whatever category it is: Qual, theme head, or         B
           listed string — OPEN as category; present as a pronounced piece)
   ---- B → A ----
4  Spell   nuclear bottom-up (SD-1); place 4th d to the left of the single            A
           di-piece (QD-5, single-di case; or step 5e as “left of Qual”)
5  Procl.  ho (and 'á) nonnuclear (SD-2, HO-2)                                       A
6  DO-6    ji → zh before consonant: 'á-ho-zh-dii-
OUTPUT     'áhozhdii- ✓ (Y&M p. 72)

DERIVATION II: 'áhodizhdoo- (p. 72, 3a future) — zh between thematic di and Mode
1–2        as I, with future Mode (doo)                                              B
3  Merge   thematic di AND future Mode as separate pronounced pieces                 B
           (supported on the surface by p. 454’s hodidiyoo- with two di-pieces)
   ---- B → A ----
4  Spell   nuclear; place 4th d to the left of future Mode, after thematic di (QD-5) A
5  Procl.  'á-ho-                                                                   A
6  DO-6    ji → zh before consonant: 'á-ho-di-zh-doo-
OUTPUT     'áhodizhdoo- ✓ (Y&M p. 72)

DERIVATION III: hozhdidoo- (p. 451, 3a future, short hodi-) — no flip
1–2        as II
3  Merge   only one di-looking piece is pronounced with future Mode                  B
           (surface: ho-di-doo in the 3rd person; not ho-di-di-…)
4  Spell   4th d to the left of that single di (QD-5 single-di case)                 A
OUTPUT     hozhdidoo- ✓ (Y&M p. 451)

DERIVATION IV (schematic): stationary-d I-interface vs moving-d II on the same string
II   (option (c)): d moves / is infixed to the left edge of future Mode (QD-5).      HYPOTHESIS
     Thematic di is already in the nuclear string to its left.
I-if (stationary d): d stays in Spec,vP syntactically; an independent placement      HYPOTHESIS
     rule pronounces its exponent in the same between-di slot.
Both yield 'áhodizhdoo-. The table string does not decide which representation is right.

HYPOTHESIS Why QD-5 must mention two pieces. Derivation II needs thematic di and future Mode as separately ordered exponents. P. 454’s 3rd-person hodidiyoo- shows two di-pieces in the future of a long table. P. 451’s 3rd-person hodidoo- shows one. The 4th-person rows track that difference: between the two pieces when both are present; before the one piece when only one is.

OPEN Not derived. Why long hodii- futures pronounce two pieces and short hodi- futures pronounce one (Y&M’s own analysis of the future was not opened). Why p. 454 distributive future fails to flip (QD-8). The z of hozdoozis (T96d second half).

7. Answers

  • SOURCED T96d, first half (order). In the opened tables, 4th-person zh follows thematic di in exactly six future cells of long hodii- paradigms (pp. 72, 454 sg, 541, 548). It precedes di in the 40 other ho+di+4th cells, including all non-future cells of those same tables and the futures of short hodi- (p. 451).
  • HYPOTHESIS Step 5e refined, not flatly falsified. An exceptionless reading “subject d above every Qualifier-di” fails on the six cells if thematic di is Qualifier. QD-5 keeps Hale’s “left of (future) Qualifier/Mode” for the placement target and allows a listed thematic di to its left. HO-4 (ho first) is untouched.
  • OPEN T96d, second half (z of hozdoozis). No sibilant-harmony source was opened. Still OPEN.
  • HYPOTHESIS Stationary-d vs moving-d. Both need the same placement generalization (QD-5). Prefix order here is compatible with I-interface and with II; it does not prove d-movement (testability report §4).

8. Word and clause; stationary d

HYPOTHESIS Focus 1 (A vs B). The between-di slot is a fact about ruleset A (word spell-out). Nothing in these tables shows where the DP complement of a subject d sits in the clause (ruleset B). QD-5 is therefore a constraint on how the exponent of d is linearized in the word, whether or not d moved in the syntax.

GRANT'S CONJECTURE Focus 2 (stationary d). Grant’s 2026-11-05 correction keeps d in place as the head of its projection while the DP moves. Under that conjecture, the six cells are evidence about the morphological placement rule for stationary d (I-interface), not about head movement of d. The same cells are also what a refined step 5e predicts under option (c). The page records the shared empirical content and leaves the choice of representation OPEN.

Historical note: the “roles, not words” idea is dropped and was not used here.

9. Limits

  • All counts are from the CU Boulder interface, not checked against print. Cell strings can contain OCR-like typos in the interface (p. 72 optative ʼnhozhdoo- looks truncated); such cells were not used as evidence for QD-5.
  • “First di” is a string heuristic. Whether that piece is Hale’s Qualifier, a thematic Mode-dependent allomorph, or a listed prefix is not given by Y&M’s table headers alone.
  • Thematic ho in these tables is often the 3s/areal listed element (HO-1), not a demonstrated human 4th-person object. The order finding is about exponents, as on space-d.
  • P. 454 distributive dahozhdiyoo- is a real counterexample to a blanket “all long hodii- futures flip” rule; QD-5 covers the six flip cells and leaves this cell to T97a.

10. Open questions

IDQuestionWhat would settle it
T97aWhy does p. 454 distributive future dahozhdiyoo- keep zh before diyoo while singular hodizhdiyoo- flips?Y&M print p. 454; Y&M on distributive + future; more long hodii- distributive futures.
T97bDo Y&M (or Kari) treat the thematic di of hodii- and the future di as two Position-VI/VII heads, or as one fused piece?Y&M grammar on the future and on Pos. VI; Kari 1976.
T97cIs there any long hodii--type table whose future 3a does not flip, other than p. 454 distributive?Full pass over print paradigms; remaining interface tables without ho in the title.
T97dDoes the same between-di placement appear with 3i subject ʼ (e.g. future cells parallel to Hale’s (22b)) when a thematic di is listed?Saved tables with 3i + thematic di + future; Hale 2001 (22b) has Qual d after ʼ, which is the non-flip order.
T97eIs the z of hozdoozis (p. 416) sibilant agreement with the stem (T96d second half)?McDonough 2003; Sapir & Hoijer 1967; Y&M phonology.

Answered in part here: T96d (order half: six flip cells; refinement of step 5e as QD-5; sibilant half still OPEN). Touched: subject-d-site step 5e; testability §3.3; Grant’s stationary-d conjecture (compatible, not proved).

Distributive future and the missing di: p. 454 dahozhdiyoo- keeps zh before diyoo because that cell has only one di-piece, while singular hodizhdiyoo- has two (T97a, T97c)

Working note. Kept as written (URL, data and status labels unchanged). Filed under The word as a spell-out domain; the consolidated argument is the essay word-edge. Rule IDs: rules index.

Restated on the unified structure (2026-11-15). Prefix order is restated as Grant’s hypothesis that headedness is fixed locally by Merge site, so the sublexical domain is linearized opposite to the phrasal domain and the prefixes are successive word-internal Merge and Agree sites (GRANT’S HYPOTHESIS, tested on the sourced template in §3.2; fit/misfit OPEN). A possible sublexical V-to-T is a separate, untested GRANT’S HYPOTHESIS (§3.3). Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §3. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

2026-11-11 audit pass 3. dP (the light-d phrase) on this page is a HYPOTHESIS label, not an established category. 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. Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference.

Qualifier-d-order left open why the same long hodii- table (Y&M p. 454) writes singular future hodizhdiyoo- (zh after the first di) but distributive future dahozhdiyoo- (zh before diyoo). Re-counting every saved long hodii--type 3a future, and pairing each with its 3rd-person twin, shows that the distributive cell is the only non-flip in that class — and that it is also the only cell in that class whose 3a future has one di-piece while the matching 3rd-person future has two. SOURCED Singular 3a hodizhdiyoo- matches 3rd hodidiyoo- in piece count (two); distributive 3a dahozhdiyoo- does not match 3rd dahodidiyoo- (one vs two). HYPOTHESIS Under QD-5, placement tracks the number of pronounced di-pieces in the 3a cell: between them when two are present, before the one piece when only one is. The p. 454 distributive cell is then the single-di case, not a counterexample. Why that cell drops a piece stays OPEN.

1. The question

T97a (from qualifier-d-order): why does p. 454 distributive future dahozhdiyoo- keep zh before diyoo while singular hodizhdiyoo- flips? T97c: is there any other long hodii--type table whose future 3a does not flip? Both bear on QD-5 (4th-person exponent between thematic di and future Mode when both are pronounced) and on QD-8 (the p. 454 asymmetry, previously underived).

2. Sources opened this run

  • SOURCED Young, Robert W. & William Morgan Sr. 1987. The Navajo Language: A Grammar and Colloquial Dictionary, revised edition. Prefix paradigm tables via the CU Boulder Y&M interface (show_paradigm.php), not print. Focal: p. 454 hodii-((2)) (saved [local working copy, not in the repository]); also p. 72 ʼáhodii-((1,2)) (55.html), p. 541 náhodii- (285.html), p. 548 naʼáhodii- (290.html), short control p. 451 hodi- (251.html), and non-flip controls p. 464 hwii- (255.html), p. 398 hahwii--type (228.html). Full pass: every 3a FUTURE cell among the 410 saved tables that contains ho/hó/hw, a di-piece, and 4th-person j/zh; plus every table whose 3rd-person FUTURE has two or more di-pieces.
  • SOURCED Hale, Ken. 2001. Navajo verb stem position and the bipartite structure of the Navajo conjunct sector. Linguistic Inquiry 32: 678–693. Re-read pp. 626–627 (future = inceptive Qualifier d plus the progressive complex) and §4 on deictic subjects left of Qualifier. Box text layer [local working copy, not in the repository].
  • SOURCED (as site context) qualifier-d-order (QD-1–QD-8; six flip cells; T97a–e).

Not opened: Y&M print p. 454; Y&M grammar on Position VI/VII and the distributive; Kari 1976. No judgment outside the interface cells.

3. Structure assumed

[ Qual [ M/S [vP dP_subj [ Voice [VP … V ]]]]]   + adjuncts (da, ná)
 nuclear:    Qual  M/S  Voice  V          (SD-1; Hale p. 681)
 future:     inceptive Qualifier d + progressive complex (Hale pp. 626–627) — SOURCED as Hale’s analysis
 listed thematic di of hodii-: a separate pronounced piece in long tables (surface; category OPEN)
 deictic subject d: displaced onto the nuclear string (QD-5 placement)
status: hypothesis (unchanged from /qualifier-d-order/; “d” is GRANT'S CONJECTURE)

OPEN Whether the listed thematic di is Hale’s Qualifier, a lower thematic head, or a listed string is still not settled by table titles. The page counts pronounced di-pieces in the cell strings.

4. Rules first

DF-1  HALE FUTURE. The future is the progressive complex “preceded by the                 SOURCED (Hale pp. 626–627)
      inceptive qualifier [d]”. So future Mode can contribute its own d-piece,
      separate from a listed thematic di.
DF-2  P.454 PIECE COUNT (sg). 3rd future hodidiyoo- has two di-pieces; 3a future           SOURCED (p. 454 table)
      hodizhdiyoo- has two, with zh between them (ho-di-zh-diyoo-).
DF-3  P.454 PIECE COUNT (distr). 3rd future dahodidiyoo- has two di-pieces; 3a              SOURCED (p. 454 table)
      future dahozhdiyoo- has ONE (da-ho-zh-diyoo-), with zh before that piece.
DF-4  T97c COUNT. Among saved long hodii-type tables that show any future flip              SOURCED (interface:
      (pp. 72, 454, 541, 548), the only 3a future that does not flip is p. 454              7 flip-class 3a futures
      distributive dahozhdiyoo-. p. 548 has no distributive 3a row in the saved table.    counted; 1 non-flip)
DF-5  QD-5 RESTATED ON PIECE COUNT. When the 3a future cell itself pronounces two          HYPOTHESIS (refines
      di-pieces, the 4th-person exponent is spelled between them (flip). When it           QD-5 / QD-8; fits DF-2,
      pronounces one, the exponent precedes that piece (no flip) — the short hodi-         DF-3, QD-4)
      pattern (hozhdidoo-, p. 451).
DF-6  P.454 DISTR IS THE SINGLE-di CASE. Under DF-5, dahozhdiyoo- is predicted once         HYPOTHESIS (theorem of
      the cell is taken as having one di-piece; it is not a QD-5 counterexample.           DF-3 + DF-5)
DF-7  DROP NOT DERIVED. Why p. 454 distributive 3a future lacks the second di that         OPEN (T98a)
      3rd dahodidiyoo- shows (*dahodizhdiyoo- is the mechanical insert-zh form) is
      not given by the tables.
DF-8  NO OTHER LONG hodii NON-FLIP. Outside the flip-class long hodii tables, short         SOURCED as classification
      hodi- (p. 451) and hwii-/hahwii-type futures (pp. 464, 398) keep zh before di         (their 3rd futures show
      in both singular and distributive; those tables’ 3rd-person futures show one         one di-piece)
      di-piece, so they are not long hodii-type for T97c.

DF-1–DF-4 and DF-8 restate opened sources or counts. DF-5 and DF-6 refine QD-5/QD-8. DF-7 is OPEN.

5. Data

Cells are copied as the CU interface shows them. Leipzig segmentations and “di-piece” counts are ours (HYPOTHESIS): a di-piece is a substring matching di/dí/dii/diy as the start of a prefix chunk in the cell. 4.SBJ = Y&M’s 3a; DISTR = distributive; FUT = future column.

D1. Y&M 1987 p. 454, hodii-((2)), FUTURE column (saved 252.html):
singular 3 hodidiyoo-; singular 3a hodizhdiyoo-
distributive 3 dahodidiyoo-; distributive 3a dahozhdiyoo-
Imperfective 3a for comparison: singular hozhdii-; distributive dahozhdii- (zh before thematic dii in both).
hodidiyoo-: ho-di-di-yoo- …-TH.DI-FUT.DI-… (two di-pieces)
hodizhdiyoo-: ho-di-zh-di-yoo- TH.DI-4.SBJ-FUT.DI-… (two di-pieces; zh between)
dahodidiyoo-: da-ho-di-di-yoo- DISTR-… (two di-pieces)
dahozhdiyoo-: da-ho-zh-di-yoo- DISTR-…-4.SBJ-… (one di-piece; zh before)

D2. Same table, 1sg future as a piece-count proxy: hodidiyeesh- (two di-pieces), matching the long pattern. Short control p. 451 1sg future hodideesh- (one di-piece).

D3. Other long hodii- futures (flip class) — singular and distributive 3a both flip where both rows exist:
p. 72 ʼáhodii-: ʼáhodizhdoo- / ʼádahodizhdoo- (zh after di in both)
p. 541 náhodii-: náhodizhdoo- / ńdahodizhdoo- (zh after in both)
p. 548 naʼáhodii-: singular naʼáhodizhdiyoo- (zh after; two di-pieces, twin of 3rd naʼáhodidiyoo-). No distributive 3a row in the saved interface table.

D4. Short / one-di controls (no flip), for QD-5’s single-di clause:
p. 451 hodi-: 3a future hozhdidoo- / dahozhdidoo-; 3rd hodidoo- / dahodidoo- (one di-piece each)
p. 464 hwii-: 3a hozhdiyoo- / dahozhdiyoo-; 3rd hodiyoo- / dahodiyoo- (one di-piece; not a long hodii- table)
p. 398 hahwii--type: 3a hahozhdiyoo- / hadahozhdiyoo- (zh before; one di-piece).

Table (Y&M page)3 FUT (di-pieces)3a FUT (di-pieces; zh vs di)Flip?
hodii-((2)) sg (454)hodidiyoo- (2)hodizhdiyoo- (2; after)yes
hodii-((2)) distr (454)dahodidiyoo- (2)dahozhdiyoo- (1; before)no
ʼáhodii- sg/distr (72)ʼáhodidoo- / ʼadahodidoo- (1 visible)ʼáhodizhdoo- / ʼádahodizhdoo- (after)yes / yes
náhodii- sg/distr (541)náhodidoo- / ńdahodidoo- (1 visible)náhodizhdoo- / ńdahodizhdoo- (after)yes / yes
naʼáhodii- sg (548)naʼáhodidiyoo- (2)naʼáhodizhdiyoo- (2; after)yes
hodi- sg/distr (451)hodidoo- / dahodidoo- (1)hozhdidoo- / dahozhdidoo- (1; before)no / no

SOURCED T97c. In the flip-class long hodii- futures of the 410 saved tables, 6 cells flip (as on qualifier-d-order) and exactly 1 does not: p. 454 distributive dahozhdiyoo-. That cell is also the only flip-class 3a future whose di-piece count is lower than its 3rd-person twin’s.

SOURCED Broader check (not required for T97c, recorded for DF-5): every saved table whose 3rd-person future has two or more di-pieces and whose 3a future keeps two pieces with a zh places that zh after the first di (between-di), including non-ho titles such as p. 17 ʼádizhdiʼdoo-, p. 315 dizhdidoo-, p. 383 hadizhdínóo-. The p. 454 distributive cell remains the unique drop-to-one among the long hodii- set.

6. Derivations

DERIVATION I: hodizhdiyoo- (p. 454, 3a singular future) — two di-pieces; zh between
1  Merge   [VP … V]; Voice; dP_subj = [dP ji [DP pro]] in Spec,vP                     B
2  Merge   M/S [FUT]; Agree values φ = 4th on M/S                                      B
3  Merge   thematic di AND future Mode’s di-piece as separate exponents                B
           (Hale DF-1; surface twin hodidiyoo- has two di-pieces)
   ---- B → A ----
4  Spell   place 4th d between the two di-pieces (DF-5 / QD-5)                         A
5  Procl.  ho-                                                                         A
6  DO-6    ji → zh before consonant: ho-di-zh-diyoo-
OUTPUT     hodizhdiyoo- ✓ (Y&M p. 454)

DERIVATION II: dahozhdiyoo- (p. 454, 3a distributive future) — one di-piece in 3a cell
1–2        as I, with distributive da                                                     B
3  Merge   only ONE di-looking piece is pronounced in the 3a future cell               B
           (SOURCED surface: dahozhdiyoo-, not *dahodizhdiyoo-; contrast 3rd
           dahodidiyoo- with two pieces — the drop itself is OPEN, DF-7)
   ---- B → A ----
4  Spell   place 4th d to the left of that single di-piece (DF-5 single-di case)       A
5  Procl.  da-ho-                                                                      A
6  DO-6    ji → zh before consonant: da-ho-zh-diyoo-
OUTPUT     dahozhdiyoo- ✓ (Y&M p. 454)

DERIVATION III (schematic): what a no-drop distributive would have been
If 3a had kept both di-pieces, DF-5 would yield *dahodizhdiyoo- (between-di),            HYPOTHESIS
matching singular. The interface cell is not that string. Tables do not say whether
the missing piece is haplology, a listed allomorph, or an interface/print issue.

DERIVATION IV: stationary-d vs moving-d on both strings
II   (option (c)): d is placed at the left edge of future Mode when two pieces           HYPOTHESIS
     are present, else at the left edge of the single di-piece.
I-if (stationary d): the same placement rule pronounces the exponent; d does not         HYPOTHESIS
     move as a head (Grant’s conjecture).
Both yield the observed cells once the piece count of the cell is given. The tables
do not choose between I-interface and II.

HYPOTHESIS Derivations I–II treat QD-5 as sensitive to the output piece count of the 3a cell. That unifies the p. 454 asymmetry with short hodi- futures (QD-4) without a special distributive rule.

7. Answers

  • SOURCED T97a (fact half). Singular and distributive 3a futures of p. 454 differ in di-piece count: hodizhdiyoo- has two (zh between); dahozhdiyoo- has one (zh before). The 3rd-person twins both have two. So the asymmetry is not “distributive refuses to flip while keeping the same pieces”; the distributive 3a cell has fewer pieces.
  • HYPOTHESIS T97a (why, under QD-5). DF-5/DF-6: placement follows the 3a cell’s own piece count. Once dahozhdiyoo- is a one-di string, zh-before is exactly what QD-5’s single-di clause (and short hodi-) already say. QD-8 is refined, not left as a bare exception.
  • OPEN T97a (why the drop). Why distributive 3a lacks the second di that 3rd dahodidiyoo- shows is not answered (DF-7 / T98a).
  • SOURCED T97c. No other long hodii--type future 3a in the 410 saved tables fails to flip. Among flip-class tables (pp. 72, 454, 541, 548), only p. 454 distributive is non-flip; p. 548 has no distributive 3a row saved.

8. Word and clause; stationary d

HYPOTHESIS Focus 1 (A vs B). Piece count and zh placement are facts about ruleset A (word spell-out). Nothing here shows where the DP complement of subject d sits in the clause.

GRANT'S CONJECTURE Focus 2 (stationary d). Grant’s 2026-11-05 correction keeps d in place as the head while the DP moves. Under that conjecture, DF-5 is an I-interface placement rule for the exponent of stationary d. The same strings are what a refined option-(c) step 5e predicts under moving-d. The tables still do not choose.

Historical note: the “roles, not words” idea is dropped and was not used here.

9. Limits

  • All counts are from the CU Boulder interface, not checked against print. A print check of p. 454 distributive future could still show dahodizhdiyoo- or another string; if so, DF-3 would need revision.
  • “Di-piece” is a string heuristic. On p. 72 and p. 541 the 3rd-person futures show only one visible di (ʼáhodidoo-, náhodidoo-) while 3a still flips (ʼáhodizhdoo-); Hale’s future-d (DF-1) plus thematic di still supplies two abstract pieces, but the 3rd-person Mode vowel (doo) does not look like di. The p. 454 argument uses the clearer two-visible-di twin.
  • p. 548’s missing distributive 3a row means T97c’s “no other non-flip” is only as strong as the saved tables.
  • Broader two-di tables without ho support between-di placement when 3a keeps two pieces, but some ji-initial cells (e.g. p. 329 jididinóo-) place the 4th-person marker before both pieces; those are not reanalyzed here.

10. Open questions

IDQuestionWhat would settle it
T98aWhy does p. 454 distributive 3a future lack the second di that 3rd dahodidiyoo- shows? Haplology, listed allomorph, or interface/print error?Y&M print p. 454; other long hodii- + diyoo distributive 3a rows (p. 548 print); Y&M/Kari on distributive + future.
T98bOn pp. 72 and 541, 3rd-person futures show one visible di but 3a still flips — is the second piece Hale’s future d absorbed into doo?Y&M on future Mode shape; Hale 2001 (24a) vs table strings.
T98cDo ji-initial two-di futures (p. 329 jididinóo-) need a separate left-edge rule, or a different segmentation of di?More ji vs zh two-di paradigms; phonology of ji/zh before d.
T97b(still open) Do Y&M or Kari treat thematic di of hodii- and future di as two heads?Y&M grammar; Kari 1976.
T97d(still open) Between-di placement with 3i subject ʼ?Saved 3i + thematic di + future cells.
T97e(still open) z of hozdoozis?McDonough 2003; Sapir & Hoijer 1967.

