‘Someone’, ‘something’ and the 3i

Legacy link recovery
Legacy section links (scroll sideways); old articles remain available.
indefinite-agent-bergative-indefinitethird-voicepassive-rankindef-personabsolutive-indefimpersonal-xindef-subjectx-on-thirdgas-subjectsun-subjectrain-fire-subjecthang-subjectlisted-dtwo-tablesnamed-unnameddriver-3ipos-labels

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 passive still merges an agent in Spec,vP, as VPISH requires for the role, but it is silent and indefinite, cannot be linked to a noun, and is existentially closed (H10b/H16a). Questions on passive verbs ask about a place or amount, never the doer: Háadi lá dibé nidahaniih? (NX-213); a passive relative picks out the patient: biʼdistał=yę́ę (NX-315). Results §12.
  • HYPOTHESIS A question word without =sh is an indefinite: John háí-shį́į́ yiztał (NX-084) (H11).
  • OPEN The rival with no agent in syntax (H10a/H16b) is not excluded, and whether the theme can raise to Spec,TP in the passive is open. The 3i forms in the legacy sections below have not yet been restated on the spine; their earlier analyses stand as contrast.

When 'someone' is the doer, the object is bi-; when 'that person' is the doer, it is silent: the agentive passive splits Hale's deictic subjects (T68b)

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

Verbs on this page → dictionary: yiztał ‘kick’ · yi-dloh ‘laugh’

Restated on the unified structure (2026-11-15). The indefinites (ʼa-, ʼadi-, areal ho-) are restated as personless pronouns in ordinary argument positions (H60): ʼadi- as the silent indefinite agent in Spec,vP that raises to Spec,TP by the default (H10b), ʼa- in VP. The 3i data on this page have not yet been individually restated. Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question.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. Status on the spine: see unified structure, §4, indefinite and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

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.

The verb 'kick' shows four subject types and three object markers on two pages of Willie 1991 SOURCED. The subject can be a plain 3rd person, an inverse, 'that person' (ji-) or 'someone' (the agentive passive ʼ-d-). The 3rd-person object marker is yi- in yiztał, bi- in biztał, Ø in jiztał and bi- in biʼdistał. 'Kiss' gives the same contrast, yidootsʼǫs vs bidiʼdootsʼǫs SOURCED. Hale 1997 says the agentive passive has "essentially identical" syntax to the 4th person, and Hale 2001 says neither prefix has a person feature SOURCED. Neither claim can tell ji- (Ø) apart from ʼ-d- (bi-). One undated Hale handout puts "indefinite" on the Ø list, and that is contradicted by Willie's forms. Hale 1997 fn. 9 gets the facts right because it lists only 1st, 2nd and 4th-person subjects for Ø. The indefinite then falls to the general b- HYPOTHESIS. This answers T68b in favour of FS-8a: the silent object needs a feature that ji- has and ʼ-d- lacks, and simply filling the deictic-subject slot is not enough HYPOTHESIS.

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 The four-way grid · 6 Three Hale statements against the grid · 7 Rulesets A and B · 8 Derivations · 9 What it means for the role schema · 10 Fidelity · 11 Open questions

Notation as on fourth-subject-p. A = the word level (prefixal, left-headed); B = the clause level (right-headed). Σ = the role schema (Dowty's proto-roles as the baseline). H = Hale's condition for yi-: a true 3rd-person subject c-commands the 3rd-person object. Merge builds structure: it puts two pieces together into a phrase. Agree copies feature values such as person and number from one piece to a probe that needs them. Person labels: 1, 2, 3; 4 = fourth person (ji- as subject); INDF = the indefinite human agent of the agentive passive (Hale's 3i with [d], written ʼ-d-; Willie's PASS, written ʼadi- on p. 36). Willie writes the object first in her glosses: "3.4.kicked" = 3rd-person object, 4th-person subject.

1. Question

T68b (from fourth-subject-p §11) asked which Hale text derives the missing yi- under ji-. Hale 1997 says the 4th person shares zero AGRs with the 3rd but is not a 3rd person "in obviation" (FS-2). Hale 2001 says the position-V prefixes have no person feature (FS-4). That gives two candidate rules. FS-8a: the subject carries a feature [4] that a "true 3rd" lacks. FS-8b: the 4th-person subject is outside the animacy ranking. The 4th person alone cannot separate them. This page adds the one other position-V subject that takes an object, the indefinite agent of the agentive passive. Hale 1997 says it is built "in the same manner as the fourth person". If FS-8a or FS-8b were simply about "a filled position V", ʼ-d- should also give Ø. It does not. The page also bears on the older T17i (pleonastic-psych): is ʼadi- another "exit" like the 4th person?

2. Sources

SOURCED Willie, MaryAnn 1991. Navajo Pronouns and Obviation. PhD dissertation, University of Arizona. Box scan; for chs. 4–6 the PDF page is the printed page + 6. Page images read this run: p. 36–38 (§2.3.1, the agentive passive), p. 115 (11)/(12), p. 152 (1b–f), p. 156 (7)–(9), p. 157 (10). jiztał is taken from p. 129 (30b) and p. 154 (3) as already recorded on fourth-subject-p and relatives (p. 154 rechecked in the text layer only this run).

SOURCED Hale, Ken 1997. "Remarks on the syntax of the Navajo verb" (hale023), Part I. PDF pp. 16–17 (main text, fn. 9, and the Speas/Jelinek remark; text layer). PDF pp. 28–29 (page images): tree (19), (20) and the agentive-passive paragraphs. The PDF's font substitutes "~" for the glottal stop and "¬" for ł in Navajo words, so Hale's (20) is described here, not copied.

SOURCED Hale, Ken (undated handout, hale062). Navajo verb handout, p. 9 (page image), (22) "Third Person". It cites McDonough 2000 and Rice 2000, so it is from 2000 or later.

SOURCED Hale, Ken 2001. Linguistic Inquiry 32: 678–693, §4 pp. 689–691 (text): (22a–c), and the remarks on 3a/3i/3s, citing Rice 2000 and Neundorf 2000 (fn. 6).

Not opened: Young & Morgan 1987 (the source of Hale's 3i label, pp. 67–69); Neundorf 2000; Rice 2000; Kari 1976.

3. Rules first

IX-1  "The prefix b- is in fact the general third person object agreement              SOURCED (Hale 1997 PDF p. 16, text)
      marker ... In all other situations, b- is the agreement marker which
      appears, where overt."  Fn. 9: Ø/b- for a 3rd object "in the c-command
      domain of a first, second, or fourth person subject" (= FS-1).
      The list has no indefinite.
IX-2  "Our analysis of the agentive passive is essentially identical to that           SOURCED (Hale 1997 PDF p. 29, image)
      sketched above for the fourth person, in its use as a human impersonal
      subject." The d-structure subject position "is occupied by the agentive
      passive subject morphology ~-d-"; it raises to Spec,EP; "~-d- is prefixed
      to E ... in the same manner as the fourth person impersonal j-."
IX-3  The agentive passive is "not actually a passive in the full sense, since         SOURCED (Hale 1997 PDF p. 29, image)
      the lexical object remains an object at s-structure, entering into the
      object agreement relation in the usual way."
IX-4  Handout (22): "Third person object (of verb) is also Ø if the subject is         SOURCED (hale062 p. 9, image)
      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)."
IX-5  The indefinite agent "must be understood as human, but no noun that              SOURCED (Willie p. 36, image)
      describes this agent may be adjoined."
IX-6  3rd-person object of the verb (AGRo in E), revised:                              HYPOTHESIS
        yi-  if H holds (true 3rd-person subject);
        Ø    if the subject bears [4] or a 1st/2nd person value
             (the class Willie p. 122 calls "definite, human");
        b-   elsewhere, including the INDF subject ʼ-d- (IX-1 elsewhere case).
      The Ø condition is stated over a feature of the subject DP, not over
      "position V is filled". This is FS-8a.
IX-7  Rejected: "Ø if position V (the deictic-subject slot) is filled".               HYPOTHESIS (falsified by (4), (6))
      Also rejected: IX-4 as written, for transitive ʼ-d-.
IX-8  Alternative for the bi- of (4)/(6): it is the inverse-type bi- (object          HYPOTHESIS (competing with IX-6)
      raised above an agent that is lower in prominence), not the plain elsewhere
      case. Hale (PDF p. 17) reports the Speas/Jelinek suggestion that the
      Inverse "involves topicalization of the object, removing it from the
      domain in which obviation applies, forcing use of the general third
      person object prefix b-" (quotation SOURCED; applying it to ʼ-d- is ours).

IX-6 and IX-8 make the same prediction for the four forms in §5. They come apart on the P object (T69a) and on scope and word order with overt nouns (T69c).

4. The data

Forms are given as printed. Lines marked "source" are Willie's own glosses. The Leipzig lines are ours, built only from those glosses and the translations. Verb words are not segmented beyond what Willie's glosses give.

Willie 1991 p. 152 (1c), (1d), (1f) (page image): 'kick' relativized
(1)yiztał=yę́ę
source: 3.3.kicked=REL
3O.3S.kick.PFV=REL
'the one he kicked' (1c)
(2)biztał=yę́ę
source: INV 3.3.kicked=REL
INV.3O.3S.kick.PFV=REL
'the one who kicked him' (1d)
(4)biʼdistał=yę́ę
source: PASS 3.3.kicked
3O.INDF.A.kick.PFV=REL
'the one who was kicked' (1f), labelled "(subject of agentive passive)"

Willie glosses (4) as "PASS 3.3", with no REL in the gloss line. Our INDF.A follows her PASS and her p. 36 description, not a segmentation.

Willie 1991 p. 129 (30b), p. 154 (3): the 4th-person subject (as recorded on fourth-subject-p and relatives)
(3)jiztał=yę́ę
source: 3.4.kicked=REL
3O(Ø).4S.kick.PFV=REL
'the one that person kicked' (p. 154 (3)); non-relative jiztał 'That person kicked it.' (p. 129 (30b))
Willie 1991 p. 156 (8), (9); p. 157 (10) (page images): 'kiss'
(5)ʼatʼéédyidootsʼǫs
source: girl 3.3.will kiss
girl3O.3S.kiss.FUT
'He will kiss the girl.' (9)
(5′)yidloh=yę́ębidootsʼǫs
source: 3.laughing=REL INV 3.3.will kiss
3S.laugh=RELINV.3O.3S.kiss.FUT
'The one who was laughing will kiss him.' (8)
(6)yidloh=yę́ębidiʼdootsʼǫs
source: 3.laughing=REL PASS.3.3.will kiss
3S.laugh=REL3O.INDF.A.kiss.FUT
'The [one] who was laughing will be kissed.' (10) ("The who was laughing" as printed [sic].) Willie introduces it: "A relative may be adjoined to the subject of an agentive passive."
Willie 1991 pp. 36–37 (page images): agentive passives with a 1st/2nd-person object
(7)niʼdilzhééh
source: 2s.PASS.shaving
2SG.O.INDF.A.shave.IPFV
'You are being shaved.' (21)
(8)shiʼdidleesh
source: 1s.PASS.painting
1SG.O.INDF.A.paint.IPFV
'I am being painted.' (23)

Willie: ʼadi- "occurs to the right of (internal to) the object pronoun", and "the patient precedes the agent in the prefix array" (pp. 36–37). So in (4) and (6) the bi- sits in the object slot, before the indefinite agent, like ni- and shi- in (7) and (8).

Willie 1991 p. 115 (12) (page image): a passive-translated 3rd-person object, unlabelled
(9)nábiʼdiiʼnil
source: 3.picked up(PL)
3O.pick.up(PL)
'[a boy and girl] were picked up' (inside (12), full sentence on fourth-subject-p (2))

Willie does not call (9) an agentive passive. Its translation is passive, and it has b- followed by a glottal stop and d, like (4) and (6). Reading it as a third bi- + ʼ-d- token is ours HYPOTHESIS (T69d). Its partner (11) náhodiiʼnil "4s.picked up (PL)" has a 4th-person object under the same unexpressed agent.

5. The four-way grid

Subject of 'kick' / 'kiss'Position V filled?Definite human (Willie p. 122)?True 3rd (H)?Verb's 3rd objectForms
plain 3rd, directnono (3rd, on the ranking)yesyi-(1) yiztał, (5) yidootsʼǫs
plain 3rd, inversenonoyes, but the object outranks itbi-(2) biztał, (5′) bidootsʼǫs
4th ji-yesyesnoØ(3) jiztał
INDF ʼ-d-yeshuman, not definite (IX-5)nobi-(4) biʼdistał, (6) bidiʼdootsʼǫs

The table cells are SOURCED, except the "definite" column for INDF, which is our inference from IX-5 and from the meaning 'someone' HYPOTHESIS. Rows 3 and 4 have the same value in the two columns that Hale's 1997 syntax (IX-2) and Hale 2001 (FS-4) can see: position V is filled, and the subject is not a true 3rd. They differ only in the column Willie uses for the 4th person, and in the result. Across the 2 verbs there are 3 pairs in which only the subject changes: (1)/(4) and (3)/(4) for 'kick', and (5)/(6) for 'kiss'. All three are consistent with this.

6. Three Hale statements against the grid

  • Handout (22) (IX-4) lists "indefinite" with 1, 2 and 4 as a Ø trigger. That predicts *Ø-ʼdistał-type forms. Willie prints biʼdistał and bidiʼdootsʼǫs SOURCED. Either (22) means the intransitive 3i ʼ-, which has no object to silence (Hale 1997 (21): ʼ- is the "subject of an intransitive"), or it is a slip. It cannot be right for the transitive ʼ-d- HYPOTHESIS (T69b).
  • Hale 2001 (FS-4): 3a and 3i are both "construed with the third person", and neither has a person feature SOURCED. On that reading the two should behave alike, and they do not. So Hale 2001 needs an added distinction between 3a and 3i that §4 does not state HYPOTHESIS.
  • Hale 1997 (IX-1, IX-2): the syntax is "essentially identical", and fn. 9's list has 1, 2 and 4 only. Since the syntax is the same, the difference must come from the subject's features, and fn. 9 lists exactly the features that give Ø. The indefinite is not on the list and fails H, so the main text's elsewhere b- applies. This derives all four rows HYPOTHESIS. We do not know whether Hale meant fn. 9's list to exclude the indefinite. He does not discuss a 3rd-person object under ʼ-d-: his (20) has a 1st-person object SOURCED.

Answer to T68b. FS-8a (a subject feature [4] that sorts with 1 and 2) derives the grid. FS-8c ("Ø whenever position V is filled", the literal reading of the handout and of Hale 2001) is falsified by (4) and (6). FS-8b (Ø because the subject is outside the animacy ranking) is not tested directly, because we have no source saying whether the indefinite agent is on the ranking. If it is off the ranking, as T17i suggested ("takes the agent out of ranking"), FS-8b wrongly predicts Ø for (4). So FS-8b survives only if the indefinite is on the ranking, which is OPEN (T69e) HYPOTHESIS. On T17i: ʼadi- is not an "exit" like the 4th person, at least for the verb's own object HYPOTHESIS.

Partial answer to T68d (a real Ø, or a yi- absorbed by ji-?). Both deictic subjects attach in the same place, prefixed to E (IX-2), and with ʼ-d- an object prefix surfaces before them (biʼd-, shiʼd-, niʼd-). Hale 2001 (22a) shijihozh 'He/She/They/One is tickling me' shows an object prefix before j- as well SOURCED. So nothing bans an object prefix in front of the deictic slot. An absorption account would need a special rule deleting yi- just before j-. This is weak support for a true Ø HYPOTHESIS. A rule of that kind could still exist in Young & Morgan, which we have not opened OPEN.

7. Rulesets A and B

StepWhereRulesetWhat happens
1VPB (head-right VP)Merge the object pro(3) with the root ('kick', 'kiss').
2vPBMerge the subject in vP (VPISH): pro(3), DP[4] (j-) or the INDF morpheme ʼ-d- itself (Hale: it "is occupied by the agentive passive subject morphology").
3MAAgree M ← subject: zero AGRs in all four rows (3rd-person construal, Hale 2001).
4EAAgree E (AGRo) ← object [3]; the object raises to Spec,MP (IX-2, IX-3).
5EPAThe subject raises to Spec,EP (EPP). j- / ʼ-d- is prefixed to E (morphological Move).
6WordASpell-out of AGRo by IX-6: yi- / Ø / b-, reading the subject's features.
7Word edgeA→B changeoverThe finished word is ordered last in its clause. A relative clause such as yidloh=yę́ę is linked to the subject and comes before the verb (B).

Only step 6 differs between rows 3 and 4. It is a word-level (A) spell-out decision, and it reads a feature that came into the structure in B at step 2. So the subject's feature has to stay visible across the hypothesized changeover at the word (not established). This is the same thing fourth-subject-p §7 found for the P object, now with the cut drawn inside the set of deictic subjects HYPOTHESIS.

8. Derivations

D1. (3) jiztał 'That person kicked it.'

1  Merge  [VP pro(3)  kick ]                            B; head-right VP
2  Merge  [vP DP[4,hum,def] [VP ...] v ]                 VPISH; DP linked to j- (Hale 1997 p. 26)
3  Merge  M; Agree M ← subject: zero AGRs               A
4  Merge  E; Agree E(AGRo) ← [3] from the object         A; object raises to Spec,MP
5  Move   subject to Spec,EP; j- prefixed to E           A (morphological Move, Hale p. 28)
6  Spell  AGRo[3]: subject has [4] → Ø  (IX-6 / FS-1)     A
7  Output jiztał                                         not segmented further

D2. (4) biʼdistał(=yę́ę) 'the one who was kicked'

1  Merge  [VP pro(3)  kick ]                            B
2  Merge  [vP ʼ-d-[hum] [VP ...] v ]                     VPISH; Hale: d-structure subject = ʼ-d- itself;
                                                          no noun may be linked to it (IX-5)
3  Merge  M; Agree M ← subject: zero AGRs               A
4  Merge  E; Agree E(AGRo) ← [3]; object to Spec,MP      A; "object agreement ... in the usual way" (IX-3)
5  Move   ʼ-d- to Spec,EP; prefixed to E                 A; same as D1 step 5 (IX-2)
6  Spell  AGRo[3]: H? no (not a true 3rd);               A
          [4]/1/2? no  → elsewhere b-  (IX-1, IX-6)
7  Output biʼdistał; =yę́ę binds the object (Willie:     B for =yę́ę at the clause edge
          "subject of agentive passive" in English)

D3. (5) ʼatʼééd yidootsʼǫs vs (6) yidloh=yę́ę bidiʼdootsʼǫs

1  Merge  [VP pro(3)  kiss ]                            B
2  Merge  [vP SUBJ [VP ...] v ]                          (5) SUBJ = pro(3); (6) SUBJ = ʼ-d-
3  Agree  M ← SUBJ: zero; E(AGRo) ← [3]                  A
4  Move   (6): ʼ-d- prefixed to E                        A
5  Spell  AGRo: (5) H holds → yi-; (6) → b-              A (IX-6)
6  Adjoin (5) ʼatʼééd linked to the object pro;          B; Willie p. 156: noun and relative are "both
          (6) yidloh=yę́ę linked to the object pro         referring expressions", adjoined alike
7  Order  [nominal] [verb word]                          B; head-final

In (5) and (6) the linked nominal is the kissed one in both (the girl; the one who was laughing). What changes is the doer: 'he' vs 'someone'. The object marker follows the doer: yi- vs bi- SOURCED (forms and translations). The rule IX-6 that relates them is our HYPOTHESIS.

9. What it means for the role schema (Focus 2)

In yiztał, jiztał and biʼdistał the role schema Σ is the same. A human kicker is the proto-agent (Willie: the indefinite agent "must be understood as human"), and the thing kicked is the proto-patient HYPOTHESIS (role assignment is ours). Hale stresses that the agentive passive is "not actually a passive". The object stays an object and the clause stays transitive (IX-3) SOURCED. So the roles do not change. What changes is a reference property of the agent: definite person ([4]), indefinite 'someone', or a plain 3rd person. That property alone selects Ø, bi- or yi- on the patient's slot HYPOTHESIS. This sharpens the point made on fourth-subject-p §9. Even with role and position (position V) held constant, the spell-out depends on a feature of the item filling the role, and a pure role schema cannot predict the verb word. This is one more point against the strong thesis that Navajo verbs take theta roles instead of lexical items. It does not rule out a weaker version in which roles are assigned schematically and features are added at spell-out. See the verb guide and the proto-role schema.

10. Fidelity

  • All Willie forms used here except (3) were read on page images this run: pp. 36–37, 115, 152, 156, 157. (3) is reused from earlier image-checked pages. The text layer garbles ł and nasal hooks, so no form was copied from it.
  • Willie glosses (4) "PASS 3.3.kicked" and (6) "PASS.3.3.will kiss". The second "3" is her label for the agent. Our INDF.A replaces it following p. 36, and Hale (3i) [d].
  • Hale 1997's Navajo forms in the PDF are in a substituted font ("~" for the glottal stop, "¬" for ł, "²" for a nasal vowel), so (20) is not copied here.
  • Willie p. 157 (10)'s translation, "The who was laughing", is missing "one" [sic].
  • Reading (9) nábiʼdiiʼnil as an agentive passive is ours. Willie only glosses "3.picked up(PL)".
  • IX-6 to IX-8, the "definite" value for INDF, the role assignments and the derivations are ours. IX-1 to IX-5 are quotations.