Answered here: T97a in part (piece-count fact SOURCED; QD-5 refinement HYPOTHESIS; drop reason OPEN); T97c (SOURCED: no other long hodii- non-flip in the saved tables). Touched: QD-5, QD-8, Hale future-d, Grant’s stationary-d conjecture (compatible, not proved).

Hale’s future Qualifier d is the onset of doo: pp. 72 and 541 flip between thematic di and doo though 3rd shows one visible di (T98b)

Working note. Kept as written (URL, data and status labels unchanged). Filed under The word as a spell-out domain; the consolidated argument is the essay word-edge. Rule IDs: rules index.

Restated on the unified structure (2026-11-15). Prefix order is restated as Grant’s hypothesis that headedness is fixed locally by Merge site, so the sublexical domain is linearized opposite to the phrasal domain and the prefixes are successive word-internal Merge and Agree sites (GRANT’S HYPOTHESIS, tested on the sourced template in §3.2; fit/misfit OPEN). A possible sublexical V-to-T is a separate, untested GRANT’S HYPOTHESIS (§3.3). Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. 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, §3. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

2026-11-11 audit pass 3. dP (the light-d phrase) on this page is a HYPOTHESIS label, not an established category. 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. Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference.

Distributive-future-d left open (T98b) why long hodii- futures on Y&M pp. 72 and 541 write only one visible di in the 3rd-person cell (ʼáhodidoo-, náhodidoo-) while 3a still flips (ʼáhodizhdoo-, náhodizhdoo-). Hale 2001 derives surface future doo from inceptive Qualifier d plus progressive Mode w, so that Qualifier is the onset of doo, not a second substring di. SOURCED Rechecked interface cells match: one visible di in 3rd, zh between that di and doo in 3a; p. 454’s diyoo twin makes both pieces look like di. HYPOTHESIS DF-5’s “piece count” is abstract under Hale (thematic di + Qualifier-onset-of-Mode), not a string-count of the substring di. For T98c, rechecked ji-initial two-di futures: p. 329 jididinóo- is left of both pieces, but the twin paradigm on the same page writes dizhdínóo- (between); left-edge vs between tracks ji vs zh more than a new cut of di.

1. The question

T98b (from distr-future-d): on Y&M pp. 72 and 541, 3rd-person futures show one visible di (e.g. ʼáhodidoo-, náhodidoo-) but 3a still flips (ʼáhodizhdoo-, náhodizhdoo-). Is the second piece Hale’s future d absorbed into doo? T98c: do ji-initial two-di futures (p. 329 jididinóo-) need a separate left-edge rule, or a different segmentation of di? T98a (why p. 454 distributive drops a piece) stays OPEN: print p. 454 is not available this run.

2. Sources opened this run

  • SOURCED Hale, Ken. 2001. Navajo verb stem position and the bipartite structure of the Navajo conjunct sector. Linguistic Inquiry 32: 678–693. Box text [local working copy, not in the repository] (PDF on disk unused for quotes). Focal: (24a) shidoołhosh and the definition that the future is the progressive complex preceded by inceptive Qualifier d (print p. 691 in the OCR header); derivation of nuclear doo from Qualifier d + Mode/Subj w (print pp. 687–688, exx. (17)–(19)); footnote on the future auxiliary doo as receptor alone historically (print p. 682 n. 3). Prior site cites of “pp. 626–627” for this passage are OCR/page-label slips; the LI article’s own pagination is 678–693.
  • SOURCED Young, Robert W. & William Morgan Sr. 1987. The Navajo Language, revised edition. Prefix paradigm tables via the CU Boulder Y&M interface, not print. Focal for T98b: p. 72 ʼáhodii-((1,2)) (55.html), p. 541 náhodii- (285.html), p. 454 hodii-((2)) (252.html), p. 548 naʼáhodii- (290.html), short control p. 451 hodi- (251.html). Focal for T98c: p. 329 dinii-((1)) (200.html) and dinii-((2)) (201.html); also p. 323 dini-((3)) (195.html), p. 17 ʼádiʼdi- (12.html), p. 315 didi-((3)) (192.html), p. 383 hadíní- (217.html). File 329.html is Y&M p. 617, not page 329.
  • SOURCED (as site context) distr-future-d (DF-1–DF-8; T98a–c); qualifier-d-order (QD-5); light-determiner testability note [local working copy, not in the repository].

Not opened: Y&M print (including p. 454); Kari 1976 (not on disk). No invented judgments.

3. Structure assumed

[ Qual [ M/S [vP dP_subj [ Voice [VP … V ]]]]]   + adjuncts (da, ná, 'á, …)
 nuclear:    Qual  M/S  Voice  V          (Hale p. 681; SD-1)
 future:     inceptive Qualifier d + progressive complex (Hale p. 691) — SOURCED
 surface doo: Qualifier d + Mode/Subj w (+ Voice), epenthesis and rounding → oo (Hale pp. 687–688)
 listed thematic di of long hodii-: a separate pronounced piece when present (surface; category OPEN)
 deictic subject d: displaced onto the nuclear string (QD-5 / DF-5 placement)
status: hypothesis (frame unchanged); “d” is GRANT'S CONJECTURE

OPEN Whether every listed thematic di is Hale’s Qualifier, a lower thematic head, or a listed string is still not settled by table titles. This page separates visible substring di from abstract pieces under Hale’s future analysis.

4. Rules first

FD-1  HALE FUTURE. The future is the progressive complex “preceded by the                 SOURCED (Hale p. 691)
      inceptive qualifier [d]”. (24a) shidoołhosh: Object–Qualifier–3rd:Perf–Voice–V.
FD-2  DOO ABSORBS QUALIFIER d. Nuclear surface doo is Qualifier d + Mode/Subj w          SOURCED (Hale pp. 687–688
      (glide onset, epenthesis, rounding → oo, glide drops). So Hale’s future d is         (17)–(19))
      the onset of doo, not a second visible di substring.
FD-3  P.72 / P.541 VISIBLE COUNT. 3rd futures 'áhodidoo- / náhodidoo- show one             SOURCED (55.html,
      visible di; 3a futures 'áhodizhdoo- / náhodizhdoo- place zh between that di           285.html; rechecked)
      and doo (…di-zh-doo-). Distributive rows flip the same way.
FD-4  P.454 DIYOO TWIN. 3rd hodidiyoo- / 3a hodizhdiyoo- show two visible di-pieces        SOURCED (252.html;
      with zh between in 3a — the clearer surface twin of the same abstract split.         DF-2)
FD-5  ABSTRACT PIECE COUNT (refines DF-5). Under FD-1/FD-2, long hodii- futures have       HYPOTHESIS (refines
      two abstract pieces when a listed thematic di co-occurs with future Mode whose        DF-5 / QD-5; fits
      onset is Hale’s Qualifier d (written doo or diyoo). QD-5 placement tracks that        FD-3, FD-4, QD-4)
      abstract count: between the pieces when two are present. Visible substring
      count of di undercounts on pp. 72/541.
FD-6  SHORT hodi- STILL ONE PIECE. p. 451 hodidoo- / hozhdidoo- keep zh before di.         HYPOTHESIS (compatible
      Taking the listed di as the sole Qualifier-side piece (Mode written doo without       with QD-4; category of
      a second abstract Qualifier onset) recovers the single-piece clause of DF-5.          listed di OPEN)
FD-7  JI LEFT OF BOTH (T98c fact). p. 329 dinii-((1)) 3a future jididinóo- places           SOURCED (200.html)
      ji before both di-pieces (ji-di-dinóo-), matching 3rd didínóo-’s two pieces.
FD-8  ZH BETWEEN ON THE TWIN PARADIGM. p. 329 dinii-((2)) 3a future dizhdínóo-              SOURCED (201.html;
      places zh between (di-zh-dínóo-); p. 323 dini-((3)) future dizhdínóo- likewise,        195.html)
      though its imperfective is ji-initial (jidini-). Same-page controls p. 17
      'ádizhdi'doo-, p. 315 dizhdidoo-, p. 383 hadizhdínóo- are between-di with zh.
FD-9  JI vs ZH PLACEMENT (T98c). When the 4th-person exponent is syllabic ji, it can        HYPOTHESIS (fits FD-7,
      sit at the left edge of the whole di-complex; when it is consonantal zh, QD-5         FD-8; why ((1)) keeps
      between-di placement applies if two abstract pieces are present. Not a new            ji in future OPEN)
      segmentation of di. Why ((1)) keeps ji while ((2)) takes zh-between is OPEN.
FD-10 DISTR LEFT-OF-COMPLEX ON P.329. Distributive 3a dazhdidínóo- on both p. 329           SOURCED as form;
      paradigms has zh before both di-pieces (da-zh-di-dínóo-), not between. Whether         analysis OPEN (T99b)
      that is DF-5’s single-complex left edge, a distributive rule, or something else
      is not settled here.

FD-1–FD-4, FD-7, FD-8 and the form half of FD-10 restate opened sources. FD-5, FD-6 and FD-9 are proposals. FD-10’s analysis is OPEN.

5. Data

Cells are copied as the CU interface shows them. Leipzig segmentations and “abstract piece” counts are ours (HYPOTHESIS). Visible di-piece = substring matching di/dí/dii/diy as the start of a prefix chunk. Abstract piece (under FD-5) = listed thematic di plus Hale’s Qualifier onset of future Mode when both are present. 4.SBJ = Y&M’s 3a; FUT = future column.

F1. Hale 2001 p. 691 (24a), future = Qualifier d + progressive Mode (box text; OCR damages some characters; category line is clear):
schematic: shidoołhosh — Object sh, Qualifier d, 3rd:Perf w, Voice, stem.
Hale’s prose: “The future is rendered by this same complex of elements preceded by the inceptive qualifier [d].”

F2. Hale 2001 pp. 687–688 (17)–(19): nuclear doo from Qualifier d + Mode/Subj w (+ Voice). Epenthetic vowels take rounding from w; the glide drops; the long vowel is written oo. So Qualifier d is the onset of surface doo. (p. 682 n. 3 also notes a future auxiliary doo as receptor alone historically; the (17)–(19) derivation is the nuclear analysis used here.)

F3. Y&M 1987 p. 72 ʼáhodii-((1,2)), FUTURE (saved 55.html), rechecked:
3 ʼáhodidoo-; 3a ʼáhodizhdoo-
distributive 3 ʼadahodidoo-; distributive 3a ʼádahodizhdoo-
visible di in 3rd: one; 3a: ʼá-ho-di-zh-doo- (zh between thematic di and doo).

F4. Y&M 1987 p. 541 náhodii-, FUTURE (saved 285.html), rechecked:
3 náhodidoo-; 3a náhodizhdoo-
distributive 3 ńdahodidoo-; distributive 3a ńdahodizhdoo-
same visible/flip pattern as F3.

F5. Y&M 1987 p. 454 hodii-((2)), FUTURE (saved 252.html), diyoo twin (from distr-future-d D1, rechecked):
3 hodidiyoo- (two visible di); 3a hodizhdiyoo- (two; zh between)
distributive 3 dahodidiyoo- (two); distributive 3a dahozhdiyoo- (one visible; zh before) — drop still OPEN (T98a).
p. 548 singular: 3 naʼáhodidiyoo- / 3a naʼáhodizhdiyoo- (two visible; zh between).

F6. Short control p. 451 hodi- (saved 251.html), rechecked:
3 hodidoo-; 3a hozhdidoo- (zh before di; one visible di).
No flip between di and doo.

F7. T98c focal — Y&M 1987 p. 329 (saved 200.html, 201.html), rechecked:
dinii-((1)): 3 FUT didínóo-; 3a FUT jididinóo- (ji before both); distr 3a dazhdidínóo- (zh before both).
dinii-((2)): 3a FUT dizhdínóo- (zh between); distr 3a dazhdidínóo- (zh before both). Imperfective 3a on both paradigms is jidinii-.

F8. Same-class and between-di controls, rechecked:
p. 323 dini-((3)): imperfective 3a jidini-; future 3a dizhdínóo- (between) — ji-left in imperfective, zh-between in future.
p. 17 ʼádizhdiʼdoo-; p. 315 dizhdidoo-; p. 383 hadizhdínóo- — zh between two di-looking pieces.

Table (Y&M page)3 FUT (visible di)3a FUTzh/ji vs di
ʼáhodii- (72)ʼáhodidoo- (1)ʼáhodizhdoo-between di…doo
náhodii- (541)náhodidoo- (1)náhodizhdoo-between di…doo
hodii-((2)) sg (454)hodidiyoo- (2)hodizhdiyoo-between
hodi- (451)hodidoo- (1)hozhdidoo-before di
dinii-((1)) (329)didínóo- (2)jididinóo-ji before both
dinii-((2)) (329)(3o yididínóo-)dizhdínóo-zh between
dini-((3)) (323)didínóo- (2)dizhdínóo-zh between

SOURCED Among the rechecked two-di future 3a cells above, jididinóo- is the clear ji-before-both case; the twin paradigm and p. 323 take zh-between. Broader auto-classification of all 410 tables is not trusted here without hand-checking (many …zhdidoo- strings are single-di left-edge, not two-di).

6. Derivations

DERIVATION I: 'áhodizhdoo- (p. 72, 3a singular future) — one visible di; two abstract pieces
1  Merge   [VP … V]; Voice; dP_subj = [dP ji [DP pro]] in Spec,vP                     B
2  Merge   M/S [FUT]; Agree values φ = 4th on M/S                                      B
3  Merge   listed thematic di; future Mode as Hale’s Qualifier d + progressive w       B
           → surface doo with Qualifier as onset (FD-1, FD-2)
           abstract pieces: thematic di | Qualifier-onset-of-doo
   ---- B → A ----
4  Spell   place 4th d between the two abstract pieces (FD-5 / QD-5)                   A
5  Procl.  'á-ho-                                                                      A
6  DO-6    ji → zh before consonant: 'á-ho-di-zh-doo-
OUTPUT     'áhodizhdoo- ✓ (Y&M p. 72)

DERIVATION II: 'áhodidoo- (p. 72, 3rd singular future) — same abstract pieces; no 4th
1–3        as I without 4th-person dP                                                  B
   ---- B → A ----
4  Spell   no 4th exponent; thematic di + doo abut                                     A
5  Procl.  'á-ho-
OUTPUT     'áhodidoo- ✓ — one visible di, two abstract pieces under FD-5

DERIVATION III: hodizhdiyoo- (p. 454) — both pieces visible as di
Same as I, with Mode written diyoo (Qualifier onset looks like di).                    B/A
OUTPUT     hodizhdiyoo- ✓ — DF-5’s clearer twin (FD-4)

DERIVATION IV: jididinóo- (p. 329 dinii-((1)), 3a) — ji left of both
1–3        two di-pieces in 3rd twin didínóo-; 4th exponent remains syllabic ji         B
   ---- B → A ----
4  Spell   place ji at left edge of the whole di-complex (FD-9), not between            A
OUTPUT     jididinóo- ✓ (Y&M p. 329 ((1)))

DERIVATION V: dizhdínóo- (p. 329 dinii-((2)) / p. 323) — zh between
1–3        as IV, but 4th exponent is zh (consonantal)                                  B
   ---- B → A ----
4  Spell   QD-5 / FD-5 between-di placement                                             A
OUTPUT     dizhdínóo- ✓ — same abstract di pieces as IV; different exponent shape

DERIVATION VI: stationary-d vs moving-d
I-if (stationary d): FD-5 / FD-9 are I-interface placement rules for the exponent       HYPOTHESIS
     of stationary d (Grant’s conjecture); d does not move as a head.
II   (option (c)): the same placement rules pronounce a displaced d.
Both yield the observed cells. The tables do not choose (testability note §1–2).

HYPOTHESIS Derivations I–II treat the pp. 72/541 “one visible di” puzzle as a spelling of Hale’s Qualifier inside doo, so DF-5 need not abandon between-di placement for those cells.

7. Answers

  • SOURCED T98b (Hale half). Hale’s future Qualifier d is the onset of nuclear doo (FD-1, FD-2). It is not expected to appear as a second substring di next to thematic di.
  • SOURCED T98b (table half). Rechecked pp. 72 and 541: 3rd ʼáhodidoo- / náhodidoo- (one visible di); 3a ʼáhodizhdoo- / náhodizhdoo- (zh between di and doo). p. 454 diyoo shows both pieces as di.
  • HYPOTHESIS T98b (why flip with one visible di). FD-5: abstract piece count = thematic di + Qualifier-onset-of-doo/diyoo. Between-di (QD-5) still applies on pp. 72/541. Visible substring count undercounts.
  • OPEN T98a. Why p. 454 distributive 3a lacks the second di is unanswered; print p. 454 was not opened.
  • SOURCED T98c (inventory). jididinóo- (p. 329 ((1))) is left of both di-pieces; twin ((2)) and p. 323 write dizhdínóo- (between); classic between-di controls (pp. 17, 315, 383) match zh-between.
  • HYPOTHESIS T98c (why). FD-9: syllabic ji can left-align to the di-complex; consonantal zh takes between-di when two pieces are present — not a different cut of di. Why ((1)) keeps ji in future while ((2))/p. 323 take zh-between stays OPEN (T99a).

8. Word and clause; stationary d

HYPOTHESIS Focus 1 (A vs B). Absorbed Qualifier-in-doo, abstract piece count, and ji/zh placement are ruleset A (word spell-out). Nothing here shows where the DP complement of subject d sits in the clause.

GRANT'S CONJECTURE Focus 2 (stationary d). Grant’s 2026-11-05 correction keeps d in place as the head while the DP moves. Under that conjecture, FD-5/FD-9 are I-interface placement rules for the exponent of stationary d. The same strings are what refined option-(c) placement predicts under moving-d. The tables still do not choose. Older wording that only moving-d remains is too strong (testability note).

Historical note: the “roles, not words” idea is dropped and was not used here.

9. Limits

  • All Y&M cells are from the CU Boulder interface, not print. A print check could still change a string.
  • Hale’s OCR page headers in the box text are used for LI pagination (678–693); prior site “626–627” cites for the same future passage are corrected here.
  • “Abstract piece” under FD-5 is an analytical claim on top of Hale, not a Y&M label. Short hodi- (FD-6) still needs an independent story of whether listed di is Qualifier.
  • T98c’s ji-before-both inventory is hand-checked on the named pages; no new unverified aggregate over all 410 tables is claimed.
  • Kari 1976 was not on disk; T97b stays OPEN.

10. Open questions

IDQuestionWhat would settle it
T98a(still open) Why does p. 454 distributive 3a future lack the second di?Y&M print p. 454; other long hodii- + diyoo distributive 3a rows; Kari/Y&M grammar on distributive + future.
T99aWhy does p. 329 dinii-((1)) keep syllabic ji in future (jididinóo-) while ((2)) and p. 323 take zh-between (dizhdínóo-)?More ((1))/((2)) pairs; Y&M/Kari on conjugation class and 4th-person shape before d.
T99bWhy is distributive 3a on p. 329 dazhdidínóo- (zh before both) while singular ((2)) is between?Other distributive two-di futures with zh; DF-5 vs a distributive left-edge rule.
T99cIs short hodi-’s listed di Hale’s Qualifier (FD-6), or is Mode doo Qualifier-less there?Y&M grammar on Position VI/VII; Kari 1976; more short vs long pairs.
T99dDoes Hale’s progressive Mode ever surface as diyoo vs doo for independent phonological reasons that would still keep Qualifier as onset?Hale 2000; McDonough on receptor epenthesis; more diyoo/doo twins.
T99eBetween-di placement with 3i subject ʼ in two-di futures (carry-over of T97d)?Saved 3i + thematic di + future cells with hand check.
T97b(still open) Do Y&M or Kari treat thematic di of hodii- and future di as two heads?Y&M grammar; Kari 1976 (not on disk this run).

Answered here: T98b in part (Hale absorption SOURCED; table recheck SOURCED; abstract-piece refinement HYPOTHESIS); T98c in part (inventory SOURCED; ji-vs-zh placement HYPOTHESIS; why ((1)) keeps ji OPEN). Left open: T98a (no print). Touched: DF-5, QD-5, Hale future-d/doo, Grant’s stationary-d conjecture (compatible, not proved).

11. Update 2026-11-07: 4th-person placement in the future is listed per paradigm; distributives copy it (T99a, T99b)

SOURCED In plain words: Hale (2001) puts the 4th-person prefix to the left of the Qualifier by default and allows “phonologically motivated exceptions.” In the saved Young & Morgan future tables, left of both di-pieces is the less common pattern (4 paradigms). Between is the usual one (18), and after both occurs twice. Neither the ji/zh shape, the conjugation marker, the 3rd-person future string, nor the imperfective decides it. Distributive 3a futures copy their own singular’s placement in 19 of 23 paradigms. All 4 mismatches move the 4th person to the left. HYPOTHESIS Placement is a property listed with each paradigm. That listing is the exception to Hale’s default, and the higher distributive da normally inherits it. FD-9 (shape decides placement) is withdrawn below. FD-5 (two pieces → between) is downgraded from a general rule to a description of the majority class.

11.1 Sources opened this run

  • SOURCED Hale 2001, LI 32: 678–693, §4 “The Deictic Subject Prefixes”, print pp. 689–690 (box text [local working copy, not in the repository], lines 543–600): Position V (3a, 3i, 3s) is “to the right of Object and to the left of Qualifier (with some phonologically motivated exceptions)”. Position V elements are “not invested with a person feature”. They are adjunct nominals linked to true arguments, ex. (22a) and (23).
  • SOURCED Young & Morgan 1987 via the CU Boulder interface (not print). I hand-read the full 3 / 3a / distributive 3 / distributive 3a FUTURE cells, plus the conjugation markers and imperfective 3a, for: pp. 17 (12.html), 18 (13.html), 20 (15.html), 22 (16.html), 72 (55.html), 315 didii-((1,2)) (193.html), 323 (195.html), 326 di-((3)) (199.html), 329 ((1)) and ((2)) (200.html, 201.html), 332 di-((2,5)) (203.html), 343 dii-((2,3)) (205.html), 383 ×2 (217.html, 218.html), 387 haʼdii- (222.html), 388 hadii-((1,2)) (223.html), 451 (251.html), 454 (252.html), 541 (285.html), 545 (287.html), 548 (290.html), 607 ×2 (320.html, 321.html), 609 (323.html), 611 (324.html), 748 (371.html). An initial machine pass misread column positions in 5-column tables, so every cell below was re-read from the raw table.

Not opened: Y&M print, Kari 1976. No forms below are invented. Cells are written as the interface shows them, with U+02BC used for the glottal stop.