11. Open questions

IDQuestionWhat would settle it
T69aA postposition object under an agentive-passive subject. IX-6 (elsewhere) and IX-8 (inverse-like) both predict b- on P, and H fails either way. Is there a sourced token, and does the verb's object then still show bi-?Willie chs. 2, 4; Young & Morgan 1987 agentive-passive paradigms. OPEN
T69bWhat does handout (22)'s "indefinite" mean: the intransitive 3i ʼ-, the indefinite object, or a slip? Does any transitive ʼ-d- form with a 3rd-person object show Ø?Young & Morgan 1987:67–69 (Hale's 3i source); Neundorf 2000. OPEN
T69cIs the bi- of biʼdistał the elsewhere b- (IX-6) or an inverse-type bi- (IX-8)? Test: does an agentive passive have the scope and word-order effects of the inverse (number-scope, two-osv)?Sourced agentive passives with overt nouns or da-. OPEN Update 2026-10-28 (12:46pm PT): partly answered on ergative-indefinite: Willie's "Ergative" (internal agent) vs Hale's ergative (subject) maps onto IX-8 vs IX-6; the deciding test is T70a (HYPOTHESIS).
T69dDo Willie p. 115 (11) náhodiiʼnil and (12) nábiʼdiiʼnil contain ʼ-d-? If so, (11) is a sourced 4th-person object under an indefinite agent, which bears on T68e.Young & Morgan 1987 entry for the 'pick up (plural)' verb. OPEN
T69eIs the indefinite agent on the animacy ranking? If not, FS-8b is falsified by (4); if so, FS-8b survives alongside FS-8a.Willie ch. 4 on the ranking; Hale & Platero 1996. OPEN Update 2026-10-28 (12:46pm PT): answered on ergative-indefinite: on Willie's criteria the indefinite agent is on the ranking, so FS-8b survives, and it is coextensive with FS-8a (HYPOTHESIS).
T69fIs the Ø trigger the morpheme j- or definiteness? The test is the impersonal 'one' use of ji- (Hale's "human impersonal subject", e.g. his (16) ʼOne looks at meʼ) with a 3rd-person object: Ø (morpheme) or bi- (definiteness)?A sourced impersonal ji- form with a 3rd-person object (Willie ch. 4; Y&M). OPEN
T69gAlgonquian: what object agreement do unspecified-actor ("X") forms show, and do they pattern like ʼ-d- (overt) or like ji- (silent)? This merges with T68f.Oxford's Algonquian chapter. OPEN Update 2026-10-28 (12:46pm PT): answered on ergative-indefinite: X is ranked 1/2>X>3, and X>1 shows the inverse theme sign with an overt object, like ʼ-d- (Meskwaki SOURCED via Oxford; the comparison is a HYPOTHESIS).

Sources: Willie 1991 pp. 36–37, 115, 152, 156, 157 (page images), pp. 129, 154 (as recorded earlier); Hale 1997 (hale023) PDF pp. 16–17 (text) and pp. 28–29 (images); Hale handout hale062 p. 9 (image); Hale 2001 LI 32 pp. 689–691. See also fourth-subject-p, relatives, hale-general-b, pleonastic-psych, subject on the head, open questions, sources.

Willie and Hale both call the 'someone' doer "ergative", but they mean different places: the agentive passive, the animacy ranking and the Algonquian impersonal (T69e, T69c, T69g)

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

Verbs on this page → dictionary: yiztał ‘kick’ · yishxash ‘bite’

Restated on the unified structure (2026-11-15). The indefinites (ʼa-, ʼadi-, areal ho-) are restated as personless pronouns in ordinary argument positions (H60): ʼadi- as the silent indefinite agent in Spec,vP that raises to Spec,TP by the default (H10b), ʼa- in VP. The 3i data on this page have not yet been individually restated. 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, §4, indefinite and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

Willie 1991 calls the agentive-passive prefix ʼadi- "an indefinite Ergative pronoun" (p. 221), and she gives the inverse's agent the same case, "Ergative" (pp. 72, 100) SOURCED. Hale 1997 also calls the d of ʼ-d- ergative, but for him it is the case of an ordinary transitive subject in the ordinary subject position (PDF p. 30, (21)) SOURCED. So the same word points to two different structures. Willie's version fits the inverse analysis of biʼdistał (IX-8), and Hale's fits the elsewhere-b- analysis (IX-6) HYPOTHESIS. On T69e: Willie's stated criteria put the indefinite agent on the animacy ranking. It is human but not definite, and she ties the ranking to third persons, the only ones that can be indefinite (pp. 86, 122) HYPOTHESIS. FS-8b therefore survives, and on all sourced data it picks out the same class as FS-8a. On T69g: in Algonquian the impersonal actor X is ranked on the person hierarchy (1/2 > X > 3), and 'people see me' carries the inverse theme sign with overt 1st-person agreement (Oxford, Meskwaki (17c)) SOURCED. That patterns with Navajo ʼ-d- (overt object), not with ji- (silent object) HYPOTHESIS.

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 Two "ergatives" · 6 Is 'someone' on the ranking? (T69e) · 7 Algonquian X (T69g) · 8 Rulesets A and B · 9 Derivations · 10 What it means for the role schema · 11 Fidelity · 12 Open questions

Notation as on indefinite-agent-b and fourth-subject-p. A = the word level (prefixal, left-headed); B = the clause level (right-headed). Σ = the role schema (Dowty's proto-roles as the baseline). Merge builds structure: it puts two pieces together into a phrase. Agree copies feature values such as person and number from one piece to a probe that needs them. Move re-merges a piece that is already in the structure higher up. INDF = the indefinite human agent of the agentive passive (Willie's ʼadi-, Hale's ʼ-d-). EXT/INT = Willie's "external"/"internal" argument. ERGW = Willie's Ergative (the case of an internal agent); ERGH = Hale's ergative (the case of the d-structure subject). Willie writes the object first in her glosses ("3.4.bit" = 3rd-person object, 4th-person subject).

1. Question

The last run (indefinite-agent-b) left three linked items. T69e: is the indefinite agent on the animacy ranking? If it is not, FS-8b ("Ø because the subject is outside the ranking") wrongly predicts a silent object in biʼdistał. T69c: is that bi- the elsewhere b- (IX-6) or an inverse-type bi- (IX-8)? T69g: what does Algonquian do with an unspecified actor? This page answers T69e and T69g from sources on the box. It gives a framework-level partial answer to T69c and states the one data test that would decide it.

2. Sources

SOURCED Willie, MaryAnn 1991. Navajo Pronouns and Obviation. PhD dissertation, University of Arizona. Page images read this run (PDF = printed + 6 from Ch. 3 on): p. 72, pp. 76–77 (42)–(51), p. 86 (70), p. 87 (71)–(72), p. 100 (§3.3.5), p. 124 (23), p. 126 (26), p. 221. Text layer only (finding aid): pp. 36–37, 50–51, 118–122, 152, 157. biʼdistał is reused from the p. 152 image reading recorded on indefinite-agent-b.

SOURCED Hale, Ken 1997. "Remarks on the syntax of the Navajo verb" (hale023), Part I. PDF p. 30 (page image): the d-component paragraph and (21) "The ergative patterning of 3i morphology". PDF pp. 26–27 (text layer): fn. 16 on the two uses of the 4th person.

SOURCED Oxford, Will. "Algonquian", to appear in the Routledge Handbook of North American Languages (box PDF). PDF p. 7 fn. 2 (X = impersonal actor), p. 8 (4) (Passamaquoddy, LeSourd 2004: 261), p. 17 (Ojibwe theme sign and hierarchy, Rhodes 1994: 432), p. 18 (17) (Meskwaki, Goddard 1994b: 191; page image for (17)).

Not opened: Young & Morgan 1987; Hale & Platero 1996; Goddard 1994b; LeSourd 2004; Rhodes 1994; Bittner 1994.

3. Rules first

EI-1  'adi- is "an indefinite Ergative pronoun ... that appears in the               SOURCED (Willie p. 221, image)
      Agentive Passive which can be used only in reference to humans"; 'a is
      "an indefinite Absolutive pronoun".
EI-2  Inverse: "the agent is the 'internal' argument, i.e., not the subject,          SOURCED (Willie p. 72, image)
      I assume that the agent pronoun has Ergative case, and the patient
      pronoun has Accusative case". p. 100: the Inverse "makes the agent an           SOURCED (Willie p. 100, image)
      internal argument, with Ergative case, and places an argument with an
      'internal' theta role in the external argument position, with Absolutive
      case." (Accusative p. 72 vs Absolutive p. 100: the source disagrees.)
EI-3  "with the Inverse, the external argument (the patient, with Absolutive          SOURCED (Willie p. 77, image)
      case) must be definite"; the internal agent may be indefinite ((46)).
EI-4  (70) Agent higher → Direct; patient higher → Inverse; equal → discourse.         SOURCED (Willie p. 86, image)
      "The voice alternation can only be used if all arguments are third
      person. This clearly is connected with the fact that only third person
      arguments can be indefinite; first and second person arguments are
      always definite."
EI-5  "Fourth person, like first and second, is definite, human, and not             SOURCED (Willie p. 122; = FS-5)
      involved in the animacy hierarchy." It "does not require a discourse            SOURCED (Willie p. 124, image)
      antecedent -- it is simply deictic", and it is used for a human "who is
      either unknown, or someone whom the speaker does not wish to identify".
EI-6  The agentive-passive morphology "is assigned to the transitive subject,         SOURCED (Hale 1997 PDF p. 30, image)
      i.e., the argument occupying the external argument (subject) position at
      d-structure ... the position in which the external argument of a
      transitive verb would be assigned ergative case, if Navajo were ergative."
      (21): '- = subject of intransitive, object of transitive; '-d- = subject
      of transitive. "the d-element is a realization of the ergative case".
EI-7  ERGW ≠ ERGH. ERGW marks an agent that is NOT in the subject (external)           HYPOTHESIS (our comparison of EI-1/2 with EI-6)
      position (inverse agent; 'adi-). ERGH marks the agent that IS in the
      subject position (every transitive subject, abstractly).
EI-8  Willie-reading of biʼdistał: EXT = 3rd patient (bi-), definite;                 HYPOTHESIS (built from EI-1–EI-3, IX-8)
      INT = INDF agent, ERGW. Point for point this is the inverse template of
      EI-2/EI-3. So the bi- is the inverse bi- (= IX-8).
EI-9  Hale-reading of biʼdistał: ʼ-d- is the subject (EI-6, IX-2); ERGH is            HYPOTHESIS (= IX-6)
      just the case value of that subject. The object fails H and is not
      under a [4]/1/2 subject → elsewhere b-.
EI-10 Ranking membership (answers T69e): an argument is off the ranking iff           HYPOTHESIS (from EI-4, EI-5, EI-1)
      it is definite AND human (EI-5's criterion). INDF is human (EI-1) and
      indefinite (EI-1), so it is on the ranking. Willie's own gloss "PASS 3.3"
      (p. 152) also files its agent as a "3", and EI-4 ties indefiniteness to
      the third person.

EI-8 and EI-9 make one different prediction, stated as a schema with no Navajo forms. Willie (72b.4) rules out "*it was bitten by him": a bi- verb cannot have an agent that ranks higher than its patient SOURCED. The INDF agent is human (EI-1) and on the ranking (EI-10). So under EI-8 a bi- + INDF verb with a non-human 3rd-person patient ('someone kicked it') should be as bad as (72b.4). Under EI-9 nothing rules it out HYPOTHESIS. This is T70a.

4. The data

Forms are given as printed and were read on page images this run, except (6). Lines marked "source" give Willie's own glosses. The Leipzig lines are ours, built only from those glosses and translations.

Willie 1991 pp. 76–77 (45)–(47) (page images): definiteness in the inverse
(1)ʼashkiiʼatʼéédbiztsʼǫs
boygirlINV.3O.3S.kiss.PFV
'The boy was kissed by the girl.' (45)
(2)ʼashkiiʼatʼééd léiʼbiztsʼǫs
boygirl INDFINV.3O.3S.kiss.PFV
'The boy was kissed by a girl.' (46)
(3)*ʼashkii léiʼʼatʼéédbiztsʼǫs
boy INDFgirlINV.3O.3S.kiss.PFV
[A boy was kissed by the girl.] (47), starred in the source

No gloss lines in the source for (45)–(47); our glosses follow Willie's p. 76 remark that a noun is definite "unless it is specifically marked indefinite with such a word as léiʼ" and her INV label for bi- forms. Willie: in the inverse the noun next to the verb is the internal argument (p. 72), so in (2) the indefinite noun is the agent (INT) and in (3) the indefinite noun is the patient (EXT).

Willie 1991 p. 87 (71)–(72) (page image): the ranking without nouns
(4)yishxash
source: 3.3.bit
3O.3S.bite.PFV
(72a) 1 'he bit him' (equal ranked); 2 'he bit it' (pat. lower ranked); 3 'it bit it' (lower equals); 4 *'it bit him' (agent lower)
(5)bishxash
source: INV3.3.bit
INV.3O.3S.bite.PFV
(72b) 1 'he was bitten by him' (equal ranked); 2 'he was bitten by it' (patient higher); 3 'it was bitten by it' (lower equals); 4 *'it was bitten by him' (higher agent)
Willie 1991 p. 152 (1f) (as read on indefinite-agent-b); p. 124 (23b) (page image)
(6)biʼdistał=yę́ę
source: PASS 3.3.kicked
3O.INDF.A.kick.PFV=REL
'the one who was kicked', labelled "(subject of agentive passive)"
(7)jishxash
source: 3.4.bit
3O(Ø).4S.bite.PFV
'That person bit it.' (23b)

(7) is a 4th-person subject with a non-human ('it') 3rd-person object and no yi-, the same pattern as jiztał. Willie labels the patient of (6) the "subject" of the agentive passive, which is her term for the external argument.

Oxford, "Algonquian", PDF p. 18 (17) (page image), Meskwaki (Goddard 1994b: 191); glosses as printed
(8a)ne-wa·pam-ekw-a
1-see-INV-3.AN.SG
's/he sees me' (3>1)
(8b)ne-wa·pam-ekwi
1-see-INV.INAN
'it sees me' (0>1)
(8c)ne-wa·pam-eko·-pi
1-see-INV-X
'people see me' (X>1)

Oxford: TA verbs distinguish animate third-person actors, inanimate actors (0) and "impersonal actors (notated X and functionally equivalent to a passive)". Independent-order hierarchy for the theme sign: "1/2>X>3>3’>0" (PDF p. 17, after Rhodes 1994: 432, stated for Ojibwe). Applying the Ojibwe hierarchy to the Meskwaki forms is our reading HYPOTHESIS.

5. Two "ergatives" (T69c, framework level)

Willie 1991Hale 1997
What "ergative" marksAn agent that is internal, i.e. not the subject: the inverse agent (EI-2) and ʼadi- (EI-1)The agent in the d-structure subject position of any transitive; overt only as the d of ʼ-d- (EI-6)
Where the INDF agent sitsInternal (from "Ergative", our inference)Subject position, then prefixed to E like j- (EI-6, IX-2)
What the patient of biʼdistał is"subject of agentive passive" (p. 152), i.e. EXTObject, "entering into the object agreement relation in the usual way" (IX-3)
Which bi-Inverse-type (EI-8 = IX-8)Elsewhere b- (EI-9 = IX-6)

The first three rows are SOURCED, except the "our inference" cell. The last row is HYPOTHESIS. Neither author draws it.

A check against EI-8. Willie also uses ʼadi- with 1st- and 2nd-person objects (niʼdilzhééh 'You are being shaved', p. 37 (21), on indefinite-agent-b). By EI-4, voice alternation needs all arguments to be third person, so those forms cannot be inverse in Willie's own sense. Yet ʼadi- is still "Ergative" there (EI-1 makes no exception). So for ʼadi-, ERGW is a fixed property of the morpheme, not a sign that the inverse applied HYPOTHESIS. That weakens EI-8. "Ergative" in Willie does not by itself show inverse syntax. It only shows that the agent is not the external argument. That is still not Hale's picture, where the agent is the subject HYPOTHESIS.

Status of T69c. The two sources disagree on the INDF agent's position, and the disagreement maps onto IX-6 vs IX-8. Neither source has an agentive passive with nouns on both sides, da-, or a non-human patient. So the scope and word-order test and the T70a test are still OPEN OPEN.

6. Is 'someone' on the ranking? (T69e)

  1. Willie's criterion for staying out of the hierarchy is a pair of features, "definite, human" (EI-5) SOURCED. INDF has the second, "used only in reference to humans" (EI-1), but not the first: it is "indefinite" (EI-1), and no noun may be linked to it (Willie p. 36) SOURCED. So INDF fails her criterion HYPOTHESIS.
  2. Willie ties taking part in the ranking to being third person, "connected with the fact that only third person arguments can be indefinite" (EI-4) SOURCED. An indefinite argument is therefore a third-person-type argument in her system, and her gloss "PASS 3.3" for (6) agrees HYPOTHESIS. She also groups "lower" with "indefinite" when she explains her translations: the internal argument "is taken to be a lower or indefinite being" (p. 88, text layer) SOURCED.
  3. Answer: INDF is on the ranking (EI-10) HYPOTHESIS. Willie never says so in so many words. The answer follows from her own statements, not from a sentence judged for this purpose.
  4. Consequence for FS-8b. FS-8b says Ø because the subject is outside the ranking. With INDF on the ranking, FS-8b predicts no Ø in (6), which is correct. FS-8b is not falsified HYPOTHESIS.
  5. Consequence for FS-8a vs FS-8b. Willie's off-ranking class is {1, 2, 4}, "definite, human". FS-8a's Ø trigger is a subject with [4] or 1/2 (IX-6). These are the same three. On every sourced form on the site the two rules pick out the same subjects. In practice they are one generalization written two ways, a feature written into the syntax (FS-8a) or exemption from the ranking (FS-8b) HYPOTHESIS. Something would separate them only if one person category were definite and human without being exempt from the ranking, or the reverse. T70d names the one candidate on the box, the narrative 4th person with an overt noun (Hale fn. 16; Willie p. 117 (14)).

T69f, refined. Willie's "definite" for the 4th person does not mean "known to the speaker". The 4th is used for a human "who is either unknown, or someone whom the speaker does not wish to identify", because it is "simply deictic" (EI-5) SOURCED. So T69f's contrast is between deictic and non-deictic, not between known and unknown HYPOTHESIS. Willie's impersonal examples (pp. 118–119 (15), (16), text layer) are intransitive or have a 4th-person P object. No sourced impersonal ji- with a 3rd-person object is on the box, so T69f stays OPEN.

7. Algonquian X (T69g)

  • X is a person category on the theme-sign hierarchy, ranked between the speech-act persons and the third person: "1/2>X>3>3’>0" (Ojibwe independent order) SOURCED (Oxford PDF p. 17).
  • When X acts on 1, the verb takes the inverse theme sign and the 1st-person object is overt (ne-): (8c) 'people see me' SOURCED (Meskwaki via Oxford (17c)).
  • So in Algonquian the unspecified human actor is on the ranking, as EI-10 proposes for Navajo INDF. The object is indexed overtly, as with Navajo ʼ-d- (shiʼdidleesh, biʼdistał) and unlike ji- (Ø). Answer to T69g: X patterns like ʼ-d- HYPOTHESIS. The comparison is ours, and the two hierarchies differ: Navajo's is an animacy scale over third persons only (EI-4), Algonquian's is a person scale.
  • A warning for EI-8. Under the Algonquian ranking, X outranks 3, so an X>3 form would be direct, not inverse. Oxford gives no X>3 form, so this is not checked. If Navajo worked like Algonquian here, biʼdistał (INDF>3) would not be an inverse, which goes against EI-8 HYPOTHESIS. On Willie's scale, by contrast, a human INDF ties with a human patient ((72b.1) 'equal ranked'), and a tie lets either voice be used. The Navajo and Algonquian rankings give different predictions here, which is T70e.

8. Rulesets A and B

StepWhereRulesetEI-8 (Willie)EI-9 (Hale)
1VPB (head-right)Merge the patient pro(3) with the root 'kick'.
2vPBMerge INDF in Spec,vP (VPISH: every agent starts inside vP).
3Case/positionBINDF stays internal (ERGW); the patient moves to the external position (Move)INDF moves to the subject position (Move); ERGH = its case
4M, EAAgree: M ← subject (zero AGRs); E (AGRo) ← patient [3].
5EAʼ-d- prefixed to E (morphological Move, Hale).
6WordAAGRo spelled out as inverse bi- (patient above agent)AGRo spelled out as elsewhere b- (H fails; no [4]/1/2)
7Word edgeA→BThe word is ordered last; =yę́ę and any nominal are placed by B.

The two readings differ only at step 3, which is a clause-level (B) choice about position. Both give the same word at step 6 HYPOTHESIS. That is why the word alone cannot decide T69c. The test has to look at clause-level effects: animacy (T70a), scope and word order (T69c). This fits the site's working picture, where the word is where ruleset B's decisions are cashed out by ruleset A spell-out (headedness) HYPOTHESIS.

9. Derivations

D1. (6) biʼdistał under EI-8 (Willie-reading)

1  Merge  [VP pro(3)  kick ]                              B; head-right VP
2  Merge  [vP INDF[hum,−def] [VP ...] v ]                  B; VPISH
3  Move   pro(3) → EXT position above vP;                  B; Willie: patient = "subject of agentive
          INDF stays INT, case ERGW                          passive"; INT agent ERGW (EI-1, EI-2)
4  Check  EXT must be definite (EI-3): pro(3) ok           interpretive; (6) 'the one who ...'
          Ranking (EI-4, 72b): agent [hum] ≤ patient?       satisfied only if the patient is human (tie)
5  Agree  M ← EXT; E(AGRo) ← [3]                            A
6  Move   ʼ-d- prefixed to E                                 A (morphological)
7  Spell  AGRo → bi- (inverse)                               A
8  Output biʼdistał                                          not segmented further

D2. (6) biʼdistał under EI-9 (Hale-reading)

1  Merge  [VP pro(3)  kick ]                              B
2  Merge  [vP ʼ-d-[hum] [VP ...] v ]                       B; Hale: ʼ-d- itself is the d-structure subject
3  Move   ʼ-d- → Spec,EP (EPP); case ERGH                  B; EI-6
4  Agree  M ← ʼ-d-: zero AGRs; E(AGRo) ← [3]; object       A; IX-3 "in the usual way"
          raises to Spec,MP
5  Move   ʼ-d- prefixed to E                                 A
6  Spell  AGRo: H? no. Subject [4]/1/2? no → b-             A; IX-1 elsewhere, IX-6
7  Output biʼdistał

D3. Schema for T70a (no Navajo form; a sourced token is needed)

   [patient: 3rd, non-human]  [agent: INDF, human]  bi- ... ʼ-d- ... V
   EI-8: step 4 of D1 fails (agent outranks patient = Willie's *72b.4)  → predicted bad
   EI-9: no ranking step                                                → predicted good

D4. (7) jishxash 'That person bit it' (for contrast)

1  Merge  [VP pro(3, −hum)  bite ]                        B
2  Merge  [vP DP[4,hum,def] [VP ...] v ]                   B; Hale fn. 16: j- as a DP when no nominal is adjoined
3  Agree  M ← DP[4]: zero AGRs; E(AGRo) ← [3]              A
4  Move   j- prefixed to E                                   A
5  Spell  AGRo[3], subject [4]/off-ranking → Ø              A; FS-8a = FS-8b here (§6.5)
6  Output jishxash

10. What it means for the role schema (Focus 2)

In (4), (5), (6) and (7) the role schema is the same: a biter or kicker as proto-agent, and the one bitten or kicked as proto-patient. The role assignment is ours HYPOTHESIS. Both sources still need properties that the schema does not give. Willie needs definiteness and the animacy ranking (EI-3, EI-4). Hale needs a case value on the agent (EI-6). And the T70a test, if it turns out as EI-8 predicts, would make the choice of bi- depend on whether the patient is human SOURCED (that the sources rely on these properties) / HYPOTHESIS (that this counts against the strong thesis). Those are properties of the items filling the roles. This is one more case where a pure "roles, not words" account of Navajo verb arguments comes up short. It does not refute a weaker account in which roles are assigned by a schema and the reference properties of the items are read at spell-out (proto-role schema, verb guide). The Algonquian X shows the same thing from the other side. A person hierarchy that includes 'people/someone' as a category of its own is a ranking of argument types, not of roles HYPOTHESIS.