11.2 Rules first

JZ-1  HALE DEFAULT. Position V (3a/3i/3s) is right of Object, left of Qualifier,          SOURCED (Hale 2001
      “with some phonologically motivated exceptions”.                                      pp. 689–690)
JZ-2  3a IS A LINKED NOMINAL, NOT PERSON. Position V prefixes carry no person feature;    SOURCED (Hale p. 690)
      they are adjunct nominals linked to a true argument (cf. ashkii in (23)).
JZ-3  SHAPE FOLLOWS POSITION. In every cell re-read here, the 4th-person exponent is        SOURCED (tables named in
      ji when nothing precedes it in the prefix string and zh when something does           11.1; pattern only)
      (jidinii- vs dazhdinii-, ʼazhdini-, hazhdíní-, dizhdínóo-).
JZ-4  THREE PLACEMENTS IN THE FUTURE. Relative to the di-pieces of the 3rd future:         SOURCED (count, 24 sg
      before both 4 paradigms; between 18; after both 2 (table 11.3).                       paradigms; p. 388 sg excluded)
JZ-5  NOT PREDICTED BY: (a) ji/zh shape (zh is before both in dazhdidínóo-, dazhdidoo-,    SOURCED (negative
      hozhdidoo-); (b) conjugation marker (before-both occurs with YI-Ø p. 343, NI- p. 326,  result over the
      YI- p. 332; between occurs with YI-Ø p. 72); (c) 3rd-person future string (p. 329      same tables)
      ((1)) vs ((2)) share 3o yididínóo-; p. 451 hodidoo- vs p. 72 ʼáhodidoo-);
      (d) imperfective 3a (jidinii- on both p. 329 paradigms).
JZ-6  DISTRIBUTIVE COPIES. Distributive 3a FUT has the same placement as its own           SOURCED (count)
      singular in 19 of 23 paradigms. The 4 mismatches (p. 329 ((2)), p. 387, p. 454,
      p. 545) all shift the 4th person LEFT; none shifts right.
JZ-7  LISTED EXCEPTION. Each future paradigm lists whether the 4th exponent is pronounced   HYPOTHESIS
      at Hale’s default (JZ-1: left of the whole di-complex) or at a listed site
      (between, or after both). Shape is then computed by JZ-3.
JZ-8  DISTRIBUTIVE INHERITANCE. Distributive da merges outside the conjunct complex, so      HYPOTHESIS (fits 19/23;
      the listed site of JZ-7 is computed on the same conjunct string and da does not        leftward mismatches
      change it. The leftward mismatches are not derived.                                    OPEN, T100a)

Status of older rules: HYPOTHESIS FD-9 is withdrawn. JZ-3 reverses its direction: placement determines whether the exponent is ji or zh, not the other way round, and zh before both pieces exists (JZ-5a). FD-5 is downgraded. It describes the majority (between) class. The four before-both paradigms keep the 4th person left of a thematic di plus the Qualifier onset of doo, so FD-5 cannot be a general rule. FD-6’s short control p. 451 (251.html) is reused below as the near-minimal partner of p. 72 in JZ-5c.

11.3 Data

The cells are SOURCED. The bracketed placement labels are our reading against the 3rd-person future of the same paradigm. FUT = future column; 4.SBJ = Y&M “3a”; DISTR = distributive row.

Y&M p. (paradigm)3 FUT3a FUTDISTR 3a FUTsg / distr
326 di-((3)) NI-(3o yididoo-)jididoo-dazhdidoo-before / before
329 dinii-((1)) YI-Ødidínóo-jididinóo-dazhdidínóo-before / before
332 di-((2,5)) YI-didoo-jididoo-dazhdidoo-before / before
343 dii-((2,3)) YI-Ødidoo-jididoo- (-shdidoo-)*dazhdidoo-before / before
17 ʼádiʼdi-ʼádidiʼdoo-ʼádizhdiʼdoo-ʼádidazhdiʼdoo-between / between
18 (ʼádiʼnii-)ʼádidíʼnóo-ʼádizhdíʼnóo-ʼádadizhdíʼnóo-between / between
20, 22 ʼadini-((2))ʼadidínóo-ʼadizhdínóo-ʼadadizhdínóo-between / between (both pages)
72 ʼáhodii-((1,2)) YI-Øʼáhodidoo-ʼáhodizhdoo-ʼádahodizhdoo-between / between
315 didii-((1,2))dididoo-dizhdidoo-dadizhdidoo-between / between
323 dini-((3))didínóo-dizhdínóo-dadizhdínóo-between / between
329 dinii-((2)) SI | YI-Ø(3o yididínóo-)dizhdínóo-dazhdidínóo-between / before ←
383 hadíní-; hadini-hadidínóo-hadizhdínóo-hadadizhdínóo-between / between (both)
387 haʼdii-hadiʼdoo-hadizhʼdoo-hadazhdiʼdoo-between / before ←
454 hodii-((2))hodidiyoo-hodizhdiyoo-dahozhdiyoo-between / before ← (one di drops, T98a)
541 náhodii-náhodidoo-náhodizhdoo-ńdahodizhdoo-between / between
548 naʼáhodii-naʼáhodidiyoo-naʼáhodizhdiyoo-(no row)between / —
607 ńdíʼní-ńdidíʼnóo-ńdizhdíʼnóo-ńdadizhdíʼnóo-between / between
609 ńdíní-ńdidínóo-ńdizhdínóo-ńdadizhdínóo-between / between
611 ńdíníi-ńdidínóo-ndizhdínóo-ńdadizhdínóo-between / between
748 yadidii-yadididool-yadizhdidool-yadadizhdidool-between / between
545 náʼádini-náʼádidínóo-náʼádidízhnóo-náʼádadizhdínóo-after / between ←
607 ńdíʼníi-ńdidíʼnóo-ńdidízhʼnóo-ńdadidízhʼnóo-after / after

Sg counts over 24 paradigms (p. 383 and pp. 20/22 counted per table) are 4 before, 18 between, and 2 after. The distr comparison covers 23 paradigms (p. 548 has no row): 19 match and 4 shift left (marked ←). The p. 388 hadii-((1,2)) singular 3a FUT cell reads hadazhdidoo-, which contains distributive da. It looks like an interface or copy slip and is excluded. Its distributive hadazhdidoo- is before both. p. 451 hodi- (3 hodidoo-, 3a hozhdidoo-, distr dahozhdidoo-) is FD-6’s one-piece control and is kept out of the two-piece counts. It still gives the JZ-5c contrast with p. 72.

Hale 2001 p. 689 (22a), Hale’s own segmentation and category line:

shijihozh
sh · j · Ø · ghozh
Hale: Object · 3a · Mode/Subj · Voice V · Leipzig: 1SG.OBJ-4.SBJ-MODE/SBJ-tickle
‘He/She/They/One is tickling me.’ — 4th person left of everything nuclear (JZ-1 default)

The p. 329 pair in Leipzig form. The segmentation is ours (HYPOTHESIS). TH = listed thematic di. FUT = future complex whose onset is Hale’s inceptive Qualifier d (FD-1/FD-2). The analysis of n inside -nóo- is OPEN.

ji-di-dinóo- (p. 329 ((1)))di-zh-dínóo- (p. 329 ((2)))da-zh-di-dínóo- (both)
4.SBJ-TH-FUTTH-4.SBJ-FUTDISTR-4.SBJ-TH-FUT

SOURCED The ((1)) 3a cell writes the second piece low (di), while 3rd didínóo-, ((2)) dizhdínóo- and distr dazhdidínóo- write it high (dí). This is one cell only. Whether it matters is OPEN (T100c).

11.4 Derivations (stationary d; VP-internal subject)

Frame (HYPOTHESIS; GRANT'S CONJECTURE for “d”): subject [dP d₄ DP] merges in Spec,vP (VP-internal
subject). d₄ stays put as head; DP moves to a clause position (Hale JZ-2: the 3a nominal is linked
to the argument). The 4th exponent is pronounced at a site fixed by JZ-7 (I-interface placement).

DERIVATION VII: jididinóo- (p. 329 ((1)); listed DEFAULT)
1  Merge   [VP … V]; Voice; vP with [dP d₄ DP] in Spec,vP                              syntax
2  Merge   listed TH di; Mode/Subj (progressive); inceptive Qualifier d (FUT, Hale)    syntax
3  Move    DP → clause position; d₄ stays (stationary)                                  syntax
4  Place   paradigm lists DEFAULT → exponent of d₄ left of Qualifier and TH (JZ-1)       I-interface
5  Shape   nothing precedes → ji (JZ-3)                                                 PF
OUTPUT     ji-di-dinóo- ✓ (Y&M p. 329 ((1)))

DERIVATION VIII: dizhdínóo- (p. 329 ((2)); listed BETWEEN)
1–3        as VII
4  Place   paradigm lists BETWEEN → exponent right of TH, left of FUT complex (JZ-7)    I-interface
5  Shape   TH precedes → zh (JZ-3)                                                      PF
OUTPUT     di-zh-dínóo- ✓ (Y&M p. 329 ((2)))

DERIVATION IX: dazhdidínóo- (p. 329 distributive, both paradigms)
1–3        as VII, plus distributive da merged outside the conjunct complex
4  Place   ((1)): DEFAULT inherited (JZ-8) → left of TH                                  I-interface
           ((2)): JZ-8 predicts BETWEEN (*dadizhdínóo-) — table has LEFT ✗              (T100a)
5  Shape   da precedes → zh (JZ-3)
OUTPUT     da-zh-di-dínóo- ✓ for ((1)); ((2)) needs an extra statement (leftward shift)

Moving-d (site option (c)) uses the same step-4 placement for a displaced d. It yields the same
strings. The tables do not choose between stationary-d and moving-d (testability note §1–2).

11.5 Answers and what is left

  • SOURCED T99b (mostly answered). Distributive does not impose a left-edge rule. Most distributive 3a futures stay between (pp. 17, 18, 20, 22, 72, 315, 323, 383, 541, 607, 609, 611, 748). p. 329’s dazhdidínóo- matches ((1))’s own singular. It mismatches only ((2)), and that mismatch is one of four, all leftward (JZ-6).
  • SOURCED T99a (narrowed). The shape and the obvious predictors do not explain why ((1)) is left and ((2)) is between (JZ-5). The same split recurs across paradigms that have the same visible 3rd future (p. 451 hodidoo-/hozhdidoo- vs p. 72 ʼáhodidoo-/ʼáhodizhdoo-; p. 343 didoo-/jididoo- vs p. 315 dididoo-/dizhdidoo-).
  • HYPOTHESIS T99a (why). JZ-7 treats the split as a listed exception site per paradigm against Hale’s default. Hale calls the exceptions phonologically motivated, but no opened source states the motivation. In this future sample the “exception” is the majority.
  • OPEN T99c, T99d, T99e, T98a and T97b are unchanged.
IDQuestionWhat would settle it
T100aWhy do the 4 distributive mismatches (pp. 329 ((2)), 387, 454, 545) all shift the 4th person left, never right?Y&M print for those pages; more distributive rows; whether da blocks a listed site.
T100bWhat lexical or phonological property marks before-both paradigms (pp. 326, 329 ((1)), 332, 343) against the between class?Y&M grammar on these stems; Kari 1976; Hale’s “phonologically motivated exceptions” sources.
T100cIs the low tone of the second piece in jididinóo- (vs high dí elsewhere) tied to before-both placement?Other before-both cells whose 3rd form has a high-tone dí; print check of p. 329.

12. Update 2026-11-07: the low tone in jididinóo- is one cell, not a before-both property (T100c); distributive mismatches step one site left (T100a)

SOURCED In plain words: On Young & Morgan p. 329 ((1)), the 3a future cell writes jididinóo-, with low tone on the second di-piece. That low tone does not appear in the other before-both future cells, including the 3s and the distributive 3a of the same paradigm (both high dí). The other three before-both paradigms write doo as the second piece, so they have no high/low dí contrast at all. Comparable two-di futures whose 3rd person writes high dí keep that high tone in 3a and in the distributive cells. The low tone is therefore not a mark of before-both placement. It is this one cell. Print would be needed to decide whether that cell is linguistically real or an interface slip. Separately, each of the four distributive mismatches from the midday count is a single step left (between→before, or after→between). The strings that inheritance would have produced are not generally banned: dadizhdínóo- is the actual distributive 3a future on p. 323, and p. 146 writes the …dadizhʼdoo- shape that p. 387 does not use. No rule that selects only those four without wrongly excluding the matching rows is available from the opened tables. This run does not invent a motivation Hale did not state.

12.1 Sources opened this run

  • SOURCED Young & Morgan 1987 via the CU Boulder interface (not print). I re-read the raw saved HTML and took FUTURE cells from the column headed FUTURE (six-column tables: 5th cell; seven-column tables with two PERFECTIVE columns: 6th cell). Before-both paradigms: p. 326 di-((3)) (199.html), p. 329 ((1)) dinii-((1)) (200.html), p. 332 di-((2,5)) (203.html), p. 343 dii-((2,3)) (205.html). Comparables whose 3rd future writes high dí: p. 329 ((2)) dinii-((2)) (201.html), p. 323 dini-((3)) (195.html), p. 18 ʼádiʼnii- (13.html), pp. 20 and 22 ʼadini-((2)) (15.html, 16.html), p. 383 hadíní- and hadini- (217.html, 218.html), p. 545 náʼádini- (287.html), p. 607 ńdíʼní- and ńdíʼníi- (320.html, 321.html), p. 609 ńdíní- (323.html), p. 611 ńdíníi- (324.html). T100a mismatch rows re-read: p. 329 ((2)) (201.html), p. 387 haʼdii- (222.html), p. 454 hodii-((2)) (252.html), p. 545 (287.html). Extra parallel opened for the inherited p. 387 shape: p. 146 báʼdi- (101.html). File 329.html is Y&M p. 617, not page 329.

Not opened: Y&M print, Kari 1976 (not on disk). Hale 2001 was not re-opened this run; section 11’s quotes do not address this tone fact or the four shifts. Cells are written as the interface shows them, with U+02BC for the glottal stop. No forms below are invented.

12.2 Rules first

TN-1  ONE-CELL LOW TONE. In the saved tables, low tone on the second di-piece of a              SOURCED (cells in
      3 / 3a / distr 3 / distr 3a FUTURE form that otherwise writes dínóo / dinóo               12.1; search of
      occurs only in p. 329 ((1)) 3a jididinóo-.                                                410 FUTURE columns)
TN-2  NOT TIED TO BEFORE-BOTH. Before-both placement does not predict that low tone:            SOURCED (negative)
      (a) pp. 326, 332, 343 write doo as the second piece, with no high/low dí contrast;
      (b) within p. 329 ((1)), 3 didínóo-, 3o yididínóo-, 3s hodidínóo-, distr 3 dadidínóo-
      and distr 3a dazhdidínóo- all write high dí; 3s is also before both pieces.
TN-3  HIGH DÍ IS OTHERWISE STABLE. Where the 3rd future of a two-di paradigm writes high        SOURCED (sample
      dí (didínóo-, ʼádidíʼnóo-, hadidínóo-, náʼádidínóo-, ńdidínóo-, ńdidíʼnóo-),               named in 12.1)
      the re-read 3a, distr 3 and distr 3a cells keep high dí on that piece.
JZ-9  LEFT-SHIFT IS ONE STEP. When distributive 3a FUT placement differs from its own           SOURCED
      singular (the 4 mismatches of JZ-6), the difference is one site leftward: between         (description of
      → before (pp. 329 ((2)), 387, 454) or after → between (p. 545). None shifts two           the 4; not an
      sites (after → before) and none shifts right.                                             explanation)
JZ-10 NO GENERAL STRING BAN. The inherited BETWEEN string *dadizhdínóo- for p. 329 ((2))        SOURCED (negative)
      is the actual distr 3a FUT of p. 323 (and the same shape occurs as ʼadadizhdínóo-
      pp. 20/22, hadadizhdínóo- p. 383). The inherited BETWEEN shape *hadadizhʼdoo- for
      p. 387 is the actual distr 3a FUT of p. 146 (bádadizhʼdoo- next to sg bádizhʼdoo-).
      A ban on those strings would overgenerate against matching rows.

Status of older rules: JZ-1–JZ-8 stand. JZ-8 still describes the 19 matching distributives and still does not derive the four leftward mismatches. JZ-9/JZ-10 narrow what those mismatches are and are not; they do not close T100a. FD-9 remains withdrawn; FD-5 remains a majority-class description.

12.3 Data

Tones in brackets are our reading of each visible di/dí piece (L = low di, H = high dí). The second piece of a doo future is written doo, not a separate dí. Bracketed placement labels are from section 11. Where a table has no plain 3 row, the 3o cell is given in parentheses.

Y&M p. (paradigm)class3 FUT3a FUTDISTR 3 FUTDISTR 3a FUT
326 di-((3))before(3o yididoo-) [L + doo]jididoo- [L + doo](3o deididoo-)dazhdidoo- [L + doo]
329 dinii-((1))beforedidínóo- [L + H]jididinóo- [L + L]dadidínóo- [L + H]dazhdidínóo- [L + H]
332 di-((2,5))beforedidoo- [L + doo]jididoo- [L + doo]dadidoo- [L + doo]dazhdidoo- [L + doo]
343 dii-((2,3))beforedidoo- [L + doo]jididoo- (-shdidoo-)* [L + doo]dadidii- [L + dii]dazhdidoo- [L + doo]
329 dinii-((2))between(3o yididínóo-) [L + H]dizhdínóo- [L + H](3o deididínóo-)dazhdidínóo- [L + H]
323 dini-((3))betweendidínóo- [L + H]dizhdínóo- [L + H]dadidínóo- [L + H]dadizhdínóo- [L + H]

SOURCED Same-paradigm controls on p. 329 ((1)): 3s future is hodidínóo- (also before both pieces; second piece high). First person futures in that table likewise write high dí (didínéesh-, didíníí-, didíníi-, didínóoh-). The expected high-tone 3a string jididínóo- does not occur in any saved table; only jididinóo- does. p. 343 distributive 3 future writes dadidii-, the same string as that table’s distributive 1 future cell; it is recorded as shown and is not used as evidence below.

Other two-di futures in the midday 24-paradigm sample that write dí / dínóo / dinóo, 3 / 3a / distr 3 / distr 3a FUTURE (all high dí on the piece that 3rd writes high):

Y&M p. (paradigm)3 FUT3a FUTDISTR 3 FUTDISTR 3a FUT
18 ʼádiʼnii-ʼádidíʼnóo-ʼádizhdíʼnóo-ʼádadidíʼnóo-ʼádadizhdíʼnóo-
20, 22 ʼadini-((2))ʼadidínóo-ʼadizhdínóo-ʼadadidínóo-ʼadadizhdínóo-
383 hadíní-; hadini-hadidínóo-hadizhdínóo-hadadidínóo-hadadizhdínóo-
545 náʼádini-náʼádidínóo-náʼádidízhnóo-náʼádadidínóo-náʼádadizhdínóo-
607 ńdíʼní-ńdidíʼnóo-ńdizhdíʼnóo-ńdadidíʼnóo-ńdadizhdíʼnóo-
607 ńdíʼníi-ńdidíʼnóo-ńdidízhʼnóo-ńdadidíʼnóo-interface row broken (see below)
609 ńdíní-ńdidínóo-ńdizhdínóo-(no plain 3 row)ńdadizhdínóo-
611 ńdíníi-ńdidínóo-ndizhdínóo-ńdadidínóo-ńdadizhdínóo-

SOURCED A search of every FUTURE cell in the 410 saved tables for a 3 / 3a / distr 3 / distr 3a form containing low-tone dinóo (or the substrings didin / jididin), followed by a re-read of each hit, leaves only p. 329 ((1)) jididinóo-. Two 2nd-person future cells on p. 22 (ʼadidiníí-, ʼndidinóoh-) also contain a low di before n; they are not 3/3a cells and are different morphology.

The p. 607 ńdíʼníi- (321.html) distributive 3a row is irregular in the interface: the fourth data cell concatenates ńdadízhʼníi-ńdadidízhʼnóo-, the FUTURE cell shows ńdadízhʼnóo (no extra di, no final hyphen), and the last cell is empty. The midday section treated the glued ńdadidízhʼnóo- as an after/after match. This run does not use that row as a clean control.

The p. 329 ((1)) tone contrast in Leipzig form. Segmentation is ours (HYPOTHESIS). TH = listed thematic di. FUT = future complex whose onset is Hale’s inceptive Qualifier d (FD-1/FD-2). The analysis of n inside -nóo- is still OPEN.

di-dínóo- (3)ji-di-dinóo- (3a)ho-di-dínóo- (3s)da-zh-di-dínóo- (distr 3a)
TH-FUT4.SBJ-TH-FUT3s-TH-FUTDISTR-4.SBJ-TH-FUT
second piece Hsecond piece Lsecond piece H (also before both)second piece H

T100a. Confirmed mismatch strings, the form JZ-8 would inherit from the singular, and the actual distributive:

Y&M p.sg 3a FUTinherited if JZ-8actual distr 3a FUTshift
329 ((2))dizhdínóo- (between)*dadizhdínóo-dazhdidínóo- (before)between → before
387 haʼdii-hadizhʼdoo- (between)*hadadizhʼdoo-hadazhdiʼdoo- (before)between → before
454 hodii-((2))hodizhdiyoo- (between)*dahodizhdiyoo-dahozhdiyoo- (before; one di drops)between → before + T98a
545 náʼádini-náʼádidízhnóo- (after)*náʼádadidízhnóo-náʼádadizhdínóo- (between)after → between

SOURCED p. 545 is the same left-shift-by-one as the other three (JZ-9): it lands at between rather than at the left edge. *dadizhdínóo- is attested as p. 323 distributive 3a future (also ʼadadizhdínóo- pp. 20/22; hadadizhdínóo- p. 383). *hadadizhʼdoo- itself is not in the saved tables, but p. 146 báʼdi- writes singular bádizhʼdoo- and distributive bádadizhʼdoo- (between/between) — the same shape. *dahodizhdiyoo- and *náʼádadidízhnóo- were not found. p. 454 distributive 3 (not 3a) future is dahodidiyoo-, which keeps both di-pieces, so the drop on the 3a row is not a general distributive reduction (T98a still OPEN). p. 329 ((2)) distributive 3a future dazhdidínóo- is letter-for-letter identical to the ((1)) distributive 3a cell on the same page; print would be needed to decide whether that is independent or a copy.

A rule “da plus a listed BETWEEN site is illegal” is false (p. 323, p. 146, and the matching between/between majority). A rule “da plus a listed AFTER site is illegal” cannot be tested cleanly from the midday sample: the only other after-class singular (p. 607 ńdíʼníi-) has the broken distributive row above. Nothing in the opened tables picks out exactly these four mismatches without also wrongly excluding matching rows. Hale’s text, as cited in section 11, does not state a reason for the shifts.