11. Fidelity

  • Willie's forms in (1)–(5) and (7) were read on 110 dpi page images this run (pp. 76, 77, 87, 124). (6) is reused from the p. 152 image reading on indefinite-agent-b. The text layer garbles ł and nasal hooks, and no Navajo form was copied from it.
  • Willie gives no gloss lines for (45)–(47). Our Leipzig lines use her INV label and her p. 76 remark on léiʼ.
  • Willie explains the "it" in her (72) translations: "on the preferred interpretation of these sentences, for both the Direct and Inverse voices, the internal argument is taken to be a lower or indefinite being" (p. 88, text layer only, not image-checked). The rank labels in (4)–(5) are hers.
  • The case of the inverse patient is "Accusative" on p. 72 and "Absolutive" on pp. 77 and 100. Both are recorded here and neither is preferred (also noted on inverse-agree).
  • Meskwaki (8) is copied from Oxford's page image with his glosses. The Ojibwe hierarchy is stated for Ojibwe and is applied to Meskwaki by us.
  • EI-7 to EI-10, D1–D4 and every conclusion in §§5–7 and §10 are ours. EI-1 to EI-6 are quotations.

12. Open questions

IDQuestionWhat would settle it
T70aIs there a sourced agentive passive (bi- + ʼ-d-) with a non-human 3rd-person patient? EI-8 predicts it is as bad as Willie's *(72b.4); EI-9 predicts it is fine.Young & Morgan 1987 agentive-passive paradigms; any text with 'someone [verb]ed it'. OPEN
T70bDoes Willie (or anyone) say whether ʼadi- forms with 1st/2nd-person objects (niʼdilzhééh) are Direct or Inverse? By EI-4 they cannot be inverse, so what does "Ergative" mean there?Willie ch. 2 and ch. 9 (pp. 36–38, 220–224) on page images. OPEN Update 2026-10-28 (~8:46pm PT): (b) answered on third-voice: neither Direct nor Inverse; a third voice (VC-13, HYPOTHESIS).
T70cWhat case does Willie give the bi- patient of the agentive passive: Absolutive (as in the inverse, p. 100) or Accusative (p. 72)? Hale's (21) makes the 3i object Absolutive-like ('- = object of transitive).Willie pp. 36–38, 152–157 in full. OPEN Update 2026-10-28 (~8:46pm PT): (c) answered on third-voice: Absolutive in both of Willie's senses, not Accusative (VC-14, HYPOTHESIS).
T70dA form that separates FS-8a from FS-8b. Candidate: the narrative 4th person with an overt noun, possibly non-human (Willie p. 117 (14), the coyote; Hale fn. 16 analyses this use as agreement). With a 3rd-person object, is the object Ø?A sourced narrative 4th-person transitive with a 3rd object (Willie ch. 4; Young & Morgan 1987, cited by Willie p. 116, not opened). OPEN
T70eAlgonquian X>3: direct or inverse? If direct (as 1/2>X>3 predicts), Algonquian is evidence against EI-8 by analogy.Goddard 1994b; Ojibwe/Cree X>3 paradigms (Nichols 1980; Wolfart 1973). OPEN
T70fT69f restated: the impersonal (non-deictic, generic) 'one' ji- with a 3rd-person object. Ø (the deictic morpheme) or bi- (generic = not deictic)?A sourced transitive impersonal; Witherspoon 1977:84 (cited by Willie p. 118). OPEN

Sources: Willie 1991 pp. 72, 76, 77, 86, 87, 100, 124, 126, 221 (page images), pp. 36–37, 50–51, 118–122, 152, 157 (text layer); Hale 1997 (hale023) PDF p. 30 (image), pp. 26–27 (text); Oxford, "Algonquian", PDF pp. 7, 8, 17, 18. See also indefinite-agent-b, fourth-subject-p, inverse, inverse-agree, fourth person, open questions, sources.

"You are being shaved": in Willie's system the agentive passive is a third voice, neither Direct nor Inverse, and her relative-clause test makes its patient an absolutive subject (T70b, T70c)

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

Verbs on this page → dictionary: yiztał ‘kick’ · jizloh ‘rope’ · yishxash ‘bite’ · yi-dloh ‘laugh’

Restated on the unified structure (2026-11-15). The indefinites (ʼa-, ʼadi-, areal ho-) are restated as personless pronouns in ordinary argument positions (H60): ʼadi- as the silent indefinite agent in Spec,vP that raises to Spec,TP by the default (H10b), ʼa- in VP. The 3i data on this page have not yet been individually restated. 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, indefinite and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

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.

Willie 1991 glosses niʼdilzhééh 'You are being shaved' and shiʼdidleesh 'I am being painted' as "PASS" (p. 37 (21), (23)). She says the construction carries "a classifier that marks the construction as intransitive" (p. 36), and she calls its agent prefix ʼadi- "an indefinite Ergative pronoun" (p. 221) SOURCED. These forms cannot be Inverse, because voice alternation needs all arguments to be 3rd person (p. 86). They cannot be Direct either, because with a 1st-person patient a Direct agent is "the subject, the external Nominative argument, and not an internal (Ergative) agent" (p. 123). So with a 1st/2nd-person patient the agentive passive is a third configuration HYPOTHESIS (T70b). Her relative-clause generalization "(16) The heads of relative clauses are absolutives", where an absolutive is "either the intransitive subject or an internal argument of the transitive verb" (p. 160), applies to biʼdistał=yę́ę 'the one who was kicked', labelled "(subject of agentive passive)" (p. 152 (1f)) SOURCED. Together these make the patient an absolutive intransitive subject, not an Accusative HYPOTHESIS (T70c). They also mean that the Willie-style analysis of biʼdistał as an inverse (EI-8) survives only if it is amended to an intransitive inverse HYPOTHESIS.

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 Neither Direct nor Inverse (T70b) · 6 Which case for the patient? (T70c) · 7 Consequence for EI-8 vs EI-9 · 8 Rulesets A and B · 9 Derivations · 10 What it means for the role schema · 11 Fidelity · 12 Open questions

Notation is the same as on ergative-indefinite. A = the word level (prefixal, left-headed). B = the clause level (right-headed). Σ = the role schema, with Dowty's proto-roles as the baseline. Merge builds structure by putting two pieces together into a phrase. Agree values features: it copies person and number from a goal onto a probe that needs them. Move re-merges a piece that is already in the structure in a higher position. INDF = the indefinite human agent of the agentive passive. EXT/INT = external/internal argument in Willie's sense, where EXT is the subject. Under VPISH every agent is first merged in Spec,vP, and "external" means the position the agent or patient ends up in. S = the single argument of an intransitive; A/O = the agent/object of a transitive. ABSa and ABSb are Willie's two uses of "Absolutive", defined in VC-11.

1. Question

The last run (ergative-indefinite) left two items for Willie's chapters 2 and 9. T70b: with a 1st- or 2nd-person patient, where voice alternation is impossible (EI-4), is the agentive passive Direct, Inverse, or neither? T70c: is its patient Absolutive, as Willie says of the inverse's external patient (p. 100), or Accusative, as she says of the inverse's patient on p. 72? This page answers both from page images. Each answer rests on a reconstruction of Willie's terms, so both are HYPOTHESIS. The page also adds one consequence for the open choice between EI-8 and EI-9 (T69c).

2. Sources

SOURCED Willie, MaryAnn 1991. Navajo Pronouns and Obviation. PhD dissertation, University of Arizona. Page images read this run: pp. 24, 32–33, 36–39 (PDF = printed + 5), and pp. 57, 71, 123, 151–152, 159–160, 221–224 (PDF = printed + 6). Pp. 37, 123, 152 and 159 were also read in 300 dpi crops. Text layer only, as a finding aid: p. 72, pp. 157–158, pp. 166–167. P. 72 was read on the page image in an earlier run (EI-2).

Earlier pages reused here: ergative-indefinite (EI-1–EI-10) and indefinite-agent-b (IX-6, IX-8).

Not opened: Young & Morgan 1987, cited by Willie pp. 36 and 57 for "agentive passive" and "simple passive". T70a, T70d, T70e and T70f still need it or other unobtained sources.

3. Rules first

VC-1 to VC-10 are Willie's statements, quoted from the page images. VC-11 to VC-16 are ours.

VC-1  The "classifier" is "typically a valence marker that marks the verbs as         SOURCED (Willie p. 24, image)
      transitive or intransitive." Stem final, classifier just before it.
VC-2  Anaphors "mark arguments that have patient theta roles ... When an              SOURCED (Willie p. 32, image)
      anaphor is present, it always requires a -d or -l classifier, marking
      the verb intransitive."
VC-3  Indefinite pronouns, "like anaphors, ... show a classifier that marks the        SOURCED (Willie p. 36, image)
      construction as intransitive." In the agentive passive "a transitive
      verb is marked with an indefinite agent pronoun, along with an object
      pronoun, producing a passive construction." The indefinite agent "must
      be understood as human, but no noun that describes this agent may be
      adjoined."
VC-4  ʼadi- "occurs to the right of (internal to) the object pronoun, proving         SOURCED (Willie pp. 36–38, images)
      that it marks the unspecified agent." Reflexive and agentive passive:
      "In both these construction types, the patient precedes the agent in the
      prefix array" (p. 37).
VC-5  "Grammatical relation is determined by the prefix order, except the fourth      SOURCED (Willie p. 222, image)
      person forms ... the patient precedes the agent, which is the canonical
      order for agents and patients in all persons."
VC-6  "The single argument of an intransitive may have an agent, patient or           SOURCED (Willie p. 57, image)
      theme theta role." yíchxǫʼ 'It is ruined' = Young & Morgan's "simple
      passive".
VC-7  "The Inverse construction is comparable to a passive in that the subject        SOURCED (Willie p. 71, image)
      has a Patient theta role, but it differs from the passive in that it is
      a transitive sentence that has both agent and patient arguments."
VC-8  With 1st/2nd/4th patients and a 3rd agent, "the lower ranked third person      SOURCED (Willie p. 123, image)
      agent is the subject, the external Nominative argument, and not an
      internal (Ergative) agent, as we see in the Inverse."
VC-9  Incorporated pronouns and their traces "are assigned theta roles and case";    SOURCED (Willie p. 221, image)
      ʼadi- is "an indefinite Ergative pronoun ... in the Agentive Passive";
      ʼa- is "an indefinite Absolutive pronoun".
VC-10 "(16) The heads of relative clauses are absolutives. That is to say, the        SOURCED (Willie p. 160, image)
      head of the relative is either the intransitive subject or an internal
      argument of the transitive verb -- and recall that with the Inverse
      voice, the agent is the internal argument." biʼdistał=yę́ę is labelled
      "(subject of agentive passive)" (p. 152 (1f)).

VC-11 Willie's case vocabulary as she uses it (our compilation; she gives no         HYPOTHESIS
      table):
        NOM   = the EXT agent (p. 123)
        ERG   = an INT agent: inverse (pp. 72, 100, 123), ʼadi- (p. 221)
        ABSa  = the EXT patient of the inverse (pp. 77, 100)
        ABSb  = S, or an INT argument of a transitive (p. 160 (16); ʼa- p. 36)
        ACC   = the inverse's patient (p. 72 only)
      ABSa and ABSb disagree on two arguments. The inverse EXT patient is ABSa
      and ACC, but it is not ABSb, because it is the EXT of a transitive (VC-7).
      The inverse agent is ERG and also ABSb.
VC-12 "Transitive" in VC-1–VC-3, VC-7 is classifier valence (construction level).   HYPOTHESIS
      "a transitive verb" in VC-3 is the lexical verb's two-role frame.
      Willie therefore separates lexical transitivity (Σ has two roles) from
      construction valence (classifier).
VC-13 (T70b) With a 1st/2nd patient the agentive passive is neither Inverse          HYPOTHESIS (from VC-3, VC-8, VC-9, EI-4)
      (no all-3rd condition, EI-4) nor Direct (its agent is ERG = INT, not
      NOM = EXT, VC-8, VC-9). It is a third voice: one argument pronoun is
      EXT (the patient), the agent is INT and ERG, and the classifier is
      intransitive. Willie's own label for it is "PASS".
VC-14 (T70c) The patient of the agentive passive is ABSb: it is labelled             HYPOTHESIS (from VC-3, VC-10, VC-11)
      "subject" (p. 152 (1f)) of an intransitive-classifier construction
      (VC-3), so it is S, and it heads a relative (VC-10). It is also ABSa if
      it is EXT. It is not ACC: ACC appears only for the inverse patient, on
      p. 72. Willie never names this argument's case.
VC-15 (Focus 1) The prefix order, patient outside agent, is the same in the          SOURCED for the order (VC-4, VC-5);
      Direct, Inverse, reflexive, agentive passive and 4>4 configurations.     HYPOTHESIS for the split
      What changes from voice to voice is clause-level (B): externality,
      case, the voice label. Ruleset A registers voice only (i) at the stem
      edge (the classifier), (ii) in the agent pronoun (Ø vs ʼadi-), and
      (iii) in the 3rd object shape (yi- vs bi-).
VC-16 (Focus 2) Willie states the template in role terms ("patient precedes          SOURCED for the wording;
      agent", "patient theta roles"). But "subject" is not a role (VC-6), and    HYPOTHESIS for the reading
      which argument is EXT depends on the arguments' person, animacy and
      definiteness (VC-8, EI-3, EI-4). This supports a weak schema thesis
      (roles order the word-level slots) and not the strong one (roles,
      not lexical items, are the verb's arguments).

4. The data

All forms are given as printed and were read on page images this run. Lines marked "source" give Willie's own glosses. The Leipzig lines are ours, built only from her glosses, her underlying forms and her translations. CL = classifier; INDF.A = the indefinite agent ʼadi-. Aspect is not glossed where Willie's gloss uses an English participle.

Willie 1991 p. 37 (20)–(23) (page image, 300 dpi): reflexive vs agentive passive
(1)ʼádílzhééh
source: REF.2s.shaving; underlying ʼádi-ni-l-zheeh
REFL-2SG.S-CL-shave
'You are shaving yourself.' (20)
(2)niʼdilzhééh
source: 2s.PASS.shaving; underlying ni-'adi-l-zheeh
2SG.O-INDF.A-CL-shave
'You are being shaved.' (21)
(3)ʼádíshdleesh
source: REF.1s.painting
REFL.1SG.S.paint
'I am painting myself.' (22)
(4)shiʼdidleesh
source: 1s.PASS.painting
1SG.O.INDF.A.paint
'I am being painted.' (23)

Willie gives underlying forms only for (20) and (21). For (3) and (4) the Leipzig line is unsegmented. In (2) and (4) the patient prefix (ni-, shi-) stands to the left of the agent (VC-4). The same is true in the reflexives, where ʼádi- is the patient anaphor (VC-2).

Willie 1991 p. 123 (21)–(22) (page image, 300 dpi): Direct with a 1st/4th-person patient
(5)łééʼchąąʼíshishxash
dogsource: 1.3.bit
dog1SG.O.3S.bite.PFV
'The dog bit me.' (21)
(6)łééʼchąąʼíhashxash
dogsource: 4.3.bit
dog4.O.3S.bite.PFV
'The dog bit that person.' (22)

Willie: in (21)–(22) "the lower ranked third person agent is the subject, the external Nominative argument" (VC-8).

Willie 1991 p. 71 (37) and p. 57 (12) (page images): Inverse and simple passive
(7)łį́į́ʼdzaanéézyiztał
horsemulesource: 3.3.kicked
horsemule3O.3S.kick.PFV
'The horse kicked the mule.' (37a)
(8)dzaanéézłį́į́ʼbiztał
mulehorsesource: INV3.3.kicked
mulehorseINV.3O.3S.kick.PFV
'The mule was kicked by the horse.' (37b)
(9)yíchxǫʼ
source: 3s.ruined
3S.ruin.PFV
'It is ruined.' (12), Young & Morgan's "simple passive"
Willie 1991 p. 152 (1c), (1d), (1f) and p. 159 (13)–(15) (page images; (1e)–(1f), (13)–(14) at 300 dpi): relative heads
(10)yiztał=yę́ę
3O.3S.kick.PFV=REL
'the one he kicked' (1c), labelled "(Direct patient; internal argument)"
(11)biztał=yę́ę
INV.3O.3S.kick.PFV=REL
'the one who kicked him' (1d), labelled "(Inverse agent; internal argument)"
(12)biʼdistał=yę́ę
source: PASS 3.3.kicked
3O.INDF.A.kick.PFV=REL
'the one who was kicked' (1f), labelled "(subject of agentive passive)"
(13)yidloh=ę́ę
3S.laugh=REL
'the one who was laughing' (13)
(14)yizloh=ę́ę
3O.3S.rope.PFV=REL
'the one that was roped' (14)
(15)bizloh=ę́ę
INV.3O.3S.rope.PFV=REL
'the one who roped him' (15); "=ę́ę binds the agent, the transitive subject of the relative" (p. 160)

Willie's source glosses: (1c) 3.3.kicked=REL, (1d) INV 3.3.kicked=REL, (13) 3.laugh=REL, (14) 3.3.roped=REL, (15) INV3.3.roped=REL. The labels in quotation marks are hers. The (1f) label does not say "internal argument", while the (1c) and (1d) labels do. The text layer of ch. 6 (pp. 166–167, (30), (31), (34), (37)–(40)) was searched only as a finding aid. In every inverse relative it shows, =yę́ę binds the agent, and none binds the inverse's external patient (this needs image confirmation: T71b).

5. Neither Direct nor Inverse (T70b)

Two of Willie's statements exclude the two named voices for (2) and (4).

Not Inverse. Willie's voice rule (p. 86, EI-4) applies only when every argument is 3rd person SOURCED. (2) has a 2nd-person patient and (4) a 1st-person patient, so neither can be an Inverse in her sense.

Not Direct. When a 1st-person patient has a 3rd-person agent and no voice alternation is possible, the agent is "the subject, the external Nominative argument, and not an internal (Ergative) agent, as we see in the Inverse" (p. 123, (5)) SOURCED. In (2) and (4) the agent is ʼadi-, which she calls "Ergative" (p. 221), and on p. 123 she uses "Ergative" to mean the opposite of "external Nominative". So the agentive-passive agent is internal, and (2)/(4) are not Direct.

What is left. One argument pronoun is external, and that is the patient. The agent is internal and Ergative, and the classifier marks the construction as intransitive (p. 36). This is VC-13 HYPOTHESIS. Willie's own gloss for it is "PASS", and on p. 36 she calls it a "passive construction". So the third voice is just her passive, kept apart from both members of the Direct/Inverse pair. The comparison has a limit. (5) and (2) differ in verb (bite vs shave) and in the agent's person (3 vs INDF), so they are not a minimal pair. The argument rests on Willie's general statements and not on that contrast.

How it differs from the Inverse (VC-7). Willie says the inverse is "comparable to a passive in that the subject has a Patient theta role" but is "a transitive sentence that has both agent and patient arguments". The agentive passive also has both an agent and a patient pronoun, yet she calls it a passive. The only property that separates the two in her text is the classifier's valence (VC-3). Hence VC-12: "transitive" here means classifier valence, not a count of argument pronouns HYPOTHESIS. Rival reading VC-12′ HYPOTHESIS: on p. 71 "the passive" means only the simple passive (9), which has one argument, and p. 71 is silent on the agentive passive. Willie cites "Chomsky 1981" there, not her own ch. 2. VC-13 survives under both readings, because it does not depend on p. 71.

6. Which case for the patient? (T70c)

Willie uses "Absolutive" in two ways (VC-11) HYPOTHESIS. On pp. 77 and 100 (ABSa) it is the case of the inverse's external patient. On p. 160 (ABSb) it is the ordinary ergative-language sense: S, or the internal argument of a transitive. These conflict for the inverse patient. As EXT of a transitive it is ABSa but not ABSb, and on p. 72 the same argument is "Accusative". They also conflict for the inverse agent, which is Ergative (p. 72) and yet ABSb by p. 160, since it is internal.

The agentive-passive patient comes out Absolutive under both senses (VC-14) HYPOTHESIS.

  • ABSb: (12) biʼdistał=yę́ę heads a relative, so by (16) it is S or INT. Willie labels it "subject", the construction is intransitive by its classifier (p. 36), and so it is S.
  • ABSa: it is the external argument (VC-13). The external patient of the passive-like inverse is Absolutive (p. 100), so the passive's external patient is Absolutive too, if the two are parallel.
  • Accusative: nothing supports it. "Accusative" occurs only on p. 72, for the inverse patient, and Willie does not use the term again in the pages read.