12.4 Derivations (stationary d; VP-internal subject)

Frame as in Derivation VII–IX (section 11): subject [dP d₄ DP] merges in Spec,vP. d₄ stays
put as head; DP moves to a clause position (JZ-2). The 4th exponent is pronounced at the
site fixed by JZ-7. Tone is a separate PF fact.

DERIVATION X: tone on p. 329 ((1)) (listed DEFAULT; JZ-7 / Derivation VII)
1–5      as VII → order ji-di-dínóo-                                                    syntax + I-interface + PF
6  Tone  3 / 3s / distr 3 / distr 3a spell the second piece high (dí)                    PF (TN-3)
         3a cell writes the second piece low (di)                                        SOURCED, one cell (TN-1)
OUTPUT   jididinóo- ✓ as written; the drop is not predicted by before-both placement
         (TN-2). 3s hodidínóo- is the same placement with high dí.

DERIVATION XI: p. 329 ((2)) distributive (JZ-8 vs the table)
1–3      as VII, plus distributive da merged outside the conjunct complex
4  Place JZ-8 inherits BETWEEN → *da-di-zh-dínóo- (*dadizhdínóo-)                        I-interface
         Table: LEFT → da-zh-di-dínóo- (dazhdidínóo-)                                    SOURCED mismatch
5  Shape da precedes → zh (JZ-3) in either case                                          PF
NOTE     *dadizhdínóo- is the actual output of p. 323 (JZ-10), so the ((2)) shift is
         not a repair of an illegal string. p. 545 is the same one-step-left operation
         landing at between (JZ-9). p. 454 adds a dropped di (T98a).

Moving-d yields the same strings and the same unmet tone fact. The tables still do not
choose between stationary-d and moving-d.

12.5 Answers and what is left

  • SOURCED T100c (answered: no / only this one cell). The low tone of the second piece in jididinóo- is not tied to before-both placement (TN-2). It does not recur in the other before-both paradigms, or in the 3s and distributive cells of the same paradigm, or in the other two-di dí futures of the sample (TN-1, TN-3). Whether that one cell is real or an interface inconsistency is a print question (T101a), not a reason to treat the low tone as a class property.
  • SOURCED T100a (narrowed, still OPEN). All four mismatches are one step left (JZ-9), and p. 545 is the same operation as the other three. The inherited strings are not generally banned (JZ-10). No tight rule that fits only these four without overgenerating against the matching rows is available. T98a (the extra di-drop on p. 454) stays OPEN and may mean that row is not the same phenomenon.
  • OPEN T100b is unchanged: marking the before-both class needs Y&M grammar or Kari 1976, which are not on disk. T99c–e, T98a and T97b are unchanged. 3i futures diʼdoo- on pp. 332 and 343 were seen while reading the before-both tables and are not pursued here.
IDQuestionWhat would settle it
T100aWhy do the 4 distributive mismatches all shift one site left, never right? What property selects just these four?Y&M print for those pages; a larger distributive sample; a stem-level property that does not also exclude p. 323 / p. 146.
T100bWhat lexical or phonological property marks before-both paradigms (pp. 326, 329 ((1)), 332, 343) against the between class?Y&M grammar on these stems; Kari 1976; Hale’s “phonologically motivated exceptions” sources. Not on disk this run.
T101aDoes print p. 329 write jididinóo- (low) or jididínóo- (high) in the ((1)) 3a future cell?Y&M 1987 print p. 329 ((1)).
T101bIs the p. 329 ((2)) distributive 3a future dazhdidínóo- independent, or a copy of ((1))’s distributive cell?Y&M print p. 329 both paradigms.
T101cWhat is the intended distributive 3a future of p. 607 ńdíʼníi-, given the broken interface row?Y&M print p. 607; a clean interface export.

13. Update 2026-11-08: the broken p. 607 row reads ńdadidízhʼnóo-, so an after-class distributive keeps its singular’s site (T101c; T100a narrowed); what the p. 329 files themselves show (T101a, T101b)

SOURCED In plain words: the broken distributive 4th-person row of Young & Morgan p. 607 ńdíʼníi- is missing one cell boundary in the interface. Restore that boundary and every cell falls into its own column. The distributive 4th-person future is then ńdadidízhʼnóo-, with zh after both di-pieces, exactly where its singular ńdidízhʼnóo- has it. So the two paradigms whose singular puts the 4th person after both pieces behave differently in the distributive: p. 607 keeps the site and p. 545 moves one step left. The after class does not force the shift, and on p. 545 the shift happens only in the future cell. For p. 329 the saved file itself carries the low tone of jididinóo- as a plain letter i, so it is not a display bug. The two p. 329 tables differ in 17 of 72 shared cells, so they are not a wholesale copy. Their future columns agree everywhere except the singular 4th person. That means the identical distributive cell is what the column does anyway and can’t be used to show copying. Print is still the last word on all three cells. HYPOTHESIS In derivational terms, the distributive da normally leaves the listed spell-out site of the 4th-person head alone. The p. 545 shift is a future-only exception. It isn’t a property of a placement class.

13.1 Sources opened this run

  • SOURCED Young & Morgan 1987 via the CU Boulder interface (not print), raw saved HTML read line by line and by character code. T101c: p. 607 ńdíʼníi- (321.html, distributive 3a row = source line 36), with same-page twin p. 607 ńdíʼní- (320.html) and parallels p. 609 ńdíní- (323.html) and p. 611 ńdíníi- (324.html). T100a comparison: p. 545 náʼádini- (287.html), all columns re-read. T101a/T101b: p. 329 dinii-((1)) (200.html) and dinii-((2)) (201.html), every cell. File 329.html is Y&M p. 617, not page 329.

Not opened: Y&M print, Kari 1976 (not on disk). Hale 2001 was not re-opened. The derivational frame is the one quoted in section 11 (JZ-1, JZ-2). The interface writes the glottal stop as ASCII '. It is given here as U+02BC. No forms below are invented. Cell readings that depend on the boundary repair are marked.

13.2 Rules first

IF-1  ROW SHAPE. Every row of 321.html has a label plus six data cells under IMPF, ITER,       SOURCED (raw HTML)
      PERF (YI Ø-Ł), PERF (YI D-L), FUT, OPT. The distr 3a row also has six, but cell 4
      holds two hyphen-final forms written together (ńdadízhʼníi-ńdadidízhʼnóo-) and
      cell 6 is empty. No other row of the table has an empty final cell.
IF-2  COLUMN SIGNATURES (321.html). Every FUT cell has two di-pieces (ńdidí…, ńdadidí…)       SOURCED (all rows)
      and ends in a hyphen. Every OPT cell has one piece (ńdí…, ńdadí…), and six of the
      ten OPT cells drop the final hyphen (ńdíʼnóosh, ńdíʼnóoh, ńdadíʼnóo, ńdadíʼnóoh,
      ńdadíʼnóo, nínáádíʼnóosh). Both PERF cells of sg 3a are the same string (ńdízhʼníi-).
IF-3  ONE-BOUNDARY REPAIR. Splitting cell 4 at its inner hyphen and moving the rest of the     SOURCED strings;
      row one column right gives PERF D-L ńdadízhʼníi- (= PERF Ø-Ł, as in sg 3a),             cell assignment is
      FUT ńdadidízhʼnóo- (two pieces, hyphen) and OPT ńdadízhʼnóo (one piece, no hyphen).      our repair (IF-1,
      Each piece then meets its column’s signature (IF-2). The unrepaired reading puts         IF-2); print unchecked
      a one-piece, hyphenless form in FUT, which no other FUT cell has.
JZ-11 AFTER CLASS DOES NOT DECIDE THE SHIFT. Of the two after-both singulars, p. 607            SOURCED (given IF-3)
      ńdíʼníi- keeps the site in the distributive (ńdidízhʼnóo- → ńdadidízhʼnóo-) and
      p. 545 shifts one step left (náʼádidízhnóo- → náʼádadizhdínóo-). Both sites
      (between and after) therefore have copying members and shifting members.
JZ-12 p. 545 SHIFT IS FUTURE-ONLY. In p. 545 the 4th exponent sits right before n in every      SOURCED (287.html,
      sg mode (náʼádizhni-, náʼádiznees-, náʼádidízhnóo-, náʼádizhnó-; written z in PERF)    all columns)
      and in every distr non-future mode (náʼádadizhni-, náʼádadiznees-, náʼádadizhnó-). Only distr FUT
      (náʼádadizhdínóo-) differs. In p. 607 ńdíʼníi-, zh sits right before ʼ in every sg
      and distr cell, the repaired distr FUT included.
IF-4  TRANSMITTED LOW TONE (T101a). In 200.html the 3a FUT cell spells the second piece with    SOURCED (character
      U+0069 (plain i). Every other second-piece dí in the FUTURE columns of 200.html and       codes)
      201.html is precomposed U+00ED. Neither file contains a combining mark
      (U+0300–U+036F). The missing acute is in the transmitted text, not in rendering.
IF-5  TWINS ARE NOT A WHOLESALE COPY (T101b). Over the 12 person rows the two p. 329 tables      SOURCED (cell-by-cell)
      share × 6 columns, 55 of 72 cells are identical and 17 differ. The FUT column is
      identical in 11 of 12 shared rows. Only sg 3a differs (jididinóo- vs dizhdínóo-).

Status of older rules: JZ-1–JZ-10 and TN-1–TN-3 stand. JZ-6’s count (19 of 23 match) is unchanged. The p. 607 ńdíʼníi- after/after match counted at midday on 2026-11-07 is now backed by IF-1–IF-3, not by reading the glued cell as is. Section 12’s worry that this row could not serve as a control is resolved for the interface (print still unchecked). JZ-8 (distributive inherits the listed site) now has a confirmed after-class member.

13.3 Data

p. 607 ńdíʼníi- (321.html), 3a and distr 3a rows exactly as transmitted (cell boundaries shown as |), then repaired:

rowIMPFITERPERF Ø-ŁPERF D-LFUTOPT
sg 3ańdízhʼníi-nínádízhʼníi-ńdízhʼníi-ńdízhʼníi-ńdidízhʼnóo-ńdízhʼnóo-
distr 3a as transmittedńdadízhʼníi-nínádadízhʼníi-ńdadízhʼníi-ńdadízhʼníi-ńdadidízhʼnóo-ńdadízhʼnóo(empty)
distr 3a repaired (IF-3)ńdadízhʼníi-nínádadízhʼníi-ńdadízhʼníi-ńdadízhʼníi-ńdadidízhʼnóo-ńdadízhʼnóo
distr 3 (control)ńdadíʼníi-nínádadíʼníi-ńdadíʼníi-ńdadíʼníi-ńdadidíʼnóo-ńdadíʼnóo

SOURCED In the repaired row every column matches the sg 3a row with da added (after ń, or after níná in ITER). The only other difference is the OPT final hyphen (IF-2). That is also the relation between distr 3 and sg 3 in the same table (ńdidíʼnóo- → ńdadidíʼnóo-). The same-page twin ńdíʼní- (320.html) has between-class sg 3a FUT ńdizhdíʼnóo- and distr ńdadizhdíʼnóo-, so the two p. 607 twins differ in the singular and their distributives differ to match. The parallels on pp. 609 and 611 are between/between (ńdizhdínóo- / ńdadizhdínóo-; ndizhdínóo- / ńdadizhdínóo-). None of the four tables writes a one-piece future anywhere, which is what the unrepaired reading would need.

The two after-class paradigms side by side (sg vs distr 3a, every mode):

Y&M p.modesg 3adistr 3asite kept?
545 náʼádini- (287.html)IMPFnáʼádizhni-náʼádadizhni-yes
PERFnáʼádiznees-náʼádadiznees-yes
FUTnáʼádidízhnóo- (after)náʼádadizhdínóo- (between)no ←
OPTnáʼádizhnó-náʼádadizhnó-yes
607 ńdíʼníi- (321.html)IMPFńdízhʼníi-ńdadízhʼníi-yes
PERFńdízhʼníi-ńdadízhʼníi-yes
FUTńdidízhʼnóo- (after)ńdadidízhʼnóo- (after; IF-3)yes
OPTńdízhʼnóo-ńdadízhʼnóoyes

p. 329 cell comparison, ((1)) 200.html vs ((2)) 201.html, over the 12 person rows both tables list (((1)) also has 3, 3s, distr 3 and REV rows, which ((2)) lacks). All 17 differing cells:

columnidenticaldiffering cells (((1)) | ((2)))
IMPF12/12—
ITER11/123a názhdinii- | nízhdinii-
PERF (first)0/12header YI-Ø Ø-Ł | SI D-L; e.g. 1 dinii- | dinésh-, 3a jidinii- | jidinees-, distr 3a dazhdinii- | dazhdinees-
PERF (YI D-L)10/122du dinoo- | dinooh-; 2pl dadinoo- | dadinooh-
FUT11/12sg 3a jididinóo- | dizhdínóo-; distr 3a dazhdidínóo- in both
OPT11/12SEMELIT. ńdinoosh- | náádinoosh-

SOURCED In ((1)) the SEMELIT. optative ńdinoosh- is the same string as the REV optative in the row directly above. All other SEMELIT. cells of ((1)) begin náá-. So this interface page does contain at least one vertical copy-like slip. It is not the distributive future cell. Character codes for T101a: 3a FUT j i d i d i n ó o - = U+006A U+0069 U+0064 U+0069 U+0064 U+0069 U+006E U+00F3 U+006F U+002D. Same-table 3s FUT h o d i d í n ó o - has U+00ED in that slot.

p. 607 ńdíʼníi- future, in Leipzig form. The segmentation is ours (HYPOTHESIS). ń is the listed initial element of the paradigm, and its analysis is OPEN. P1, P2 = the two di-pieces of the 3rd future. Which one is thematic and which is Hale’s inceptive Qualifier is not settled for this paradigm by the opened sources. ʼnóo is the rest of the complex, also OPEN.

ń-di-dí-ʼnóo- (3)ń-di-dí-zh-ʼnóo- (3a)ń-da-di-dí-zh-ʼnóo- (distr 3a, IF-3)
Ń-P1-P2-RESTŃ-P1-P2-4.SBJ-RESTŃ-DISTR-P1-P2-4.SBJ-REST

13.4 Derivations (stationary d; VP-internal subject)

Frame as in VII–XI: subject [dP d₄ DP] merges in Spec,vP (VP-internal subject). d₄ stays put as
head; its DP complement moves to a clause position (JZ-2: the 3a nominal is linked to the
argument). Steps 1–3 are clause syntax (ruleset B, right-headed). At the word, ruleset A
(left-headed spell-out) fixes where the exponent of d₄ is pronounced (JZ-7) and its shape (JZ-3).

DERIVATION XII: ńdadidízhʼnóo- (p. 607 ńdíʼníi- distr 3a FUT; listed AFTER)
1  Merge   [VP … V]; Voice; vP with [dP d₄ DP] in Spec,vP                              B (syntax)
2  Merge   listed paradigm material (ń, P1, P2) + future Mode/Qualifier (Hale)          B (syntax)
3  Merge   distributive da outside the conjunct complex; Move DP → clause position;   B (syntax)
           d₄ stays (stationary)
4  Place   paradigm lists AFTER → exponent of d₄ right of P2, left of ʼ (JZ-7);         A (word)
           da does not reset it (JZ-8)
5  Shape   material precedes → zh (JZ-3)                                               A / PF
OUTPUT     ń-da-di-dí-zh-ʼnóo- ✓ (Y&M p. 607, IF-3 repaired cell)

DERIVATION XIII: náʼádadizhdínóo- (p. 545 distr 3a FUT; listed AFTER)
1–3        as XII
4  Place   JZ-8 inherits AFTER → *náʼá-da-di-dí-zh-nóo- (*náʼádadidízhnóo-)              A (word)
           table: one step left → náʼá-da-di-zh-dí-nóo-                                SOURCED mismatch
5  Shape   zh (JZ-3)
NOTE       In the same paradigm, step 4 inherits AFTER/before-n in IMPF, PERF and OPT
           (JZ-12). So the extra statement needed for p. 545 is limited to the future,
           where the second piece is present. It is not a rule about the AFTER class
           (JZ-11: p. 607 shows that class inheriting).

Moving-d (site option (c)) gives the same strings with the same step-4 statements. As before,
the tables fix order, not which piece moved, so they do not choose stationary-d over moving-d.

13.5 Answers and what is left

  • SOURCED T101c (answered for the interface). The intended distributive 3a future of p. 607 ńdíʼníi- is ńdadidízhʼnóo-. The string is in the transmitted row. It belongs in the FUT column once the one missing cell boundary is restored (IF-1–IF-3). Every other cell then lands in its own column. A print check of p. 607 remains the final test and is not claimed.
  • SOURCED T100a (narrowed, still OPEN). “da plus a listed AFTER site is illegal” is false: p. 607 inherits AFTER (JZ-11). The p. 545 shift is confined to the future cell of its paradigm (JZ-12). Together with JZ-10, no placement class (between or after) and no string ban selects the four shifts. The remaining question is lexical or future-specific per paradigm.
  • SOURCED T101a (narrowed). The low tone of jididinóo- is in the transmitted text (U+0069), not produced by display or CSS (IF-4). Whether the interface’s data entry matches print p. 329 is still a print question.
  • SOURCED T101b (narrowed). The two p. 329 tables are not a wholesale copy (IF-5: 17 differing cells, including non-predictable ones such as ITER 3a názhdinii- vs nízhdinii-). Their FUT columns agree in every shared row except sg 3a. So the identical distributive 3a future dazhdidínóo- is the column’s normal behaviour and doesn’t count as evidence of copying. One copy-like slip elsewhere on the page (SEMELIT. OPT) shows that slips do occur. The interface cannot decide this cell, and it stays a print question. The p. 607 twins give a contrast. There the singulars differ and the distributives differ to match, so the interface does not simply reuse a twin’s distributive cell.
  • OPEN T100b, T98a, T97b unchanged (need Y&M grammar, Kari 1976 or print, not on disk). T99c–e unchanged. The 3i futures diʼdoo- on pp. 332 and 343 are still not pursued (T97d/T99e).
IDQuestionWhat would settle it
T100aWhat selects the four distributive futures that shift one site left (pp. 329 ((2)), 387, 454, 545), given that neither the placement class nor a string ban does (JZ-10–JZ-12)?Y&M print for those pages; a stem-level or future-specific property that does not also cover p. 607 ńdíʼníi-, p. 323 or p. 146.
T101aDoes print p. 329 ((1)) write jididinóo- or jididínóo-? (Interface text has plain i, IF-4.)Y&M 1987 print p. 329.
T101bIs ((2)) distr 3a FUT dazhdidínóo- independent? (Not a wholesale copy; the cell itself is undecidable from the interface.)Y&M print p. 329, both paradigms.
T102aIs the p. 545 distributive future shift (náʼádadizhdínóo-) tied to the presence of the second di-piece? It is the only mode of that paradigm with two pieces, and the only one that shifts (JZ-12).Other after-site or between-site paradigms whose non-future modes can be compared with the distributive future; Y&M print p. 545.

14. Update 2026-11-08 (later): the leftward distributive shift is not four stray cells. Across every interface paradigm with a one-di 3a future it occurs in 23 of 57, and neither the imperfective, the singular string nor the number of di-pieces predicts it (T102a answered, T100a reframed)

SOURCED In plain words: sections 11–13 treated the distributive future that puts the 4th-person zh one step left of its singular’s site as four exceptions in a 23-paradigm sample. Scanning all the saved interface paradigms shows it is a second ordinary pattern. Among the 57 paradigms whose singular 3a future writes zh right after one di, 34 keep that order in the distributive and 23 put zh before the di. Nothing I could read off the tables decides which, so for now it is a paradigm-by-paradigm fact.

14.1 Sources opened this run

  • SOURCED Young & Morgan 1987 via the CU Boulder interface (not print): the 410 saved paradigm HTML files ([local working copy, not in the repository]). 312 contain a FUTURE column and exactly two 3a rows (singular and distributive). Page numbers below are the “(Young & Morgan, 1987, p. N)” line inside each file; file name in brackets. Rows read in full, all columns: p. 23 20.html, p. 28 23.html, p. 122 92.html, p. 330–334 202.html–204.html, p. 326 199.html, p. 387 222.html, p. 454 252.html, p. 545 287.html, p. 329 ((2)) 201.html. For the other rows I read only the 3a FUT and 3a IMPF cells, as printed in the tables below.

Not opened: Y&M print, Kari 1976. The interface writes the glottal stop as ASCII '; given here as U+02BC. The interface has data-entry slips (section 13: IF-5; p. 28 also has ʼadńdiń- in the 1du future), so a single odd cell is not strong evidence either way.

14.2 Rules first

SC-1  SCAN DEFINITION. For each of the 312 files take the 3a cell of the FUTURE column in        SOURCED (method)
      the first 3a row (singular) and the second 3a row (distributive). In each string find
      the tokens zh and di (string match, tone and hyphens ignored, so a "di" inside another
      morph is also counted). Compare the order of the tokens in the two strings. This is a
      string-level classification, not a parse; the 57+2 rows in the tables below are the
      ones I list, and I read their FUT and IMPF cells.
SC-2  RESULT. Singular 3a FUT with one di then zh (class dz, 59 files): distributive keeps       SOURCED (strings)
      dz in 34 (copy); puts zh before di in 23 (left shift, table A); and in 2 further
      files the singular has zh before di and the distributive has di before zh (right
      shift, table B). Singulars already zh before di (zd, 64 files) keep that order.
SC-3  NOT DECIDED BY THE IMPERFECTIVE. Where the 3a IMPF has zh before di (as in             SOURCED (strings)
      ʼazhdi-, p. 23), the distributive future copies the singular in 22 files and shifts in
      13. Where the 3a IMPF has zh with no di-token: copies in 6, shifts in 8.
SC-4  NOT DECIDED BY THE SINGULAR STRING. Identical singular 3a futures go both ways:             SOURCED (strings)
      ʼádizhdoo- copies on p. 23 (ʼádadizhdoo-, 18.html) and shifts on pp. 30 and 31
      (ʼádazhdidoo-, 24.html, 25.html); ʼadizhdoo- copies on p. 22 (ʼadadizhdoo-, 17.html)
      and shifts on p. 23 (ʼadazhdidoo-, 20.html); ńdizhdoo- copies on p. 614 (ńdadizhdoo-,
      327.html) and shifts on pp. 616 and 617 (ńdazhdidoo-, 328.html, 329.html).
SC-5  NOT TIED TO A SECOND di-PIECE (T102a). All 23 left shifts in class dz have one                 SOURCED (strings)
      di-token in the singular future. p. 545 (two pieces) is one more instance, not the
      defining one.
SC-6  STRUCTURE IS SHARED. Whichever linear order is spelled out, the Merge sites of the          HYPOTHESIS
      4th-person d and the thematic d are the same in the copying and the shifting
      paradigm (steps 1–3 below). The difference is confined to the word-internal
      spell-out (step 5), where the data show two attested orders and no rule choosing
      between them. (Not forced by data; stationary-d and moving-d give the same strings.)

14.3 Data

Table A. Left shifts: singular 3a future has di then zh; distributive 3a future has zh before di (23 files).

Y&M page (file)singular 3a FUTdistributive 3a FUT
23 (20.html)ʼadizhdoo-ʼadazhdidoo-
30 (24.html)ʼádizhdoo-ʼádazhdidoo-
31 (25.html)ʼádizhdoo-ʼádazhdidoo-
60 (48.html)ʼahidízhʼnóo-daʼahizhdíʼnóo-
117 (89.html)ʼadízhnóo-ʼadazhdínóo-
121 (91.html)ʼadízhnóo-ʼadazhdínóo-
173 (111.html)bidádizhdoo-bidádazhdidoo-
187 (120.html)bighádizhʼdoo-bighádazhdiʼdoo-
212 (138.html)bidizhʼdoo-bidazhdiʼdoo-
217 (142.html)bidízhʼnóo-bidazhdíʼnóo-
282 (178.html)chʼídizhʼdoo-chʼídazhdiʼdoo-
299 (187.html)dádizhdoo-dádazhdidoo-
330 (202.html)dizhdoo-dazhdidoo-
354 (206.html)dzidizhdoo-dzidazhdidoo-
355 (207.html)dzídzadizhdoo-dzídzadazhdidoo-
387 (222.html)hadizhʼdoo-hadazhdiʼdoo-
404 (230.html)hádizhʼdoo-hádazhdiʼdoo-
406 (232.html)hadizhʼdoo-hadazhdiʼdoo-
469 (257.html)ʼidízhʼnóo-ʼidazhdíʼnóo-
524 (274.html)nabídízhnóo-nabídazhdínóo-
616 (328.html)ńdizhdoo-ńdazhdidoo-
617 (329.html)ńdizhdoo-ńdazhdidoo-
656 (352.html)dízhnóo-dazhdínóo-

Table B. Right shifts: singular has zh before di; distributive has di before zh (2 files). In both, the distributive 3a row is odd next to the 3rd-person distributive cell (p. 28 3rd distr dadiʼdoo-; p. 122 3rd distr dadíʼnóo-), so these may be slips; they are not counted as a pattern.

Y&M page (file)singular 3a FUTdistributive 3a FUT
28 (23.html)ʼazhdidoo-dadizhʼdoo-
122 (92.html)ʼazhdínóo-dádízhʼnóo-

Copy files (34, class dz → dz), by Y&M page: 13, 14, 19, 22, 23, 72, 105, 141, 146, 149, 172, 178, 185, 213, 215, 373, 387, 409, 442 (two paradigms), 524, 541, 545 (286.html), 546 (two paradigms), 555, 558, 614, 636, 645, 673, 675, 676, 700. A page number can be shared by two paradigms; p. 387 has one copier (220.html, hadizhdoo- → hadadizhdoo-) and one shifter (222.html), and p. 545 has the copier 286.html beside the two-piece shifter 287.html of section 13.

14.4 Derivation: one structure, two spell-outs (schematic; HYPOTHESIS)

Only the strings in 14.3 are Navajo data. The steps use slot labels, not new forms. Take p. 23 20.html as the focal pair: singular 3a FUT ʼadizhdoo-, distributive 3a FUT ʼadazhdidoo-, where the same paradigm’s IMPF shows zh before di (ʼazhdi-, ʼadazhdi-).

Step 1  MERGE (clause syntax, ruleset B, right-headed): V (stem) with its arguments; the
        light head d_obj takes its DP; the 4th-person (3a) subject d is merged as a
        spine head (Syntax-bot split); the noun DPs stay outside the word.
Step 2  MERGE: Distr (da-) and the future/Mode head above the thematic piece (di); T
        above that. Agree: 4th-person d values zh; da- values distributive.
Step 3  MOVE (clause syntax): the verb complex moves through v, Mode, T; only head d’s are
        taken up into the word (stationary-d: d stays; moving-d: d adjoins; same strings).
Step 4  WORD BOUNDARY: ruleset B ends. The word is built from the heads collected in
        steps 1–3 and ruleset A (left-headed, template-like) linearizes them.
Step 5  SPELL-OUT of zh relative to the thematic di:
          copy paradigms (34 files, e.g. p. 22)   sg ʼa-di-zh-doo,   distr ʼa-da-di-zh-doo
          shift paradigms (23 files, e.g. p. 23)  sg ʼa-di-zh-doo,   distr ʼa-da-zh-di-doo
        Nothing in steps 1–4 differs between the two. SC-3 and SC-4 show the choice is
        not read off the IMPF or the singular string, so it has to be a per-paradigm
        statement in step 5 (or a different thematic item). This is a restatement, not an
        explanation.

14.5 Answers and what is left

  • SOURCED T102a (answered, no). The shift is not tied to a second di-piece (SC-5). Most shifting paradigms have one.
  • SOURCED T100a (reframed, still OPEN). The earlier four are not a rare minority: in the full interface 23 of 57 one-di paradigms shift (SC-2). JZ-8 (“distributive inherits the singular’s site”) describes the 34 copiers only. The earlier statement “never right” holds for the 23 listed here but two other files go the other way (table B, possibly slips). What selects the shifters is not the imperfective order (SC-3) and not the singular string (SC-4); I did not test stem class, ablaut, or Kari’s paradigm numbers.
  • HYPOTHESIS The section-11 to 13 wording that treats the shift as an exception to be explained away, and any text that calls the copy pattern “the rule”, is too strong. Both orders are attested; the structure (SC-6) predicts neither.
  • OPEN Unchanged: T100b, T98a, T97b, T101a/b (need print or Y&M grammar, Kari 1976, not on disk).
IDQuestionWhat would settle it
T103aWhat distinguishes the 23 shifting paradigms from the 34 copying ones, given SC-3 and SC-4? Candidates not yet tested: stem class (Y&M paradigm label), ablaut, the other distributive modes beyond IMPF, the 3o rows.Y&M grammar / Kari 1976 on stem classes; a cross-tab of the 57 paradigms against paradigm label and the 3o/3i future cells.
T103bDo print pp. 28 and 122 really write the right-shifting distributive 3a futures (dadizhʼdoo-, dádízhʼnóo-)?Y&M print pp. 28 and 122.
T103cFor the 64 singulars that already have zh before di, does any distributive move right (the scan found none in class zd→dz beyond table B)? Are there singular 3a futures with other token patterns (e.g. ʼdi… with zh inside another piece) that the string scan miscounts?Hand check of the zd and zdd/dzd classes in [local working copy, not in the repository].

2026-11-11 · English phrasal verbs and Navajo verbs

2026-11-14: the “On the spine” summary at the top of this page supersedes this section’s framing. Its data, judgments and citations are unchanged.

2026-11-11 audit pass 3. Tree fix: in the English panels A and C the object sits in Spec,PP and is not the complement of throw, but the LF gave it two theta roles (Theme of throw and the OUT role). The LF now gives it one role (OUT); being thrown follows from Cause(e,s). Panel B now says why its complex head can relate the object to both. Five figure-caption badges that carried extra words were split into a label plus plain text. In the Navajo panels the low directional AdvP is a VP adjunct, so its LF is now OUT(e) (an event modifier) instead of OUT(e,t), which had given the object a second role. See the pass 3 report.

2026-11-11 audit pass 4. The theta checker now also counts relation predicates (OUT, IN, UP, CAUSE, BECOME…), so the pass-3 “throw the rat out” error would be caught automatically. Panel B, where one complex V⁰+P⁰ head relates the object to both throwing and OUT, is marked “θ-composite: r” as a stated hypothesis. See the pass 4 report.

2026-11-11 audit pass 6. The Navajo panel A title now reads “GRANT’S CONJECTURE: our V→v→T derivation”. Before, the visible title said only “Hypothesis”, and the GRANT’S CONJECTURE label (untested, neither proven nor refuted) was in the panel summary only. See the pass 6 report.

2026-11-13 audit pass 7. Grant’s instruction (2026-11-13): VPISH is not a word-order account; it frames the verb’s arguments so that syntax and semantics line up. In ordinary tensed clauses, V→v→T and subject raising Spec,vP→Spec,TP (on to Spec,CP when needed) are the working default, not GRANT’S CONJECTURE to be tested against a no-movement rival. This review was posted before that instruction. Its derivations, labels and verdicts on verb movement are rewritten here to use the default: the former no-movement rival was first kept as labelled contrast and then, on Grant’s instruction later on 2026-11-13, removed from the figure (panel B; it carried no form not in panel A), each derivation says what else has to move, and examples that seemed to resist the default are kept as tests of it (OPEN). No Navajo form, gloss, judgment or citation changed. See the 2026-11-13 note and the pass 7 report.

SOURCED Navajo has more than one useful counterpart to English phrasal verbs. A directional preverb can belong with a stem across intervening prefixes (Hale 2001:678–679). A separate postposition can also change a verb’s meaning: Chee & Yazzie call such combinations “phrasal verbs” (2025:22, 34–35). Neither fact establishes English-style particle shift in Navajo. HYPOTHESIS The comparison should therefore ask what combines, what can separate, and what meaning or participant changes, rather than demand a single yes/no answer.

Correction · 2026-11-11. SOURCED The initial version’s “no verified postposition + verb pair” was a limit of that evidence set, not a gap in the literature: Chee & Yazzie supply explicit pairs. HYPOTHESIS The first diagrams also used head-movement arrows to stand for purely morphological assembly. That was misleading. The revised Navajo figure reserves those arrows for the raising account and shows phonological spell-out without them. The English pronoun restriction is now explicitly restricted to unstressed pronouns; Booij’s Dutch account is not lumped together with a complex-head account.

HYPOTHESIS Our only starting assumptions are VPISH and theta roles. In our implementation, the subject DP first merges in left Spec,vP; “VPISH” names that hypothesis, not a category. Every additional category, movement, case mechanism, Result head, lexical-storage rule or spell-out operation discussed below is an author’s proposal or our labeled hypothesis. A source’s framework is not adopted merely because its data are used. NONE: no light determiner is needed by the present comparison; its existence is not decided.

1. English: meaning and separability are different tests

SOURCED Jackendoff distinguishes idiomatic, directional and aspectual uses (2002:73–80). A particle’s meaning cannot be read from its spelling alone: up in lifting, abandoning, and finishing is not automatically the same semantic contribution. Aspectual particles need not mark an endpoint: continuative on and away differ from completive up (pp. 76–80). The placement contrasts below illustrate standard English facts; their wording is constructed, not quoted corpus data (compare Jackendoff p. 70).

Combination / useConstructed English exampleWhat it tests
run out · spatialThey ran out.A directional use; no object here, so no particle-shift test. “The milk ran out” has a different, exhaustion reading.
climb up · spatialThey climbed up.Upward motion. “They climbed up the ladder” includes a path phrase, not evidence that the ladder is a shifted object.
pick up · liftingShe picked up the book. / She picked the book up.Particle placement alternates around a full object.
give up · abandonShe gave up the plan. / She gave the plan up.An idiomatic meaning can still allow the alternation.
turn off · deactivateShe turned off the lamp. / She turned the lamp off.A conventional result reading; not simply physical turning.
look after · care forShe looked after the child. / *She looked the child after.A prepositional-verb control, not the same placement pattern.
eat up · complete consumptionShe ate the cake up.An aspectual/completive use, not upward motion.
use up · exhaust a resourceShe used the money up.Another completive use; compare the sourced examples below.

SOURCED The ordinary pattern with an unstressed object pronoun is illustrated by Jackendoff’s “Lila looked it up” / “*Lila looked up it” (2002:70, (4a)). Our constructed parallel is She picked it up / *She picked up it. Do not turn this into a ban on every focused/stressed pronoun in every variety. Long objects, discourse and prosody also affect placement; merely counting words is not an analysis.

HYPOTHESIS (editorial Leipzig-style gloss) The constructed example below is annotated by us, not quoted from a source. PST = past; DEF = definite. The gloss is an annotation of the constructed sentence, not an analysis of English stem allomorphy.

Shepickedthebookup
3SG.Fpick.PSTDEFbookup

HYPOTHESIS Glossing the particle as up, rather than assigning an automatic PATH or RESULT role, keeps the disputed semantic analysis out of the gloss.

2. What the English analyses disagree about

SOURCED The following are accounts in the literature, not settled facts. Direct readings and secondary reports are distinguished; no original page is claimed as read when only a later discussion was available.

AccountProposal and pressure pointEvidence actually read
Small-clause family · Kayne 1985; den Dikken 1995The object and particle participate in a predication below the verb. The family is not one tree: in den Dikken’s version the DP begins as the particle’s complement and moves into the small-clause subject position. That differs from simply drawing [DP particle] as the initial order.Ramchand & Svenonius 2002, author manuscript PDF pp. 2–4, especially (3)–(7). Kayne and den Dikken are covered through this explicit secondary discussion; their originals were not page-checked here.
Complex predicate / complex head · Johnson 1991As reported by Ramchand & Svenonius, V and particle form a complex morphological word that raises to a higher μ head. They challenge that word analysis using cross-language separation and headedness evidence.Their manuscript pp. 2, 5–6; Johnson’s original article was not directly page-checked.
Particles as P · Svenonius 2003His Figure/Ground proposal distinguishes a particle with a Figure but no Ground from a preposition with a Ground. His examples permit shifting a moving thing, not a path/location just because it follows P. This is a structural proposal, not proof that every particle has a spatial meaning.Direct: pp. 433–435, especially (10)–(20). Compare Jackendoff 2002:70 on back and together, particles that do not double as ordinary prepositions.
Result decomposition · Ramchand & Svenonius 2002The object first combines low with the particle, but that phrase is not itself treated as a complete small clause. Separate verbal and Result structure supplies the event interpretation. Particle movement to R versus overt object movement gives the two orders. Their fine-grained thematic positions and EPP/lexicalization conditions are additional hypotheses.Direct: manuscript pp. 2, 6–9, (13)–(19). Their account aims to combine insights from both main families, not simply endorse either one.
Dutch constructional idioms · Booij 2002Dutch separable verbs are word combinations with word-like functions. They can be stored and can change valency, yet syntactic separation argues against treating them as ordinary prefixed words. His solution is a constructional idiom, not Johnson’s morphological complex head.Direct: pp. 21–25, 29–31. His storage argument explicitly notes that syntactic idioms can be stored too (p. 26).
Different English constructions · Jackendoff 2002Idiomaticity supports lexical listing, not necessarily grammatical wordhood or even a constituent containing just V and particle. Directional and aspectual uses have different semantic and argument-structure properties despite overlapping placement behavior.Direct page images: pp. 68–80, especially 73, 75–77, 80.

Analysis A: Small predication: object in left Spec,PP Hypothesis

Analysis A: Small predication: object in left Spec,PPAll heads are initial. The subject DP is in left Spec,vP; the object is the left specifier of a predicative PP, not an SC/O/Prt category. PP predication and causation are our semantic hypothesis. The object gets exactly one theta role, from the small predicate (OUT); throw does not take it as its complement here, so there is no second Theme(e,r) role. That the rat is what gets thrown is an inference from Cause(e,s), not a theta role. LF: ∃e∃s[throw(e) ∧ Agent(e,a) ∧ rat(r) ∧ Cause(e,s) ∧ OUT(s,r)] Full structure in the list below the figure.vPDPagent av′v⁰VPV′V⁰throwPPDPthe dead ratP′P⁰out

Analysis B: Complex-head hypothesis Hypothesis

Analysis B: Complex-head hypothesisAll selecting heads are initial. V⁰ is a complex V⁰+P⁰ head; the object DP is its complement, not a right specifier. This panel assumes base complex-head assembly, not an unshown particle-movement derivation. The complex V⁰+P⁰ head is one predicate and assigns the object one role, written here as Theme(e,r) ∧ OUT(s,r) because the complex predicate entails both. (θ-composite: r) LF: ∃e∃s[throw(e) ∧ Agent(e,a) ∧ Theme(e,r) ∧ rat(r) ∧ Cause(e,s) ∧ OUT(s,r)] Full structure in the list below the figure.vPDPagent av′v⁰VPV′V⁰DPthe dead ratV⁰throwP⁰outcomplex V

Analysis C: Separate result layer: particle-to-P option Hypothesis

Analysis C: Separate result layer: particle-to-P optionAll selecting heads are initial and both specifiers are left. A higher P⁰ is hypothesized to express result and the particle raises to it. This P result layer is our abstraction, not an assertion that Ramchand–Svenonius call their R head P. The low object stays in left Spec,PP. The object gets exactly one theta role, from the small predicate (OUT); throw does not take it as its complement here, so there is no second Theme(e,r) role. That the rat is what gets thrown is an inference from Cause(e,s), not a theta role. LF: ∃e∃s[throw(e) ∧ Agent(e,a) ∧ rat(r) ∧ Cause(e,s) ∧ OUT(s,r)] Full structure in the list below the figure.vPDPagent av′v⁰VPV′V⁰throwPPP′P⁰[result]PPDPthe dead ratP′P⁰outresult headhead-move
(1) Our vP comparison of three particle-verb derivations Hypothesis English throw / the dead rat / out: Ramchand & Svenonius 2002 manuscript (13)–(18), pp.6–8. A–C are our category-normalized schematic hypotheses, not author facsimiles. Historical contrasts only: small-clause family, their pp.2–4 (Kayne/den Dikken); complex-head family, p.5 (Johnson); result discussion, pp.6–8. The P⁰ result abstraction in C is editorial. The constructed agent placeholder and event LFs are labeled HYPOTHESIS. The drawings do not themselves settle particle shift or pronoun placement.
Bracket notation (copy)

A: [vP [DP agent a] [v′ v⁰ [VP [V′ [V⁰ throw] [PP [DP the dead rat] [P′ [P⁰ out]]]]]]] B: [vP [DP agent a] [v′ v⁰ [VP [V′ [V⁰ [V⁰ throw] [P⁰ out]] [DP the dead rat]]]]] C: [vP [DP agent a] [v′ v⁰ [VP [V′ [V⁰ throw] [PP [P′ P⁰ [PP [DP the dead rat] [P′ [P⁰ out]]]]]]]]]

Structure as a list

Analysis A

  • vP
    • DP (id agent): agent a
    • v′
      • v⁰ (head; id v)
      • VP
        • V′
          • V⁰ (head; id V): throw
          • PP
            • DP: the dead rat
            • P′
              • P⁰ (head; id particle): out

Analysis B

  • vP
    • DP (id agent): agent a
    • v′
      • v⁰ (head; id v)
      • VP
        • V′
          • V⁰ (head; id complexV; hosts the “complex V” (complexV))
            • V⁰ (head; id verb): throw
            • P⁰ (head; id particle): out
          • DP: the dead rat

Analysis C

  • vP
    • DP (id agent): agent a
    • v′
      • v⁰ (head; id v)
      • VP
        • V′
          • V⁰ (head; id V): throw
          • PP
            • P′
              • P⁰ (head; id result; [result]; hosts the “result head” (particle + result))
              • PP
                • DP: the dead rat
                • P′
                  • P⁰ (head; id particle; head-moves to result): out

HYPOTHESIS (our redrawn vP derivations) All subject DPs start in left Spec,vP. A makes the object the left specifier of a predicative PP; B puts V⁰ and P⁰ inside one complex V⁰; C tests a separate result layer with particle-to-P movement. The P⁰ result abstraction is ours, not an assertion that an earlier author’s R is P. These are category-normalized hypotheses, not author facsimiles. Each panel has an explicit event LF; none alone explains the unstressed-pronoun restriction. OPEN: which derivation survives the full diagnostics.

3. Navajo: three different kinds of evidence

SOURCED All new forms below were checked against page images. Spellings and translations are retained; glottal stops are U+02BC. Abstract source segmentation is not a claim that the pieces concatenate without phonological changes. Earlier Speas examples NX-089–093 remain in the index: the talking dependency, the wriggle/finish-eating contrast, and the sit/remain-behind pair (1986:205–206).

3a. Direction and aspect inside the word

Checked exampleExact source translationWhat is licensed
NX-094 Siláo tłʼóóʼgóó chʼíshidiníłdązh.“The policeman jerked me outdoors (to get me out of a fight).”Hale 2001:678–679, (1)–(3): preverb chʼí, “out horizontally”; object, qualifier, mode/subject and voice intervene before the stem.
NX-095 haashááh“I climb up”McDonough 2019:28, (1): ha- is glossed “up”. Speas 1986:212 (3.16b) also lists ha “up and out”; that inventory label is not an interchangeable translation for every occurrence.
NX-096 yah ʼiishʼnééh“I crawl inside”McDonough 2019:40, (8): yah is glossed “into an enclosed space”.
NX-097 yah ʼiishteeh“carry AnO inside”McDonough 2019:39, Table 7: the source labels this transitive and the crawling example intransitive. Different bases and valence markers are involved; this is not a minimal prefix contrast.
NX-106 binánídishkid“I ask you about it.”Chee & Yazzie 2025:22, (7), n. 3: ná is REV, reverting to a prior location, state or condition. Compare ITR in their (14) below; do not silently collapse the two labels.

SOURCED (source interlinear, not ours) McDonough 2019:28 prints the following analysis for NX-095. VL is her valence label; IPFV is imperfective. These pieces are not surface syllables:

ha-ishøghááh
‘up’-IPFV.1VL-walk.IPFV

SOURCED For NX-094, Hale’s source segmentation is chʼí · sh · d · n-´ · ł · dązh, labeled Preverb · Object · Qualifier · Mode/Subj · Voice · V (stem). He describes the last four as thematic material relating to arms/legs, a portmanteau of situation/perfective aspect and third-person subject, transitivity, and jerky movement with perfective inflection (pp. 678–679). HYPOTHESIS A short editorial Leipzig gloss would be OUT.HORIZ · 1SG.OBJ · THM · SIT.PFV.3SG.SBJ · TR · JERK.PFV. This is our abbreviation of his prose, not a source-printed gloss or a claim that OUT assigns a new argument.

OPEN: a separate ʼa- meaning “away” and a separate ʼíí- meaning “into” were not established by these checked examples. In NX-096 the source gives the “into” gloss to yah; the following inflected form is not evidence that every similar-looking initial sequence means “into”. Likewise, the “away out of sight” translation of NX-102 does not identify each prefix’s independent meaning.

3b. Postposition + verb meaning shifts

SOURCED Chee & Yazzie explicitly call these “phrasal verbs”. Their published pairs give the comparison that the first version lacked. The translations below are theirs, not translations composed from our glosses.