The morphology cannot decide this. The patient prefix keeps its object shape in every voice (ni- in (2), shi- in (4) and (5), bi- in (8) and (12)), and grammatical relation is read off prefix order (p. 222). Case in Willie's sense is a clause-level (B) label with no word-level exponent of its own (T71d).

7. Consequence for EI-8 vs EI-9 (T69c)

EI-8 (the Willie reading) analysed biʼdistał as "point for point the inverse template": the 3rd-person patient is EXT and the INDF agent is INT. If that template is the transitive inverse (VC-7), the patient is EXT of a transitive. That makes it ABSa but not ABSb, and (16) then predicts it cannot head a relative. But (12) shows it heading one. So (16) and (12) together rule out EI-8 in its transitive form HYPOTHESIS. Two versions survive:

  • EI-8′ (new, this page) HYPOTHESIS: the agentive passive is an intransitive inverse. The patient is EXT and S (ABSa and ABSb), and the agent is INT and ERG. This is VC-13 extended to 3rd-person patients, and it matches Willie's own wording on p. 36.
  • EI-9 (Hale) HYPOTHESIS: ʼ-d- is the transitive subject and the patient is the internal object. The patient is then ABSb as an internal argument of a transitive, and (16) is satisfied without any amendment.

So the relative-clause test does not choose between Willie and Hale. It does remove one version of Willie. (16) is Willie's generalization over "the relative clauses we have seen so far" (p. 160), not a stated theorem, which is why this is a HYPOTHESIS and not a refutation. T71a below is the test that splits EI-8′ from EI-9.

8. Rulesets A and B (Focus 1)

StepLevelRulesetWhat happens
1vPBMerge the verb root with the patient pro. Merge the agent (pro or INDF) in Spec,vP (VPISH).
2ClauseBChoose the voice. All arguments 3rd: Direct or Inverse by ranking (EI-4). Otherwise a 3rd agent is EXT/NOM (VC-8). INDF agent: stays INT/ERG, and the patient is EXT/S (VC-13).
3ClauseBAssign case labels to the chain positions (VC-9, VC-11). They have no exponent of their own.
4WordAAgree: the object slot takes the patient's person and number (ni-/shi-/yi-/bi-). The subject slot takes the agent's (Ø, sh-, …), or ʼadi- is inserted internal to the object.
5WordAThe patient prefix goes outside the agent prefix in every voice (VC-4, VC-5). The classifier, next to the stem, spells the construction's valence (VC-1, VC-3).
6Word edgeA→B=yę́ę attaches to the finished word. Its binding condition (16) is stated over B-level notions (S/INT), so B information must still be readable after A has spelled out the word.

If the changeover hypothesis is right (it is not established), the switch at the word falls at step 6. The enclitic is phonologically part of the word (A), but what it may bind is decided by clause-level positions (B) HYPOTHESIS. The word-level order of steps 4–5 is the same in all four voices. Voice reaches A at only two points: the stem edge (the classifier) and the choice of exponent (Ø vs ʼadi-, yi- vs bi-) (VC-15).

9. Derivations

Each step names its operation and ruleset. Merge builds structure, Agree copies person and number onto a probe, and Move re-merges a piece higher up. The forms are Willie's. The structure is ours HYPOTHESIS.

D1  niʼdilzhééh 'You are being shaved' (p. 37 (21)), Willie reading (VC-13)
    1  Merge  [VP pro(2sg) √zheeh]                                    B   patient = Σ P-Patient
    2  Merge  v[CL: intransitive]; Merge INDF in Spec,vP               B   agent = Σ P-Agent, human (VC-3)
    3  Move   pro(2sg) → EXT (subject); INDF stays INT                 B   no voice alternation: not all 3rd (EI-4)
    4  Case   pro(2sg) = S → ABSb (and ABSa as EXT); INDF = ERG        B   VC-11, VC-14
    5  Agree  object slot ← [2sg] → ni-                                A   prefix shape unchanged by step 3
    6  Spell  INDF → ʼadi-, internal to ni- (VC-4); CL → l              A   Willie: ni-ʼadi-l-zheeh
    7  Out    niʼdilzhééh
    Hale reading (EI-9): step 3 Moves INDF (ʼ-d-) to subject; pro(2sg) stays INT (ABSb).
    Steps 5–7 are identical. The word does not distinguish the readings.

D2  łééʼchąąʼí shishxash 'The dog bit me' (p. 123 (21)), Direct
    1  Merge  [VP pro(1sg) √bite]; Merge pro(3) in Spec,vP             B
    2  Move   pro(3) → EXT; pro(1sg) stays INT                         B   VC-8
    3  Case   pro(3) = NOM; pro(1sg) = ABSb (INT of a transitive)      B   VC-11
    4  Agree  object slot ← [1sg] → shi-; subject slot ← [3] → Ø       A
    5  Out    shishxash; the noun łééʼchąąʼí is adjoined and linked to EXT   B   (the classifier is not segmented in the source)

D3  dzaanééz łį́į́ʼ biztał 'The mule was kicked by the horse' (p. 71 (37b)), Inverse
    1  Merge  [VP pro(3) √kick]; Merge pro(3) in Spec,vP               B
    2  Move   patient pro(3) → EXT; agent stays INT                    B   EI-4: patient ranks ≥ agent
    3  Case   patient = ABSa (p. 100) = ACC (p. 72); agent = ERG       B   VC-11 conflict
    4  Agree  object slot ← [3], inverse → bi-                          A
    5  Out    biztał; transitive classifier (VC-7)                     A

D4  biʼdistał=yę́ę 'the one who was kicked' (p. 152 (1f)), relative test
    1–3 as D1 with pro(3): patient EXT; INDF INT/ERG                   B
    4  Spell  biʼdistał                                                A
    5  Merge  =yę́ę at the word edge; it binds the patient                A→B
    6  Check  (16): head must be S or INT (ABSb)
              EI-8 (transitive inverse): patient = EXT of transitive → ✗ predicted bad, but attested
              EI-8′ (intransitive inverse): patient = S → ✓
              EI-9 (Hale): patient = INT → ✓

10. What it means for the role schema (Focus 2)

Two facts from this run bear on the thesis that Navajo verbs take theta roles rather than lexical items as arguments.

  • For a role schema at the word level. Willie states the prefix order in role terms. The patient precedes the agent "in all persons" (p. 222), and anaphors "mark arguments that have patient theta roles" (p. 32) SOURCED. She also separates "a transitive verb" (a two-role lexical frame) from a construction "marked as intransitive" by the classifier (p. 36). That is a split between the roles the verb selects (Σ) and the construction's valence (VC-12) HYPOTHESIS.
  • Against the strong thesis. Which role ends up as subject is not fixed by role: "The single argument of an intransitive may have an agent, patient or theme theta role" (p. 57) SOURCED. It depends on properties of the particular arguments: person (VC-8), definiteness (EI-3), humanness (VC-3) and ranking (EI-4). Those are properties of items, not of roles.

So this run supports only the weak thesis, that roles order the word-level slots, and it gives no new support to the strong thesis (VC-16) HYPOTHESIS. The strong thesis stays OPEN, and it is not established. No chemistry analogy is drawn from this page.

11. Fidelity

  • All fifteen numbered forms were read on page images this run. (1)–(4), (5), (12), (13) and (14) were also checked in 300 dpi crops. Glosses marked "source" are Willie's. Our Leipzig lines add no morphemes she does not give, except the aspect label PFV on her past-tense English glosses, following the convention of earlier pages.
  • Willie p. 39 (26) ʼasdzą́ą́ ʼíi?chxǫʼ and the transitive classifier symbol in the sentence before it are not clearly legible on the page image. Both are left out, and no classifier is named for (26).
  • The classifier of biʼdistał is not given in Willie's gloss. "Intransitive classifier" for it rests on her general statement on p. 36, not on a segmentation (T71c).
  • Willie's terms are inconsistent across chapters. "Absolutive" is used two ways (pp. 77/100 vs p. 160). The inverse agent is "internal ... not the subject" on p. 72 but "the transitive subject of the relative" on p. 160. The inverse patient is "Accusative" only on p. 72. VC-11 compiles this usage. It is not her table.
  • (16) is her generalization over "the relative clauses we have seen so far". The ch. 6 search for counterexamples was text-layer only.
  • Young & Morgan 1987, the source of the terms "agentive passive" and "simple passive", was not opened.

12. Open questions

IDQuestionWhat would settle it
T70bAgentive passive with a 1st/2nd patient: Direct, Inverse, or neither?Answered on this page: neither, a third voice (VC-13) HYPOTHESIS.
T70cIts patient: Absolutive or Accusative?Answered on this page: Absolutive in both of Willie's senses, not Accusative (VC-14) HYPOTHESIS.
T71aCan the INDF agent head a relative ('the one who kicked him/her', with ʼadi-)? Under (16), EI-8′ (INT agent = ABSb) says yes. VC-3's ban on adjoined nouns, together with relatives being adjoined (p. 158), says no. Under EI-9 the agent is the transitive subject, so (16) says no.A sourced relative on an agentive-passive verb that binds the agent, or a statement that this is impossible (Willie ch. 6 on images; Young & Morgan 1987). OPEN
T71bIs there any inverse relative whose head is the external patient ('the one who was kicked by him' on a bi- verb)? ABSb (16) predicts none.Willie ch. 6 pp. 161–175 on page images. Only the text layer was searched. Update 2026-10-29: answered on external-head: yes, (20b) 'The girl who was kissed by the boy' with Willie's (21) (SOURCED). The data were already on relatives (REL-EXT); the statement in §4 that every inverse relative binds the agent was wrong for (20b). (16) holds for relatives without nouns, which includes (12), so §6–§7 stand (HYPOTHESIS).
T71cWhich classifier does biʼdistał carry, and is the intransitivity in the classifier (Willie p. 36) or in the d-element of ʼ-d- (Hale's ergative d)?A segmented agentive-passive form with its classifier named (Young & Morgan 1987; Hale 1997 on the d-effect). OPEN
T71dWillie's case labels have no exponent of their own (§6). Is there any diagnostic of Absolutive vs Accusative in Navajo apart from prefix shape, such as relativization, postposition choice or the 4th-person pattern?Willie ch. 3 §3.3 and ch. 9 on images; Hale 1997 on case. OPEN
T71eFor the inverse agent, which is Willie's considered position: "internal ... not the subject" (p. 72) or "the transitive subject of the relative" (p. 160)?Willie ch. 3 §3.3.5 vs ch. 6 §6.2 read together on images. Update 2026-10-29: partly answered on external-head: ch. 6 calls the inverse agent internal twice (pp. 160, 161); "transitive subject" is loose (HYPOTHESIS). OPEN for ch. 3.
T70a, T70d–fUnchanged from ergative-indefinite.OPEN (Young & Morgan 1987, Goddard 1994b, Witherspoon 1977 not obtained)

Sources: Willie 1991 pp. 24, 32–33, 36–39, 57, 71, 123, 151–152, 159–160, 221–224 (page images), pp. 72, 157–158, 166–167 (text layer). See also ergative-indefinite, indefinite-agent-b, fourth-subject-p, inverse, fourth person, open questions, sources.