Verb aloneWith postpositionSource
NX-100 naashné
“I play”
NX-101 bee naashné
“I make him cry”
2025:22, (8). The prose glosses bee “by means of it”.
NX-102 ánáshwoʼ
“I usually walk away out of sight.”
NX-103 biká ánáshwoʼ
“I usually help him.”
2025:34, (14a–b). The postposition is glossed 3sg.for.
NX-104 honiigaii
“The area became hotter.”
NX-105 bitah honiigaii
“He became sick, ill.”
2025:35, (15a–b). The postposition is glossed 3sg.among.

SOURCED (source gloss, not an editorial decomposition) For NX-103, Chee & Yazzie print:

biká[á-ná]-[sh]-[l-woʼ]
3sg.for[thm- itr]Infl[1sg.sbj]AStem[val- walk.itr]VStem

SOURCED The source’s l and its thematic/iterative labels are retained. HYPOTHESIS The help and illness pairs are closer to an English prepositional idiom such as look after than to the placement alternation in pick the book up: they have a postposition bearing person information and a meaning belonging to the combination. “Closer” is a comparison, not an identity of structures. OPEN: lexical storage, productivity and the exact theta composition of each pair. “Make him cry” is not, by itself, proof of a productive syntactic causative.

3c. Does the disjunct prefix add an argument?

SOURCED McDonough prints ʼáshdįįh twice: “I disappear” (NX-098) and “I get rid of it” (NX-099), with different valence-marker analyses, ø versus ł (2019:40, (8)). She says valence is a property of the word form, not of the Model Paradigms alone. Chee & Yazzie describe postpositions as able to add an object or indirect object (2025:22). HYPOTHESIS These are reasons to test participants separately from the directional meaning. They do not isolate a disjunct prefix as the cause of a change in argument count. OPEN: a controlled “same stem and inflection, only this disjunct prefix differs” valency test.

4. Rice on scope: fixed order is not necessarily an arbitrary template

SOURCED Rice’s own freely readable chapter summaries distinguish three cases: fixed scope relations give fixed order; reversible scope can give order variation with interpretation; no scope relation can permit variation, with language-specific ordering still needing stipulation (2000, Chapter 6 summary; the chapter spans pp. 79–125). Chapter 15’s summary explicitly argues against an arbitrary template as the explanation (chapter pp. 373–384). These are publisher summaries, not a claim that the full chapters were read.

SOURCED Hale gives the critical structural comparison directly (2001:679–681). In his report of Rice, the preverb and stem form a low constituent; right-headed structure and stem raising through Voice to Mode/Subj produce the surface arrangement. Hale then reverses some headedness and scope assumptions and proposes a phonological skeleton account with “no head movement in the traditional sense” (p. 681). Bird’s freely readable 2001 review of Rice adds that scope predicts ordering within its domain but phonological, morphological or syntactic requirements can override it; the review is unpaginated.

Analysis A: Working default: our V→v→T derivation Hypothesis

Analysis A: Working default: our V→v→T derivationThe subject first merges in left Spec,vP and moves to left Spec,TP. chʼí is modeled as a low directional AdvP; dązh is the cited V stem. That category/attachment and OUT predicate are hypotheses. T⁰ and v⁰ are initial, lexical V⁰ final. V head-raises through v to T while the directional AdvP remains low; assembling its exponent into the same pronounced word requires a further PF operation. The AdvP adjoins to VP, so it can only modify the event (OUT(e): the event's path is outward); it cannot give the object a second theta role. The object's single role is Theme, from V. V→v→T and subject Spec,vP→Spec,TP are the working default in tensed clauses (audit pass 7, 2026-11-13). As drawn, T⁰ is initial (HYPOTHESIS), so the raised verb would be pronounced before everything left in vP. The Navajo verb-final order then needs something else to move above T: the remnant VP (object, PPs, adverbs), as in panel D of the assembly figure (HYPOTHESIS). The alternative is a final T⁰, where nothing else moves. Which is right is OPEN. LF: ∃e[jerk(e) ∧ Agent(e,a) ∧ Theme(e,t) ∧ OUT(e)] Full structure in the list below the figure.TPDPagent aT′T⁰vPDPagent av′v⁰VPAdvPchʼíVPV′DPtheme tV⁰dązhV+v+Thead-movehead-moveMove
(2) Our directional-prefix test on one vP/TP skeleton Hypothesis NX-094, Hale 2001:678–679(1–3): chʼíshidiníłdązh, with separately printed chʼí and dązh. No new Navajo is constructed. Word-internal interruption is the sourced phenomenon; low AdvP, T/v identity, subject movement and head movement are our hypotheses. Panel A is the working default (V→v→T, subject to Spec,TP). Hale 2001:681 supplies a brief no-head-movement contrast, not an axiom or a full model reproduced here. A box over T marks the V+v+T chain; it is not an independent syntactic word category.
Bracket notation (copy)

A: [TP [DP agent a] [T′ T⁰ [vP ⟨DP⟩ [v′ v⁰ [VP [AdvP chʼí] [VP [V′ [DP theme t] [V⁰ dązh]]]]]]]]

Structure as a list

Analysis A

  • TP
    • DP (id S-high; landing site of Move from subject): agent a
    • T′
      • T⁰ (head; id T; hosts the “V+v+T” (V + v + T))
      • vP
        • DP (id subject; struck copy; moved to S-high): agent a
        • v′
          • v⁰ (head; id v; head-moves to T)
          • VP
            • AdvP: chʼí
            • VP (id lowVP)
              • V′
                • DP (id object): theme t
                • V⁰ (head; id V; head-moves to v): dązh

HYPOTHESIS (our matched vP/TP comparison) The figure shows the working default: V→v→T plus separate subject→Spec,TP. (A second panel B, which held the same core and subject movement fixed but used PF/template assembly without V-head movement, was removed on Grant’s instruction on 2026-11-13; it carried no Navajo form not shown in panel A.) The directional AdvP remains low; incorporating its exponent into the pronounced word is a separate PF hypothesis. All specifiers are left. There is no extra PF-output branch: a dashed word box marks the complex head or its host. Hale’s no-head-movement proposal is a brief contrast, not our premise. OPEN: whether the directional exponent needs any PF placement beyond what the default gives, and, with T⁰ initial as drawn, whether the remnant VP moves above T for the verb-final order. A template or Mirror Principle does not decide this by itself. Dated correction, 2026-11-11: these figures replace earlier non-category labels and the mistaken “VPISH” node.

5. Telicity: a real English contrast, still no Navajo minimal pair

SOURCED Dehé prints “John ate the cake for an hour” versus “*John ate up the cake for an hour” (2002:6, (12a)), and “Ann used {up} her money {up}” ((12b)). Her braces mark alternative particle positions, not two copies in one sentence. This supports a completive contribution of up on the reported reading. Jackendoff’s “Elena drank the milk (completely) up” and “Elena drank it up/*drank up it” give a parallel (2002:76–77, (18a), (20a)).

HYPOTHESIS The careful generalization is “some particle uses impose or strengthen a completion reading”, not “particles always telicize”. An object’s quantity and the context also matter; eat the cake can already have a bounded reading. Our next test must hold those constant. SOURCED Jackendoff’s continuative particles are counterexamples to a blanket endpoint claim (pp. 77–80). OPEN — Navajo: no sourced pair here holds the stem, participants, mode and context constant while adding/removing just a directional prefix and testing duration or completion. Imperfective does not mean atelic, and perfective does not by itself prove telicity. “Climb up”, “inside” and “finished” in translations do not fill that gap.

6. Questions, negation and passives: the eight-cell check

SOURCED Jackendoff prints “*Up which answer did Bill look?” (2002:70, (3b)); Dehé likewise stars “*Up what did he toss?” and “*Out what did he figure?” (2002:10, (18c)). These are restrictions on taking the particle along with a wh-phrase, not evidence that particle verbs forbid questions. Dehé also prints a positive active wh-example, “Whati did they look up the information about ti?” (p. 19, (12b)), against “*Whati did they look the information about ti up?” (p. 18, (10b)). The indices and t are the source’s extraction notation, not pronounced words. This fills the English active/affirmative/wh cell with an order-sensitive extraction contrast. These are her printed judgments, not a new claim of agreement across English speakers. No question is manufactured in Navajo. NX-106’s translation “I ask you about it” is a declarative with the verb ask, not a question.

Question / polarityActivePassive
Polar · affirmativeEnglish OPEN; Navajo OPENEnglish OPEN; Navajo OPEN
Polar · negativeEnglish OPEN; Navajo OPENEnglish OPEN; Navajo OPEN
Wh · affirmativeEnglish SOURCED: wh-extraction contrast, Dehé pp. 18–19; also starred particle pied-piping. Navajo OPEN.English OPEN; Navajo OPEN
Wh · negativeEnglish OPEN; Navajo OPENEnglish OPEN; Navajo OPEN

NONE: no Navajo particle/preverb/postposition minimal question–negative–passive paradigm has been verified in this addition. Older general question examples elsewhere in the index are not silently reassigned to this construction. Non-question passive evidence in the English literature likewise cannot stand in for a passive-question cell.

7. The comparison, without conflating the constructions

TestEnglish particle verbNavajo preverb + stemNavajo postposition + verb
SeparationSOURCED: a full object may separate V and particle; weak-pronoun placement is restricted.SOURCED: prefixes can intervene inside a word. English-style object/particle shift is OPEN, not demonstrated.SOURCED: separate-word forms are attested here. Whether a noun can intervene in each specific pair is OPEN.
MeaningSOURCED: directional, idiomatic and aspectual uses overlap in placement behavior.SOURCED: directional and theme-dependent meanings. A fixed independent gloss for every prefix is OPEN.SOURCED: explicit play/make-cry, walk/help and heat/illness meaning shifts.
ArgumentsSOURCED: literature accounts disagree on whether an object belongs first with P, V or a Result projection.OPEN: no prefix-only argument-count contrast isolated. Different valence bases are not that test.SOURCED: authors describe argument addition. Exact theta decomposition of the idiomatic pairs is OPEN.
Wordhood / storageSOURCED: idiomatic listing does not entail grammatical wordhood (Jackendoff); Dutch separable combinations are a useful third comparison (Booij).SOURCED: one phonological word can contain a discontinuous lexical dependency (Hale).HYPOTHESIS: meaning may be listed for a phrase. Spacing alone proves neither storage nor free syntactic separability.
EndpointSOURCED: completive up contrasts with continuative uses.OPEN: no controlled telicity pair verified.OPEN: a meaning shift is not a telicity test.

HYPOTHESIS In plain words: Navajo has both word-internal verb partners and separate postposition–verb expressions with special meanings. English particle verbs share some of those properties, but not all. The strongest next test is not another translation match: it is a controlled separation, scope or argument-structure contrast that one analysis predicts and another does not.

Sources and exactly what was read

2026-11-11 — HYPOTHESIS / OPEN: Prefix–stem interruption does not establish where head direction changes. The dated head-by-head audit distinguishes selecting heads, argument prefixes, clitic hosts and three proposed boundaries; no unique word-edge switch is demonstrated.

2026-11-15 · First-principles review 16: Willie on subject and object prefixes, reflexives and reciprocals, the agentive passive, the indefinite ʼa- and postpositions (ch. 2, §§2.1–2.5)

2026-11-14: the “On the spine” summary at the top of this page supersedes this section’s framing. Its data, judgments and citations are unchanged.

HYPOTHESIS Plain-words verdict. The clearest result is about the passive. Navajo uses one prefix shape, ʼadi-/ʼádi-, for both “-self” and the unnamed doer of the agentive passive. It goes in the object’s place in the word when it means “-self” (“you shave yourself”) and in the subject’s place when it means “someone” (“you are being shaved”). So where a prefix goes depends on the role of the participant it stands for, not on which prefix it is. The passive keeps a real doer, an unnamed person, in the subject position (H10b), and the person being shaved keeps the object prefix. Under the V-to-T default that doer is the subject, and nothing else has to move: this passive does not make the object into the subject. One rule, “a noun can only go with a third person” (H31), plus the fact that the reflexive, the unnamed doer, the indefinite ʼa- and the area ho- have no person at all (H60), explains Willie’s observation that none of them can be paired with a noun. The distributive da- again counts only the subject (H21a passes). Willie says nothing can come between a postposition and the verb (her “cannot be interrupted”). That follows from the V-to-T default with all heads final and all specifiers on the left, with no special verb-sentence unit. Two things are NOT UNDERSTOOD: why the reflexive and the passive both change the verb’s “classifier” while the indefinite object ʼa- does not (H59 is only a description), and *(41c), still, as in review 15. Telicity: the mode paradigm, the reflexive/passive pairs and the three “ruin” sentences keep the verb fixed, so their contrasts are not about telicity; the choice of “to Flagstaff” vs “as far as Flagstaff” changes how bounded the path is without changing the syntax. There are no questions and no negation on these pages; the da- here is the distributive, not the negative da. The passive cells of the grid gain no question or negative example.

Starting point

Assumed, by instruction: VPISH (each verb’s subject argument first merges in Spec,vP, where it gets its theta role; this frames the arguments and is not a word-order account), theta roles, and the working default for tensed clauses: V→v→T head movement, with the subject moving Spec,vP→Spec,TP (on to Spec,CP or C when needed). Willie’s statement that “there are no non-finite verb forms” (p. 24) means every clause here has a T⁰ for the verb to reach. HYPOTHESIS Everything else is tested: H1–H57 from reviews 1–15 (here mainly H10a/H10b, H21a, H31, H54a/H54c, H55), head direction (HS-1, T⁰ final; the derivations are written for HS-1), Agree, and the order of prefixes inside the word (OPEN on this page’s “On the spine” summary; restated 2026-11-15 as Grant’s Merge-site account of the word/clause headedness mismatch, under which the mirror-image order is the prediction). The live question for word order is what else moves. Every derivation ends in a logical form (LF) whose scope order matches c-command, and each lists the hypotheses it uses beyond VPISH, theta roles and the default. Silent pronouns are written with their person: x₁ (first), y₂ (second), x₃, y₃ (third), u (the indefinite human agent), r (an anaphor), a (areal). The role labels (Agent, Patient, Theme, Goal, Experiencer) are ours. Willie’s analysis (incorporated pronouns with traces, nouns as adjoined non-arguments, the Maximal Verb-Sentence “MVS”) appears only as labeled contrast.

The source’s claims, as its own

  • SOURCED Prefix order (§2.1). “The subject is the innermost (rightmost) argument, closest to the verb stem, while the object is to the left of (external to) the subject. These pronominal prefixes are definite in reference” (p. 27). “the relative order of the prefixes that mark the verbal arguments is consistent, with the patient always to the left of the agent, and the postpositional phrases leftwards of the direct arguments” (p. 50). Mode: “the stem changes from one aspect/mode to another” (p. 30).
  • SOURCED Distributive da- (§2.1.1). “In direct constructions, when da- appears, the subject must be plural (three or more), and the number of the object is irrelevant. Da- is not an argument, but marks a semantic feature of the subject argument” (p. 32).
  • SOURCED Anaphors (§2.1.2). “These anaphors mark arguments that have patient theta roles, and cannot have independent reference. These anaphoric prefixes do not vary with person or number. When an anaphor is present, it always requires a -d or -l classifier, marking the verb intransitive. No pronoun or nominal that is coreferent with the reflexive may be adjoined to the sentence” (p. 32). A plural reflexive also has a reciprocal reading (14′), (15′); with the reciprocal “only one reading is possible” (p. 34).
  • SOURCED Agentive passive and impersonal (§2.3). These prefixes “exclude adjoined coreferent nominals” and show “a classifier that marks the construction as intransitive” (pp. 35–36). In the agentive passive “the indefinite agent must be understood as human, but no noun that describes this agent may be adjoined”; its prefix is “identical in phonological shape to the reflexive anaphor”, but “occurs to the right of (internal to) the object pronoun, proving that it marks the unspecified agent, and is not an object” (p. 36). “In both these construction types, the patient precedes the agent in the prefix array” (p. 37). The ʼa- prefix “never permits an adjoined coreferent nominal” (p. 38); (26) “is a transitive construction with an indefinite inanimate object argument, and permits an adjunct to the third person subject pronoun” (p. 39).
  • SOURCED Postpositions (§2.4). “The postpositional object pronouns and the possessive pronouns are identical in form” (p. 39). “The grammatical postpositions occur only with pronominal objects and are a component of the MVS. The lexical postpositions occur only with nominal (lexical) objects” (p. 40). “It is not possible to use -goo, a lexical postposition, with an animate object” (p. 43). “While it is not possible to stack two grammatical or two lexical postpositions, it is possible to have the kind of combination shown in (36)” (p. 44). The MVS “(a) cannot be interrupted; (b) … may constitute a single phonological unit” (p. 50, (44)).
  • SOURCED Pleonastic ho- (§2.5). The same claims and examples as ch. 4 §4.1 (review 15), printed earlier with some differences (below); ch. 2 adds that “Example (38) above also has a pleonastic ‘areal’ subject” (p. 46).

The checked data

SOURCED Exact forms, source glosses and translations, with example numbers and printed pages. Glosses list object before subject (nishédléézh “2s.1s.painted” is “I painted you”). Printing notes and crops are on each index entry.

  • Person and mode (pp. 29–31). NX-477 (2): déyá “1s.will go”, “I will go” · NX-478 (3): naashnish “1s.working”, “I am working.”; prose (citing Hale 1977): also “he worked”, with -sh marking perfective · NX-479 (4a): yi-sh-kʼá (printed with hyphens that the gloss line does not mirror): “3s.1s.grind:IMPF”, “I am grinding it.” · NX-480 (4b): ná-sh-kʼááh (printed with hyphens that the gloss line does not mirror): “iter.3s.1s.grind.IMPF.”, “I usually grind it.” · NX-481 (4c): yíkʼą́ “3s.1s.grind.PERF.”, “I ground it” · NX-482 (4d): deeshkʼááł “3s.1s.grind.FUT.”, “I will grind it.” · NX-483 (4e): wóshkʼááł “3s.1s.grind.OPT.”, “I wish to grind it.” · NX-484 (5): ashkii, shishgish “1s.3s.cut”, “The boy cut me.” · NX-485 (6): ashkii, nishgish “boy 2s.3s.cut”, “The boy cut you.” · NX-486 (7): nishédléézh “2s.1s.painted”, “I painted you.” · NX-487 (8): shédléézh “3s.1s.painted”, “I painted him.”
  • Distributive (p. 32). NX-488 (9): dayiiłtsą́ “PL.3.3.saw”, “They saw him/them.”, *“He saw them.”
  • Reflexives and reciprocals (pp. 33–35). NX-489 (10): yiiłtsą́ “3.1s.saw”, “I saw it.” · NX-490 (11): ʼádiistsą́ “REF.1s.saw”, “I saw myself.” · NX-491 (12): yiichʼid “1-dual-scratching”, “We are scratching it.” · NX-492 (13): ʼádiichʼid “REF-1-scratching”, “We are scratching ourselves.” · NX-493 (14): ʼádadiidleesh “REF.PL.1dl.painting”, “We are painting ourselves.” / (14′) “We are painting each other.” · NX-494 (15): ʼádadiilzhííh “REF.PL.1dl.blackening”, “We are blackening ourselves.” / (15′) “We are blackening each other.” · NX-495 (16): ʼahiidleesh “REC.1dl.painting”, “We are painting each other.” · NX-496 (17): ʼahiichʼid “REC.1dl.scratching”, “We are scratching each other.” · NX-497 (18): daʼahiidleesh “DIST.REC.1dl.painting”, “We are painting each other.” · NX-498 (19): daʼahiichʼid “DIST.REC.1dl.scratching”, “We are scratching each other.”
  • Reflexive vs agentive passive (p. 37). NX-499 (20): ʼádílzhééh (printed underlying form ʼádi-ni-l-zheeh) “REF.2s.shaving”, “You are shaving yourself.” · NX-020 (21): niʼdilzhééh (printed underlying form ni-ʼadi-l-zheeh) “2s.PASS.shaving”, “You are being shaved.” · NX-500 (22): ʼádíshdleesh “REF.1s.painting”, “I am painting myself.” · NX-021 (23): shiʼdidleesh “1s.PASS.painting”, “I am being painted.”
  • Indefinite ʼa- (pp. 38–39). NX-022 (24): ʼííchxǫʼ “unspec. subj. ruined”, “Ruining took place.” · NX-023 (25): (hatáál) yíchxǫʼ “(The ceremony) it is ruined.” · NX-501 (26): ʼasdzą́ą́ ʼííłchxǫʼ (no gloss line) “The woman did the ruining (of it).”
  • Postpositions (pp. 41–44). NX-502 (29): shimá bil déʼáázh “1.mother 3.with dl.walk”, “I am going with my mother.” · NX-503 (31): Kinłání=góó déyá “Flagstaff=to I will go”, “I will go to Flagstaff.” · NX-504 (32): Kinłání=jįʼ déyá (=jįʼ glossed “to”) “I will go as far as Flagstaff.” · NX-505 (33): ʼawééʼ bichʼįʼ yishááł “baby 3.to 1.walking”, “I am walking to the baby.” · NX-506 *(34): ʼawééʼ-góó yishááł “(I am walking to the baby.)” · NX-507 (35): hooghan=góó déyá “I am going home.”, *“I am going to the hoghan.” · NX-508 (36): dibé baghan bitsʼą́ą́jįʼ yishááł “sheep 3s.home 3s.from.at 1s.walking”, “I am walking away from the corral.” · NX-509 (38): Tsintah hodiʼil “Among the trees/forest, it is dense.” · NX-510 (39): Nihitah deeyá “Among us he will come.”
  • Pleonastic ho-, ch. 2 printing (pp. 45–49). NX-511 (40a): ditą́ʼ · NX-512 (40b): tó ditą́ʼ · NX-513 (40c): hoditą́ʼ · NX-514 *(40d): tó hoditą́ʼ · NX-515 (40e): táyiʼ hoditą́ʼ · NX-516 (41a): shił nizhóní · NX-517 (41b): shił hózhóní “I like it there (the area).” · NX-518 *(41c): kintah shil nizhóní “I like the town.” · NX-519 (41d): kintah-di shił hózhóní “I like it there in town (the area).” · NX-520 #(41e): kintah-di shił nizhóní “I like him/it [when I’m] in town.” · NX-521 (42): Shił béé hózin “[about X, knowing is with me.] I know it.” · NX-522 (43): Kee yóóʼeelwod=ígíí bił béé hózin “He/she knows that Kee ran away.”

OPEN Two printings of the pleonastic examples. Ch. 2 (40)–(43) and ch. 4 (1)–(4) (NX-445–457) differ, and both are indexed as printed: (i) ch. 2 prints “deep” with a final glottal stop (ditą́ʼ, NX-511–515), ch. 4 without (ditą́); we do not decide which spelling is right (SOURCE CONFLICT). (ii) (41e) is marked # (good string, meaning not the one under study), where ch. 4 (2e) has *; Willie’s prose on both pages gives the good reading. This bears out review 15’s reading of *(2e) as a starred reading, not a bad string. (iii) (41c) prints shil with a plain l; ch. 4 prints shił. (iv) Translations differ for (41b), (41d), (42); ch. 2 has no counterpart of ch. 4’s *(1f).

Hypotheses tested, revised or added

  • HYPOTHESIS H58 · prefix slots follow structural position (new). Whatever rule orders the person prefixes inside the verb word refers to the participant’s place in the structure: the prefix for the argument merged inside VP (Patient) is outer, and the prefix for the argument merged in Spec,vP (Agent) is inner. It does not refer to a list of morphemes. H58 does not say how the word is spelled out, which stays OPEN (tested against Grant’s Merge-site account on the unified structure §3.2). H58a: agreement for the VP argument is spelled on a head above the head that spells agreement for the Spec,vP argument. This needs that head to skip the closer subject, a problem we cannot yet solve (OPEN). H58b: a word-level spell-out rule orders agreement by role.
  • HYPOTHESIS H59 · the classifier with anaphors and the agentive passive (new; a description, NOT UNDERSTOOD). The d-effect classifier (Willie’s “-d or -l”) spells a value of v⁰ that appears when the VP argument is an anaphor bound by the Spec,vP argument, or when the Spec,vP argument is the indefinite human ʼadi-. It is not “one fewer argument”: the reflexive has two theta roles. It is not “a personless argument”: the indefinite object ʼa- of (26) comes with the transitive ł (Willie, p. 39).
  • HYPOTHESIS H60 · personless pronouns (new). The anaphors ʼádi-/ʼahi-, the passive agent ʼadi-, the indefinite ʼa- and the areal ho- (H54a) carry no person value. With H31 (a noun is third person and can be linked only to a third-person participant), no noun can be linked to any of them. This merges four of Willie’s separate statements into one.
  • HYPOTHESIS H62 · reflexive vs reciprocal (new). ʼádi- is an anaphor whose value is identical to its binder’s. With a plural binder, “we paint us” is true whether each paints himself or each paints the others, so both (14) and (14′) are available. ʼahi- adds a distinctness condition (each paints a different member) and needs a plural binder, so it has one reading.
  • HYPOTHESIS H63 · stem shape by mode (new). The stem’s shape (-kʼá, -kʼááh, -kʼą́, -kʼááł in (4a–e), as printed) is chosen inside the V+v+T complex by the mode value of T⁰ (or an aspect head on the way), a local context created by V→v→T.
  • HYPOTHESIS H64 · grammatical vs lexical postpositions (new). A grammatical postposition has an agreement probe. Its object (a pronoun, or a noun such as ʼawééʼ) sits in its left specifier, and the prefix (bi-, shi-, nihi-) is the agreement, as in a possessed noun (dibé baghan “sheep 3s.home”). A lexical postposition has no probe and takes a place-denoting complement directly (Kinłání=góó). Willie’s “grammatical: only pronominal objects; lexical: only nominal objects” becomes agreement vs no agreement.
  • HYPOTHESIS H65 · inclusory “with” (new, one example). In (29) the dual subject is “my mother and I”, and the “with her” phrase names the other member. Rival H65′: the dual subject is “I and someone else”, and the mother is a third companion.
  • HYPOTHESIS Tested again: H10a/H10b (passive with no agent vs a silent indefinite agent: H10b supported for the agentive passive), H21a (Dist⁰ just above vP agrees with the closest argument: passes), H31 (passes on (5), (6), (26), (29), (33), (36); see H60), H54a/H54c (pass on (38)), H55 (still the only account of *(41c)).

Derivations, step by step

  1. One string, two derivations: NX-478 (3) naashnish. Reading 1 “I am working”: merge V “work”; merge v and x₁ in Spec,vP (Agent); merge T with the imperfective value; V→v→T, x₁ raises to Spec,TP; the word spells first-person subject agreement as -sh-. LF: ∃e[work(e) ∧ Agent(e,x₁) ∧ ongoing(e)]. Reading 2 “he worked” (prose): the same, with a silent x₃ and the perfective value, which is what -sh spells here (Willie, citing Hale 1977). LF: ∃e[work(e) ∧ Agent(e,x₃) ∧ completed(e)]. Result: the ambiguity needs no movement difference, only two different heads on the verb’s spine (subject agreement vs mode) with the same exponent. Uses: H63 (mode on T⁰ or an aspect head), HS-1.
  2. Mode paradigm: NX-479–483 (4a–e). All five: merge V “grind” with a silent y₃ as its complement (Theme); merge v and x₁ in Spec,vP (Agent); merge T with the mode value; V→v→T; x₁ raises to Spec,TP. The mode values are IMPF (4a), iterative + IMPF (4b, with the prefix ná-), PERF (4c), FUT (4d), OPT (4e). LF (4c): ∃e[grind(e) ∧ Agent(e,x₁) ∧ Theme(e,y₃) ∧ completed(e)]. LF (4e): WISH(x₁, ∃e[grind(e) ∧ Agent(e,x₁) ∧ Theme(e,y₃)]), with WISH contributed by T⁰ above the event, which matches c-command (as in review 14). The stem shape changes with the mode while the participants stay fixed. Under the default, V ends up in the same complex as the mode head, so H63 is local. Result: fits the default, but does not decide it. A rival in which T⁰ is spelled on the verb by a post-syntactic lowering would be just as local (labeled contrast, not adopted). The iterative changes the stem too ((4a) vs (4b)), so whatever iterative ná- spells is also on the verb’s spine (OPEN which head). Uses: H63, HS-1.
  3. Noun with a first/second-person object: NX-484, NX-485 (5), (6). Merge V “cut” with a silent x₁ (5) or y₂ (6) as its complement (Patient); merge v and ashkii in Spec,vP (Agent); merge T; V→v→T; ashkii raises to Spec,TP and is pronounced first. H31: the noun can be linked only to a third person, and only the agent is third person, so there is one reading, as printed. LF (5): ∃e[cut(e) ∧ Agent(e,boy) ∧ Patient(e,x₁)]. Willie’s comma (an adjoined noun) is not needed. Uses: H31.
  4. Two pronouns: NX-486, NX-487, NX-489, NX-491 (7), (8), (10), (12). E.g. (7) nishédléézh: y₂ Patient in VP, x₁ Agent in Spec,vP, default. LF: ∃e[paint(e) ∧ Agent(e,x₁) ∧ Patient(e,y₂)]. In every case the source gloss puts the object before the subject, matching Willie’s “patient always to the left of the agent” (p. 50). Nothing moves beyond the default. The prefix order is H58 (below, step 7).
  5. Distributive: NX-488 (9) dayiiłtsą́. Merge V “see” with y₃ (Theme); merge v and X in Spec,vP (Agent); merge Dist⁰ just above vP. By H21a, Dist⁰ agrees with the closest argument it c-commands, the agent, which must be plural. Merge T; V→v→Dist→T; X raises to Spec,TP. LF: ∃e[see(e) ∧ Agent(e,X) ∧ Theme(e,y₃)], |X| ≥ 3, y₃ singular or plural. *“He saw them”: a singular agent fails Dist⁰’s agreement, whatever the number of the theme. Result: H21a passes with silent arguments, as it did with nouns (NX-272, *NX-273) and in the inverse (NX-274, plural agent only). Uses: H21a.
  6. Reflexives and reciprocals: NX-490, NX-492–498 (11), (13)–(19). (11) ʼádiistsą́: merge V “see” with the anaphor r as its complement (Patient, as Willie says); merge v and x₁ in Spec,vP (Agent). x₁ c-commands r inside vP, so r is bound at once. VPISH gives the binding configuration with no movement. Merge T; x₁ raises to Spec,TP. LF: ∃e[see(e) ∧ Agent(e,x₁) ∧ Theme(e,r) ∧ r = x₁]. Compare (10), the same verb with a free y₃. (14) ʼádadiidleesh: the agent X is plural (Dist⁰, H21a; three or more); r = X. Reflexive reading (14): ∀x∈X ∃e[paint(e) ∧ Agent(e,x) ∧ Patient(e,x)]; reciprocal reading (14′): ∀x∈X ∃y∈X[y ≠ x ∧ ∃e[paint(e) ∧ Agent(e,x) ∧ Patient(e,y)]]. Both satisfy “r = X” (H62). (16) ʼahiidleesh: the reciprocal anaphor brings y ≠ x and needs a plural binder (dual here), so only the second LF. (18), (19) add Dist⁰: three or more. Willie: no noun may be linked to the reflexive. Under H60 + H31, r has no person, so no noun can be linked to it. OPEN The printed forms place da after ʼá in the reflexive (ʼádadiidleesh) but before ʼahi in the reciprocal (daʼahiidleesh). The source gloss does not segment the words, so this is an observation for the prefix-order rule, not a result. Uses: H21a, H60, H62, H31.
  7. Reflexive vs agentive passive: NX-499 (20), NX-020 (21), NX-500 (22), NX-021 (23); tree panels A, B. (20) ʼádílzhééh (ʼádi-ni-l-zheeh): r (Patient) in VP, y₂ (Agent) in Spec,vP; default. LF: ∃e[shave(e) ∧ Agent(e,y₂) ∧ Patient(e,r) ∧ r = y₂]. (21) niʼdilzhééh (ni-ʼadi-l-zheeh): y₂ (Patient) in VP, spelled by the object prefix ni-; u, the indefinite human pronoun (Agent), in Spec,vP, spelled ʼadi-; merge T; by the default u raises to Spec,TP. LF: ∃e[shave(e) ∧ Agent(e,u) ∧ Patient(e,y₂)], u = some human or other. What the pair forces. (a) The same-shaped prefix is outer when it is the Patient (20) and inner when it is the Agent (21); second-person ni- does the reverse. The slot follows the participant’s structural place, not the morpheme (H58; SOURCED-supported, narrow). A template that lists prefixes in fixed slots needs two homophonous entries to state this; Willie in effect lists two (anaphor vs “unspecified agent”). (b) The passive prefix sits in the agent’s slot and means a human agent, so the agentive passive merges a real agent argument in Spec,vP: H10b, not H10a, for this construction. The simple passive (25) yíchxǫʼ (NX-023) is still OPEN between them. (c) The patient keeps its object prefix, so nothing in the word shows it becoming the subject. Under the default the Spec,vP agent u is what raises, and the patient stays in VP: this passive is not object promotion, and nothing beyond the default moves. Whether a patient noun could still raise is OPEN (no noun with these verbs is printed). (d) The anaphor and u both lack person (H60), so no noun can be linked to either. That gives Willie’s “no noun that describes this agent may be adjoined” (p. 36). (22)/(23) repeat the pattern with “paint”. Uses: H10b, H58, H60, H31, H59 (the classifier).
  8. Indefinite ʼa-: NX-022 (24), NX-023 (25), NX-501 (26). (24) ʼííchxǫʼ: the only argument is the indefinite ʼa- (merged as V’s complement, the thing ruined; label ours). It raises to Spec,TP if a lone theme does so (H26, OPEN since review 8). LF: ∃e∃z[ruin(e) ∧ Theme(e,z)], z = something. No noun can be linked (H60 + H31), matching “never permits an adjoined coreferent nominal”. (25) (hatáál) yíchxǫʼ: a third-person theme, so the noun is allowed (H31). (26) ʼasdzą́ą́ ʼííłchxǫʼ: merge V “ruin” with indefinite ʼa- (z) as complement (Patient); merge v (transitive ł, Willie) and ʼasdzą́ą́ in Spec,vP (Agent); default; the noun raises to Spec,TP and is pronounced first. LF: ∃e∃z[ruin(e) ∧ Agent(e,woman) ∧ Patient(e,z)], ∃z inside the clause, below nothing else (no scope interaction). Result: H60 + H31 pass. H59 is shown to be descriptive: a personless object here comes with ł, not the d-effect (NOT UNDERSTOOD why). Willie’s prose calls the (24) argument “animate” on p. 38 and “inanimate” on p. 39 (kept, not resolved; NX-022 note). Uses: H60, H31, H26 (OPEN), H59.
  9. Grammatical postpositions: NX-505 (33), NX-510 (39), NX-502 (29), NX-508 (36); tree panel C. (33) ʼawééʼ bichʼįʼ yishááł: merge P -chʼįʼ “to” with ʼawééʼ in its left specifier; P’s probe agrees with it (third person, bi-; H64, H31). Merge V “walk” with the PP as complement (Goal); merge v and x₁ (Agent); default. LF: ∃e[walk(e) ∧ Agent(e,x₁) ∧ Goal(e,baby)]. (39) Nihitah deeyá: P -tah “among” agreeing with a silent first-person plural (nihi-); x₃ agent; future. LF: ∃e[come(e) ∧ Agent(e,x₃) ∧ among(e,us) ∧ future(e)]. (36) dibé baghan bitsʼą́ą́jįʼ yishááł: dibé in the specifier of the possessed noun baghan (b- agreement); that noun phrase in the specifier of grammatical -tsʼą́ą́ “from” (bi- agreement); lexical =jįʼ (glossed “at”) on top: [PP [PP [DP dibé baghan] bitsʼą́ą́] jįʼ]. LF: ∃e[walk(e) ∧ Agent(e,x₁) ∧ Source(e,corral)] (Willie’s translation; “sheep’s home” = corral). H64 explains why two grammatical postpositions cannot stack: the upper one’s probe would find only a PP, which has no person to agree with (cf. H60). Why two lexical ones cannot stack is NOT UNDERSTOOD; only a selection stipulation states it. No stacked form is printed; this is a prose claim. (29) shimá bil déʼáázh: shimá in the specifier of -ł/-l “with” (printed plain l); the dual subject X (silent) is in Spec,vP and then Spec,TP. H65: ∃e[go(e) ∧ Agent(e,X) ∧ X = x₁ ⊕ mother(x₁)], the PP naming the non-speaker member; H65′: X = x₁ ⊕ someone, plus the mother as companion. The translation “I am going with my mother” with a dual verb favors H65 (OPEN; one example). Uses: H64, H31, H65.
  10. “Cannot be interrupted” (44a) derived. In all of (29), (33), (36), (39) and (42)–(43), the postposition phrase is inside VP (complement or specifier). The verb word is in T⁰, and every head is final with every specifier on the left, so the pronounced order is Spec,TP – Spec,vP – (VP: PP – V) – v – T. No position is pronounced between the VP-internal PP and the verb word in T⁰. Material can come between them only if the PP itself moves leftward. Result: Willie’s first MVS property follows from the default + HS-1 + left specifiers, with no MVS constituent (HYPOTHESIS). (44b) “a single phonological unit” is a sound-level fact not tested here. OPEN These pages print no interrupted string, starred or otherwise; the test is a starred adverb or noun between a grammatical PP and its verb (to look for in ch. 3).
  11. Lexical postpositions: NX-503, NX-504, NX-506, NX-507, NX-509 (31), (32), *(34), (35), (38). (31) Kinłání=góó déyá: merge lexical P =góó with the place noun as complement (no agreement, H64); the PP is V’s Goal; x₁ agent; future. LF: ∃e[go(e) ∧ Agent(e,x₁) ∧ Goal(e,Flagstaff) ∧ future(e)], with “toward the general area” as =góó’s meaning. (32): the same with =jįʼ “up to a specific point”. The two differ only in P’s meaning (Willie, p. 43). *(34) ʼawééʼ-góó yishááł: =góó selects a place, and a baby is not a place. That is selection, not syntax; contrast (33), where the grammatical P agrees with the noun. (35): the “home” reading is lexicalized; *“to the hoghan” is an idiom fact with no structural content. (38) Tsintah hodiʼil: lexical -tah with the noun tsin “trees”; the areal pronoun a is the Theme of “dense”, and the place phrase, adjoined to TP, restricts it (H54a, H54c). LF: ∃s[dense(s) ∧ Theme(s,a) ∧ among(a,trees)]. Result: H54a/H54c pass out of sample (a new verb), and no noun is linked to the areal argument (H60). Uses: H64, H54a, H54c, H60.
  12. Pleonastic examples, ch. 2 printing: NX-511–522 (40)–(43). Derived as review 15 derived (1)–(4): H54a for (40c), (40e), (41b), (41d); H31 + H54a for *(40d); H54b or H54′ for (42), (43), with the =ígíí clause either adjoined (H56a) or moved (H56b). The # of (41e) fits review 15’s second derivation of (2e) (place phrase as an adjunct locating the liker, with a third-person theme). *(41c) kintah shil nizhóní stays NOT UNDERSTOOD: H31 and the default predict it good; only the stipulation H55 excludes it.

Analysis A: (20) NX-499 ʼádílzhééh “You are shaving yourself”: the agent is a second-person pronoun; the patient is the anaphor ʼádi-, bound inside vP Hypothesis

Analysis A: (20) NX-499 ʼádílzhééh “You are shaving yourself”: the agent is a second-person pronoun; the patient is the anaphor ʼádi-, bound inside vPPanel A. Merge V “shave” with the anaphor r as its complement (Patient; label ours); merge v and the silent second-person pronoun y₂ in Spec,vP (Agent). y₂ c-commands r already inside vP, so r is bound there (VPISH alone supplies the binding configuration; no movement is needed for it). Merge T; y₂ raises to Spec,TP. r has no person value of its own (H60), so no noun can be linked to it (H31). The d-effect classifier (printed l in the underlying ʼádi-ni-l-zheeh) is our H59, a value of v⁰ that co-occurs with the anaphor. This is the working default: V⁰→v⁰→T⁰, and the subject first merges in left Spec,vP, where it gets its role (VPISH), and raises to left Spec,TP; its lower copy is silent. With T⁰ final (HS-1, HYPOTHESIS) the verb word is pronounced last in TP. Nothing else moves. LF: ∃e[shave(e) ∧ Agent(e,y₂) ∧ Patient(e,r) ∧ r = y₂] Full structure in the list below the figure.TPDPAgent: silent second-person pronoun y₂ (subject prefix)T′vPT⁰DPAgent: silent second-person pronoun y₂ (subject prefix)v′VPv⁰V′DPPatient: anaphor r, spelled ʼádi- (outer, object-side prefix)V⁰ʼádílzhééhverb wordhead-movehead-moveMove

Analysis B: (21) NX-020 niʼdilzhééh “You are being shaved”: the same-shaped ʼadi- is now the agent, an indefinite human pronoun in Spec,vP; the second person stays the patient Hypothesis

Analysis B: (21) NX-020 niʼdilzhééh “You are being shaved”: the same-shaped ʼadi- is now the agent, an indefinite human pronoun in Spec,vP; the second person stays the patientPanel B. Merge V “shave” with the silent second-person y₂ as its complement (Patient), spelled by the object prefix ni-; merge v and the indefinite human pronoun u in Spec,vP (Agent), spelled ʼadi- in the subject slot, to the right of ni- (H10b). Merge T; by the default u raises to Spec,TP. The patient keeps its object prefix, so nothing in the word shows it becoming the subject; under the default it stays in VP. u has no person value (H60), so no noun can name the agent, as Willie reports. This is the working default: V⁰→v⁰→T⁰, and the subject first merges in left Spec,vP, where it gets its role (VPISH), and raises to left Spec,TP; its lower copy is silent. With T⁰ final (HS-1, HYPOTHESIS) the verb word is pronounced last in TP. Nothing else moves. LF: ∃e[shave(e) ∧ Agent(e,u) ∧ Patient(e,y₂)], u = some human or other Full structure in the list below the figure.TPDPAgent: indefinite human pronoun u, spelled ʼadi- (inner, subject-side prefix) (H10b)T′vPT⁰DPAgent: indefinite human pronoun u, spelled ʼadi- (inner, subject-side prefix) (H10b)v′VPv⁰V′DPPatient: silent second-person pronoun y₂, spelled ni- (object prefix)V⁰niʼdilzhééhverb wordhead-movehead-moveMove

Analysis C: (33) NX-505 ʼawééʼ bichʼįʼ yishááł “I am walking to the baby”: the noun sits in the postposition phrase, which stays in VP; no position lies between it and the verb Hypothesis

Analysis C: (33) NX-505 ʼawééʼ bichʼįʼ yishááł “I am walking to the baby”: the noun sits in the postposition phrase, which stays in VP; no position lies between it and the verbPanel C. Merge P “to” (bichʼįʼ) with the noun ʼawééʼ in its left specifier; P agrees with it (bi-, third person, H31, H64). Merge V “walk” with the PP as its complement (Goal); merge v and the silent first-person x₁ in Spec,vP (Agent). Merge T; x₁ raises to Spec,TP. With every head final and every specifier on the left, no position is pronounced between the PP and T⁰, so nothing can come between the postposition and the verb word: Willie’s “cannot be interrupted” (44a) follows without a separate verb-sentence constituent. This is the working default: V⁰→v⁰→T⁰, and the subject first merges in left Spec,vP, where it gets its role (VPISH), and raises to left Spec,TP; its lower copy is silent. With T⁰ final (HS-1, HYPOTHESIS) the verb word is pronounced last in TP. Nothing else moves. LF: ∃e[walk(e) ∧ Agent(e,x₁) ∧ Goal(e,baby)] Full structure in the list below the figure.TPDPAgent: silent first-person pronoun x₁T′vPT⁰DPAgent: silent first-person pronoun x₁v′VPv⁰V′PPV⁰yishááłDPʼawééʼ “baby” (Spec,PP; linked by Agree to the bi- of the postposition, H64)P′P⁰bichʼįʼverb wordhead-movehead-moveMove
(1) Reflexive vs agentive passive, and a postposition phrase: ʼádílzhééh (NX-499), niʼdilzhééh (NX-020), ʼawééʼ bichʼįʼ yishááł (NX-505) Hypothesis NX-499, NX-020, NX-505: Willie 1991 ch. 2 (20), (21), (33), printed pp. 37 and 43: ʼádílzhééh “(ʼádi-ni-l-zheeh in underlying form)”, glossed “REF.2s.shaving”, “You are shaving yourself.”; niʼdilzhééh “(ni-ʼadi-l-zheeh in underlying form)”, glossed “2s.PASS.shaving”, “You are being shaved.”; ʼawééʼ bichʼįʼ yishááł, glossed “baby 3.to 1.walking”, “I am walking to the baby.” Leaves are printed words. T⁰, v⁰, the silent pronouns, the position of the noun inside the PP and the role labels are our assignments.
Bracket notation (copy)

A: [TP [DP Agent: silent second-person pronoun y₂ (subject prefix)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [DP Patient: anaphor r, spelled ʼádi- (outer, object-side prefix)] [V⁰ ʼádílzhééh]]] v⁰]] T⁰]] B: [TP [DP Agent: indefinite human pronoun u, spelled ʼadi- (inner, subject-side prefix) (H10b)] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [DP Patient: silent second-person pronoun y₂, spelled ni- (object prefix)] [V⁰ niʼdilzhééh]]] v⁰]] T⁰]] C: [TP [DP Agent: silent first-person pronoun x₁] [T′ [vP ⟨DP⟩ [v′ [VP [V′ [PP [DP ʼawééʼ “baby” (Spec,PP; linked by Agree to the bi- of the postposition, H64)] [P′ [P⁰ bichʼįʼ]]] [V⁰ yishááł]]] v⁰]] T⁰]]

Structure as a list

Analysis A

  • TP
    • DP (id S1-high; landing site of Move from S1): Agent: silent second-person pronoun y₂ (subject prefix)
    • T′
      • vP
        • DP (id S1; struck copy; moved to S1-high): Agent: silent second-person pronoun y₂ (subject prefix)
        • v′
          • VP
            • V′
              • DP: Patient: anaphor r, spelled ʼádi- (outer, object-side prefix)
              • V⁰ (head; id V1; head-moves to v1): ʼádílzhééh
          • v⁰ (head; id v1; head-moves to T1)
      • T⁰ (head; id T1; hosts the “verb word” (V1 + v1 + T1))

Analysis B

  • TP
    • DP (id S2-high; landing site of Move from S2): Agent: indefinite human pronoun u, spelled ʼadi- (inner, subject-side prefix) (H10b)
    • T′
      • vP
        • DP (id S2; struck copy; moved to S2-high): Agent: indefinite human pronoun u, spelled ʼadi- (inner, subject-side prefix) (H10b)
        • v′
          • VP
            • V′
              • DP: Patient: silent second-person pronoun y₂, spelled ni- (object prefix)
              • V⁰ (head; id V2; head-moves to v2): niʼdilzhééh
          • v⁰ (head; id v2; head-moves to T2)
      • T⁰ (head; id T2; hosts the “verb word” (V2 + v2 + T2))

Analysis C

  • TP
    • DP (id S3-high; landing site of Move from S3): Agent: silent first-person pronoun x₁
    • T′
      • vP
        • DP (id S3; struck copy; moved to S3-high): Agent: silent first-person pronoun x₁
        • v′
          • VP
            • V′
              • PP
                • DP: ʼawééʼ “baby” (Spec,PP; linked by Agree to the bi- of the postposition, H64)
                • P′
                  • P⁰ (head): bichʼįʼ
              • V⁰ (head; id V3; head-moves to v3): yishááł
          • v⁰ (head; id v3; head-moves to T3)
      • T⁰ (head; id T3; hosts the “verb word” (V3 + v3 + T3))

The V-to-T default on these pages

HYPOTHESIS All 46 new clauses and the four cited ones fit the default. Nothing beyond it moves on these pages: objects, postposition phrases and the patient of the passive stay in VP, and nouns before the verb are either the raised subject (Spec,TP) or inside a VP-internal phrase. Two cases are not covered by the default as stated, and are OPEN, not counterexamples: a lone theme ((24), (25)), whose raising to Spec,TP is H26; and the role-less ho- of “know” under H54b, which would merge directly in Spec,TP. Neither resists the default. NEG and tense/aspect: no negation appears. Mode sits on T⁰ (or an aspect head on the spine) and conditions the stem (H63).

Telicity

  • SOURCED Mode paradigm (4a–e): one verb (“grind”), one set of participants, five modes; the stem and prefixes change with mode while the predicate, and so its telicity, is fixed. The stem alternation is not telicity-driven (narrow).
  • SOURCED Reflexive vs passive: (20)/(21) “shave” and (22)/(23) “paint” hold the verb and the imperfective fixed; the voice contrast and the slot of ʼadi- do not depend on telicity (narrow).
  • SOURCED “Ruin” (24)–(26): one root, three constructions; the ban on a noun with ʼa- holds with the predicate fixed (narrow).
  • SOURCED Path (31)/(32): =góó “toward the general area” vs =jįʼ “up to a specific point” changes how bounded the path is (our description of Willie’s contrast) with the same verb, the same order and the same structure. A boundedness difference here leaves the surface grammar unchanged (narrow).
  • OPEN Reflexives and reciprocals occur with “see” (perfective), “scratch” (an activity; our classification, tagged atelic) and “paint/blacken” (imperfective). Willie claims no telicity effect, and the verbs differ in mode as well, so there is no minimal pair. Whether telicity restricts the anaphors is untested.

Question / negation / passive grid

SOURCED No question (no -ísh/-sh, daʼ, háí, haʼátʼíí) and no negation (no doo … da, lágo, hanii) appears on pp. 24–51 (text layer searched, renders checked). The da- of (9), (14), (15), (18), (19) is the distributive (glossed PL/DIST), not the negative da. Passive: the agentive passives (21)/(23) (NX-020/021) are affirmative declaratives, already in the grid since review 2. They now have a derivation under the working default (step 7: the indefinite agent is the subject, and the patient stays the object). No cell changes. Still OPEN: polar-aff × passive; polar-neg × inverse/passive; wh-neg × inverse/passive; any negative or questioned agentive passive. Sources to search: Willie ch. 7 (next); the NLA grammar draft chapters on voice, if on the box; the Navajo Questions database for entries glossed “pass” with -ísh (review 4 found only a candidate, NX-217, whose gloss lacks “pass”).

Light n

OPEN Willie notes that postposition objects and possessors have “identical” prefixes (p. 39; tables (27) p. 40 and (1) p. 28: shi-, ni-, yi-/bi-, ha-/ho- on postpositions and possessed nouns, as on verbal objects). If a light n introduced the possessor the way light v introduces the agent (Grant’s DP/TP symmetry point), one might expect subject-type marking on possessors; what is printed is object-type marking. This does not decide anything: the shape of an agreement prefix need not reveal which head introduced the argument. We record it only as a datum for the light n question. The possessor sits in the left specifier of the noun phrase in (36), whatever head introduces it.

Agreements and disagreements with Willie (labeled contrast)

  • Agree: patient prefix left of agent prefix (H58 states it by structure); da- counts the subject (H21a); anaphors and indefinites take no noun (H60 + H31); the passive ʼadi- is an agent (H10b).
  • Disagree: that anaphors make the verb “intransitive”: the reflexive has two theta roles (Agent, Patient), so the classifier does not count arguments (H59). That nouns are non-arguments adjoined to pronouns: not adopted; nouns occupy argument positions (Spec,vP→Spec,TP, VP, Spec,PP). That the MVS is a constituent with its own properties: (44a) follows from the default (step 10), so no MVS is posited.
  • Restated: grammatical vs lexical postposition as agreement vs no agreement (H64); *(34) as selection.

Predictions and what is still wanted (OPEN)

  • A third-person noun with an agentive passive verb is predicted good, linked to the patient (H31); it would also show whether the patient raises (H10b + default say it need not). Not printed here.
  • A noun linked to the antecedent of a reflexive with a third-person agent is predicted good (H60 says only the anaphor itself is personless).
  • An adverb or noun between a grammatical postposition phrase and its verb is predicted bad unless the PP has moved (step 10).
  • Stacked postpositions (two grammatical, two lexical) as printed starred strings, to test H64’s account and the NOT UNDERSTOOD lexical case.
  • A second inclusory “with” example, e.g. with a plural verb, to separate H65 from H65′.
  • Why the d-effect classifier appears with anaphors and the passive agent but not with indefinite ʼa- objects (H59, NOT UNDERSTOOD); the active counterparts of (20)–(23) are not printed here.
  • The prefix-order rule itself (H58a vs H58b; the placement of da in (14) vs (18)).
  • Which spelling of “deep” is right, ditą́ʼ or ditą́ (SOURCE CONFLICT between the two printings).

2026-11-15 · H59 and Japanese jibun/honorifics: a comparison

2026-11-14: the “On the spine” summary at the top of this page supersedes this section’s framing. Its data, judgments and citations are unchanged.

HYPOTHESIS Plain-words verdict: partial. For the classifier (H59), Japanese has no exact counterpart. Jibun leaves the verb unchanged. But Japanese does have the pattern of one voice suffix with several related uses: -(r)are marks the passive and also respect for the subject. The usual account of that suffix is that it plays down the doer (“agent defocusing”, Shibatani as reported in J2). Applied to Navajo, this gives a one-condition restatement of H59, called H59α′ below: in a two-role verb, d/l appears when the doer is not a separate, specific individual. Either the doer is the same as the one acted on (reflexive, reciprocal) or it is “someone” (agentive passive). That predicts the transitive ł of ʼasdzą́ą́ ʼííłchxǫʼ, where the woman is a specific doer, instead of listing it as an exception. It still doesn’t say why d spells this, so H59 stays NOT UNDERSTOOD at that level. A broader version that covers one-role verbs, H59α, is contradicted by a sourced Navajo fact: with “someone works” the l classifier is deleted rather than d added. Navajo also does not use the classifier for respect, which Japanese -(r)are does. For the ji- of respect and report (H57a), the parallel is real. Japanese respect for the subject or object is analyzed as agreement, with separate subject and object forms, works inside embedded clauses, and leaves the arguments unchanged. Navajo’s polite ji-/ho- fits that profile, which favors H57a over H57b (supporting, not decisive). Three axes give nothing for H59: jibun and gender, jibun as a first-person word in formal or hierarchical speech, and the addressee-politeness -mas. Empathy (Kuno) bears on the inverse and the fourth person, not on the classifier.

The Japanese facts, as the sources give them

  • SOURCED Subject and object honorifics are agreement (J1). “Since Shibatani (1977), subject honorification has been treated as an instance of (abstract) subject verb agreement”, with evidence credited to Toribio (1990), Ura (2000) and Hasegawa (2002) (J1 p. 454). J1 (1): Tanaka sensei-ga hon-o o-yomi-ni-nat-ta “Prof. Tanaka-Nom book-Acc read-SH-past”, “Prof. Tanaka read the book” (p. 454). The morphology, from Harada (1976, p. 504) as J1 (3) quotes it: “HP+INF ni nar-” for subject honorification and “HP+INF su-/itas-” for object honorification (p. 455). J1 (6): Taro-ga Tanaka sensei-o o-tasuke-si-ta “Taro-Nom Prof. Tanaka-Acc help-OH-past”, “Taro helped Prof. Tanaka” (p. 456). With a ditransitive, object honorification goes only to the indirect object: J1 *(8) Hanako-ga Mary-ni Tanaka sensei-o go-syookai-si-ta “Hanako introduced Prof. Tanaka to Mary” (p. 456). J1 explains this as defective intervention under Agree (abstract; §3).
  • SOURCED A reflexive cannot be honored (J1 p. 454 n. 1). *Tanaka sensei-ga zibun-o o-tasuke-si-ta “Prof. Tanaka-Nom self-Acc help-OH-past”, “Prof. Tanaka helped himself”, credited to Norbert Hornstein (p.c.): “As Woolford (1999) extensively showed, anaphors typically resist agreement. We therefore expect anaphors to resist honorification.”
  • SOURCED One voice suffix, four uses (J2). “The Japanese passive morpheme -(r)are … has four distinct uses that are historically related: passive, spontaneous, potential (ability), and honorific” (§1, p. 1). Passive J2 (1a): Max-ga Pat-ni nagur-are-ta “Max-Nom Pat-Dat hit-RARE-Past”, “Max was hit by Pat.” (§2.1, p. 2). Honorific J2 (12b): Sensei-ga hon-o kaw-are-ta “teacher-Nom book-Acc buy-RARE-Past”, “The teacher bought a book. (hon.)”, which J2 calls “synonymous to o-V-ninaru”, though more formal (§2.4, pp. 5–6). In the honorific use, -(ra)r/-(r)are “do not affect the argument pattern of the stem” (§3.4, p. 9). As J2 reports Shibatani (2000), the honorific use rests on “the avoidance of direct mention of the agent”, achieved by “defocusing the direct link between the honorified agent and the action” (§3.4, p. 9). J2 itself treats the passive’s core as “lack of control” and leaves open whether the honorific came from the spontaneous or the passive (pp. 2–3, 9–10). Shibatani 1985 (not opened) is the source of the agent-defocusing view of the passive. Its published abstract relates the passive to “reflexives and reciprocals” and to “spontaneous, potential, honorific, and plural formations”.
  • SOURCED Three uses of zibun (J3 §5.3.1, pp. 79–80). “Reflexive zibun must be bound to its co-argument subject. Empathic zibun is long-distance bound to a structurally commanding subject, and must be the empathy locus of the minimal clause or NP containing it.” The logophoric use gives the de se reading. J3 (19), from Kuno 1978:213 as J3 cites it: *Taro-wa [boku-ga zibun-ni kasi-ta] okane-o nakusite-simat-ta rasii, “It seems that Taro lost the money I lent him”. This is bad because empathic zibun “cannot occur as a co-argument of a first person pronoun”. Reflexive J3 (17) Taro-wa Hanako-ni zibun-o e-no moderu-tosite wariatete-{yat/kure}-ta “Taro assigned himself to Hanako as a portrait model (for her sake)”: the verb carries no reflexive morphology.
  • SOURCED Empathy and the inverse (J3 §7.1, pp. 166–173). Kuno’s empathy hierarchies include “It is easier for the speaker to empathize with the referent of the subject than with the referents of other NPs” (Kuno 1987:211 as J3 quotes it, p. 169). J3 classes the direct/inverse opposition and obviation, “attested in … Algonquian (e.g. Cree), Athabaskan (e.g. Navajo)”, as “devices to encode restrictions on the empathy relation”. It treats yaru/kureru “give” as “lexicalized direct/inverse verbs” (p. 172).
  • SOURCED Addressee politeness is different (J4). J4 separates content-oriented markers (referent honorifics) from utterance-oriented ones, following Harada’s (1976) ‘propositional’ vs ‘performative’ distinction (p. 1 n. 2). Japanese -mas “conveys the speaker’s politeness toward the addressee and the formality of the utterance situation”, J4 (8) Peter-wa sushi-o tabe-mas-i-ta “Peter ate sushi.”. It “cannot be embedded as standard indirect discourse” (Miyagawa 2012:89 as cited): J4 (9) Hanako-wa [dare-ga kuru/*ki-mas-u ka] sitte i-mas-u “Hanako knows who is coming.” (p. 4). Referent honorifics “are readily embeddable” (Korean (4), p. 3). J4 proposes that content-oriented markers “do not themselves directly mark politeness information, but rather are dependent on a (possibly null) c in the root clause that does so” (p. 2). Basque allocutive agreement marks a familiar addressee’s gender on the auxiliary (J4 (7), p. 3, from Hualde & Ortiz de Urbina 2003).
  • SOURCED Jibun, gender and formality (J5). As a first-person word, jibun is “non-standard”, and according to Kanemaru (1997) it is “considered as masculine form and a reflexive pronoun”. It is “often associated with men in sports or in militaristic groups” (§2.3.2, p. 12). Abe (2010:43) reports speakers who chose it as the most “neutral” first-person word, neither feminine like watashi/atashi nor masculine (J5 p. 12). Survey answers in J5: “I use ore in private situations, I use jibun when talking with superiors” (Male, 18) (p. 39).

H59: what the Japanese templates do with the Navajo classifier

SOURCED The Navajo facts at stake (all already on the site): anaphors “always” require “a -d or -l classifier” (Willie p. 32). The agentive passive, NX-020 niʼdilzhééh, shows “a classifier that marks the construction as intransitive” (pp. 35–36). The indefinite object of NX-501 ʼasdzą́ą́ ʼííłchxǫʼ comes with transitive ł (p. 39; review 16). Young & Morgan’s 3i subject “someone” on “work”, naʼanish, is listed with the note “See naashnish, to work, but note deletion of l-classifier” (Y&M p. 549 def 24 as given by the online dictionary interface, not checked against the printed book; indef-subject §4; cf. Willie’s NX-478 naashnish). Y&M p. 779 also gives the d-effect “in the 1st person dpl” (voice-allomorphy T94d).

  • HYPOTHESIS H59α (from the -(r)are template, broad). The d/l value of v⁰ (Voice) spells a defocused Spec,vP argument, one that is not a specific individual separate from the other participants. Contradicted by naʼanish. There an indefinite human subject, as defocused as a subject can be, comes with an l classifier removed, not with d added. That holds if the Y&M note is right; the online reading is not checked against print.
  • HYPOTHESIS H59α′ (narrowed to two-role verbs; the best new candidate). In a vP with an Agent in Spec,vP and a second participant inside VP, v⁰ is spelled d/l when the Agent is not distinct from the VP participant or not specific. In the reflexive, r = x (H62). In the reciprocal, the agent set and the patient set are the same set X (H62). In the agentive passive, the agent is u, “some human or other” (H10b). Otherwise v⁰ keeps its listed transitive value. What it gains: H59 listed two triggers and an unexplained exception. H59α′ has one condition, stated on the Agent, and the ł of NX-501 follows: the woman is a specific Agent distinct from the indefinite thing ruined, so nothing triggers d. naʼanish has one role, so α′ is silent about it, and its l-deletion stays NOT UNDERSTOOD. What it does not gain: why the spell-out is d. That is still a listed fact, and Japanese shows only that the same bundle of uses recurs on one voice morpheme, not why. LF is unchanged from review 16, step 7. Reflexive (20): ∃e[shave(e) ∧ Agent(e,y₂) ∧ Patient(e,r) ∧ r = y₂]. Passive (21): ∃e[shave(e) ∧ Agent(e,u) ∧ Patient(e,y₂)]. α′ reads the condition off Spec,vP, the Agent position under VPISH, before the default raises the Agent to Spec,TP. No new movement.
  • HYPOTHESIS The Japanese honorific route is not taken. In Japanese, the same defocusing suffix also honors the subject (J2 (12b)). If Navajo did the same, a respected Agent would bring d. The polite fourth-person agent of NX-468 John jiztał “She kicked him, John.” is printed with the same -ztał as plain NX-033 yiztał. The relative of the agentive passive, NX-315 biʼdistał=yę́ę, is printed differently. Willie segments none of the three, so this is consistent but weak (T71c). Navajo marks respect on the person prefix (ji-), not on Voice.
  • HYPOTHESIS H59β (from J1’s anaphor fact). d/l is the form v⁰ takes when its object probe meets an anaphor that it cannot agree with, as Japanese object honorific agreement cannot target zibun. Navajo anaphors “do not vary with person or number” (Willie p. 32), which is the same anaphor effect and supports H60 for anaphors. But β covers only the anaphor half. In the agentive passive the object prefix (ni- in NX-020) agrees normally, so β needs a second clause for the passive and is no better than H59. It is kept as the account of why anaphors are personless, not as the account of the classifier.

H57a: the respect/report ji- against Japanese referent honorifics

HYPOTHESIS H57a-J. The respect use of ji-/ho- is a Navajo referent honorific. It has the same profile as Japanese subject and object honorification. (i) It is agreement with an argument (J1). In Navajo it sits on a third-person noun argument, so H31 is kept (H57a). (ii) It has separate subject and object forms: Japanese o-V-ni nar- vs o-V-su- (J1 (3)), and Navajo “ji-/ho- are used to identify the subject and object respectively” (Willie p. 111, review 15; NX-468 John jiztał, NX-469 John haztał). (iii) It leaves the arguments as they are, as honorific -(r)are does (J2 p. 9). In NX-470 shimá ʼéí Baaʼ joolyé the mother is still the subject. (iv) It can sit inside an embedded clause, as referent honorifics can (J4 p. 3). The report ji- appears inside the =go clause of NX-473. J4’s proposal that such markers depend on a root-clause c (p. 2) is the same shape as H57a’s “Agree with a higher head”. Derivation of NX-470 under H57a-J: V “be called” with the name, the subject in Spec,vP, V→v→T, the subject to Spec,TP; a final head cRESP above TP (HS-1) agrees with the subject, spelled ji- in the verb word. LF: the assertion ∃s[called(s) ∧ Theme(s,my.mother)], plus, not at issue, RESPECT(speaker, my.mother) from cRESP, which c-commands TP. Scope matches c-command. This makes H57a typologically ordinary, and it counts for H57a over H57b. It is not proof: nothing in Japanese decides the Navajo case.

NONE No Navajo counterpart on our pages: an addressee-politeness marker like -mas (root-only, speaker-to-addressee). None of the Navajo respect or report forms reviewed so far is about the addressee.

Axes that give nothing for H59

  • NONE Jibun and gender. In Japanese, gender lives in the choice of first-person word (J5). The only gender-on-the-verb case met here is Basque addressee agreement (J4 (7)). No source on this site shows Navajo marking gender anywhere, so there is nothing to compare.
  • NONE Jibun as a formal or hierarchical “I”. Japanese turns its reflexive into a self-reference word for speaking up the hierarchy (J5 pp. 12, 39). No Navajo source on the site uses ʼádi- for the speaker. Navajo uses the fourth person for respect toward others (review 15). The reflexive–respect link Grant pointed to exists in both languages, but through different pieces, and neither touches the classifier.
  • HYPOTHESIS Empathy and logophoric zibun bear elsewhere. J3’s empathy treatment of direct/inverse, which names Navajo, bears on the inverse ranking (H53) and on the fourth person bound by a wisher or sayer (H51, H43), not on H59.

Predictions to test (OPEN)

  • OPEN H59α′, passive on a new verb: any two-role verb with the agentive-passive ʼadi- has d/l, including verbs whose active is Ø or ł. Test: a segmented source naming the classifier of biʼdistał (T71c).
  • OPEN H59α′ vs a morpheme fact: a two-role verb with an indefinite Agent other than ʼadi- (for example, if Y&M’s 3i ʼa- occurs as the subject of a transitive verb) should also take d/l. If it keeps Ø or ł, then α′ is wrong and d belongs to the ʼadi- morpheme.
  • OPEN Respect stays off Voice: a segmented transitive with a polite ji- Agent keeps its active classifier. A d here would make Navajo like Japanese -(r)are.
  • OPEN Anaphor with a respected binder: a reflexive whose binder is a polite ji- subject keeps ʼádi- unchanged and keeps d/l (anaphor effect + α′).
  • OPEN 1st dual-plural d-effect: α′ does not predict it, since it is a subject φ value, not distinctness. If it is the same d (T94d), α′ is incomplete.
  • OPEN H57a-J, intervention: with two internal arguments (a postposition object and a direct object), polite ho- should go to the higher one, as Japanese object honorification does (J1 *(8)).
  • OPEN H57a-J, root licensing: report ji- in an embedded clause under a plain third-person “say” should lose the report reading, if it depends on a root report head as J4 proposes for referent markers.
  • OPEN Logophoric parallel: if the wisher-bound and sayer-linked fourth persons (H51, H43) are logophoric like zibun, they should allow only de se readings (J3 ch. 5). This needs a judgment no source on the site prints.

HYPOTHESIS Standing items. The V-to-T default is untouched: no Navajo clause is re-derived except NX-470 under H57a-J, which fits it. Telicity: nothing here bears on it. Light d stays withdrawn. Light n is not touched.