'Someone' has no person but is still human: the animacy ranking reads [human], not person, and Willie's "PASS 3.3" is a paradigm label (T74d)

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

Verbs on this page → dictionary: yiztał ‘kick’

Restated on the unified structure (2026-11-15). The indefinites (ʼa-, ʼadi-, areal ho-) are restated as personless pronouns in ordinary argument positions (H60): ʼadi- as the silent indefinite agent in Spec,vP that raises to Spec,TP by the default (H10b), ʼa- in VP. The 3i data on this page have not yet been individually restated. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Status on the spine: see unified structure, §4, indefinite and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

T74d asked which of two things in Willie 1991 the animacy ranking uses for INDF, the ʼadi- 'someone' agent of the agentive passive. On pp. 50–51 Willie lists ʼadi among the indefinite prefixes, apart from the first to fourth persons. In chapter 6 she glosses biʼdistał=yę́ę as "PASS 3.3.kicked", which looks like it files the agent as a third person. This run read the page images again and set the glosses side by side. In chapter 2 the same prefix is glossed with the patient's person only: "2s.PASS", "1s.PASS" (pp. 37, (21), (23)). In chapter 6 "3.3" is the label Willie puts on every 3-on-3 form of 'kick', Direct, Inverse and passive alike (p. 152, (1c), (1d), (1f)). And the ranking itself is stated over the animacy scale, not over persons (p. 86, (70)). So the ranking can read the [human] that p. 36 gives INDF, and INDF needs no person value at all HYPOTHESIS.

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 Answer to T74d · 6 What is left of p. 86 · 7 Rulesets A and B · 8 Derivations · 9 The role schema · 10 Fidelity notes · 11 Open questions

Notation is the same as on agent-ext and person-head. No Syntax · Semantics · Interface conventions file is in the repo or the context folder. A = the word level (prefixal, left-headed). B = the clause level (right-headed). Σ = the role schema, with Dowty's proto-roles as the baseline. EXT/INT = Willie's external/internal argument. INDF = the ʼadi- agent. Merge builds structure: it puts two pieces together into a bigger one. Agree builds nothing. It copies feature values such as person and number from one place to another. Move is Merge again of a piece that is already in the structure.

1. Question

EI-10 (ergative-indefinite) says INDF is on the animacy ranking, because it is human and indefinite, and an argument is off the ranking only if it is definite and human (EI-5's criterion). It also leaned on two supports that INDF counts as a third person: the gloss "PASS 3.3" (p. 152, and "PASS.3.3" on p. 157 (10)), and p. 86's line that "only third person arguments can be indefinite". agent-ext §8 found that pp. 50–51 put ʼadi in a separate class from the persons, which weakened the second support. T74d asks which of these the ranking actually uses. The question matters because the case against reading biʼdistał as an inverse (EI-8) rests on EI-10 (T73d).

2. Sources

SOURCED Willie, MaryAnn 1991. Navajo Pronouns and Obviation. PhD dissertation, University of Arizona. Page images rendered from the PDF and read this run: pp. 35–38 (PDF 40–43; offset printed + 5 in this part), 50–51 (PDF 55–56), 152 and 157 (PDF 158 and 163; printed + 6 here). P. 86 (PDF 92) was read from the PDF's text layer this run and matches the image-based quote already on ergative-indefinite (EI-4).

Already on the site and not new: (20)–(23) on pleonastic-psych and reflexive-reciprocal; the p. 36 "must be understood as human" quote (IX-5, indefinite-agent-b); (1c), (1d), (1f) and (10) on relatives. New here: setting the chapter 2 and chapter 6 glosses side by side, and the p. 35 boundary line "This completes the survey of the definite direct pronominals" before §2.3.

Not opened: Young & Morgan 1987 (the source of the term "agentive passive"), Hale et al. 1977 and Witherspoon 1977 (Willie's sources for the animacy hierarchy, p. 86).

3. Rules first

AP-1  §2.2 ends "This completes the survey of the definite direct pronominals"     SOURCED (Willie p. 35, image)
      and §2.3 opens "a group of pronominal prefixes that mark indefinite and
      non-referential arguments", which includes the agentive-passive prefix.
AP-2  "the indefinite agent must be understood as human, but no noun that          SOURCED (Willie p. 36, image; = IX-5)
      describes this agent may be adjoined."
AP-3  Chapter 2 glosses the agentive passive with the patient's person and          SOURCED (Willie p. 37, (21), (23), image)
      PASS, and nothing for the agent: (21) "2s.PASS.shaving", (23)
      "1s.PASS.painting".
AP-4  Chapter 6 glosses three 3-on-3 forms of 'kick' with the same "3.3":           SOURCED (Willie p. 152, (1c), (1d), (1f), image)
      (1c) "3.3.kicked" (Direct), (1d) "INV 3.3.kicked", (1f) "PASS 3.3.kicked".
      P. 157 (10) "PASS.3.3.will kiss" follows the same pattern.
AP-5  (70): the voice is chosen by "the animacy scale": agent higher → Direct;       SOURCED (Willie p. 86; = EI-4)
      patient higher → Inverse; equal → discourse. "The voice alternation can
      only be used if all arguments are third person. This clearly is
      connected with the fact that only third person arguments can be
      indefinite; first and second person arguments are always definite."
AP-6  "3.3" in chapter 6 is a label for the 3-on-3 paradigm cell that PASS or       HYPOTHESIS (from AP-3, AP-4)
      INV modifies. It is not a person value for INDF: in chapter 2, where the
      patient is 1st or 2nd, the agent slot gets PASS and no number.
AP-7  The ranking compares animacy (AP-5), and INDF's animacy is given directly    HYPOTHESIS (from AP-2, AP-5)
      by AP-2 ([human]). So the ranking needs no person value for INDF.
AP-8  Feature split for INDF: [human] yes, [definite] no, [person] none.           HYPOTHESIS (from AP-1, AP-2, AP-6, EI-1)
      Person and animacy are separate features in Willie's system, and the
      ranking reads only animacy.
AP-9  EI-10′ (replaces the support clause of EI-10; EI-10's criterion is            HYPOTHESIS (from AP-7, AP-8, EI-5)
      kept): an argument is off the ranking iff it is definite AND human.
      INDF is on the ranking because it is indefinite (AP-1) and human
      (AP-2). Neither "PASS 3.3" nor "only third person arguments can be
      indefinite" is needed.
AP-10 The precondition "all arguments are third person" (AP-5) is read as            HYPOTHESIS (reading of AP-5 with AP-1)
      "no argument is 1st, 2nd or 4th" (no definite human person). A
      personless INDF meets it, so a 3rd patient with INDF can be ranked.
      Whether Willie means this or a positive [3] on every argument is OPEN.

4. The data

All forms are copied from Willie's page images. Willie's gloss is given first, then a Leipzig line. Willie segments (21) herself; the other forms are not segmented further here.

(21), p. 37niʼdilzhééh (ni-ʼadi-l-zheeh in underlying form)
Willie: 2s.PASS.shaving · Leipzig: 2SG.OBJ-INDF.AGT-CL-shave
'You are being shaved.'
(23), p. 37shiʼdidleesh
Willie: 1s.PASS.painting · Leipzig: 1SG.OBJ.INDF.AGT.paint
'I am being painted.'
(1c), p. 152yiztał=yę́ę
Willie: 3.3.kicked=REL · Leipzig: 3OBJ.3SBJ.kick.PFV=REL
'the one he kicked' (Willie: "Direct patient; internal argument")
(1d), p. 152biztał=yę́ę
Willie: INV 3.3.kicked=REL · Leipzig: INV.3OBJ.3SBJ.kick.PFV=REL
'the one who kicked him' (Willie: "Inverse agent; internal argument")
(1f), p. 152biʼdistał=yę́ę
Willie: PASS 3.3.kicked=REL · Leipzig: 3OBJ.INDF.AGT.kick.PFV=REL
'the one who was kicked' (Willie: "subject of agentive passive")

In the Leipzig line for (1f) the "3.3" is not copied, because AP-6 reads it as a paradigm label. That reading is a HYPOTHESIS; Willie's own gloss is printed beside it.

5. Answer to T74d

The gloss leg. If "PASS 3.3" meant "a 3rd patient and a 3rd agent", chapter 2 should gloss (21) as something like "2s.3.PASS". It does not. It gives the patient's person and then PASS, which takes the agent's place in the gloss (AP-3). In chapter 6 the "3.3" is shared by a Direct, an Inverse and a passive form of the same verb (AP-4), so it reads as the name of the cell (third-person patient, transitive 'kick') that INV or PASS then modifies (AP-6) HYPOTHESIS. This leg of EI-10 is now as weak as the person-class leg.

What the ranking reads. Willie states the ranking over "the animacy scale" (AP-5) SOURCED. Animacy is exactly what p. 36 fixes for INDF: "must be understood as human" (AP-2) SOURCED. So the ranking can place INDF without asking what person it is (AP-7). INDF then has [human] and no [definite], and no person (AP-8) HYPOTHESIS.

Effect on EI-10. EI-10's criterion survives, and its two person-based supports can be dropped (AP-9, EI-10′). The case against EI-8 (reading biʼdistał as an inverse) still depends on EI-10′, which still has no direct evidence. That would be a form where an agentive passive's choice is visibly decided by ranking INDF against a non-human patient (T70a). T73d's request for direct evidence therefore stays OPEN, but T74d's "which one" is answered: animacy HYPOTHESIS.

6. What is left of p. 86

P. 86 says "only third person arguments can be indefinite" (AP-5). Read with a positive [3] feature, this files INDF as a 3 and clashes with AP-1 and AP-8. Read as "only arguments that are not 1st or 2nd can be indefinite", it is a statement about the definite persons and says nothing positive about INDF's person (AP-10) HYPOTHESIS. Willie's next clause, "first and second person arguments are always definite", is the contrast she draws, which fits the second reading. Which reading she intends is OPEN (T75a).

7. Rulesets A and B

StepLevelRulesetWhat happens
1vPBΣ assigns proto-agent and proto-patient. INDF carries [human, −def] and no person (AP-8).
2ClauseBChoose the voice. Precondition: no 1st/2nd/4th argument (AP-10). Rank by animacy (AP-5, AP-7). INDF agent: stays INT; the patient becomes EXT (VC-13 with a 1st/2nd patient; with a 3rd patient, Willie p. 152 "subject of agentive passive").
3WordAAgree copies person only from arguments that have it: the patient's into the object slot. The agent slot is spelled out as ʼadi-/ʼ-d-, not by a person value.
4WordASpell-out, prefixes left of the stem (left-headed).

The word is again the changeover point. Every ranking decision is made in B on features (animacy, definiteness), and A spells out person where there is a person to spell out HYPOTHESIS.

8. Derivations

D1  (21) ni'dilzhééh 'You are being shaved.' (Willie p. 37)
1  Merge  [VP pro(2sg)  shave ]                            B; head-right VP
2  Merge  [vP INDF[hum,−def,no person] [VP ...] v ]         B; VPISH (AP-8)
3  Check  voice: a 2nd-person argument is present, so       B; AP-5, AP-10; = VC-13
          no Direct/Inverse choice; ranking not consulted
4  Move   pro(2sg) → EXT; INDF stays INT                     B; VC-13 (third voice)
5  Agree  object slot ← [2sg] from pro(2sg)                  A
          agent slot: no person to copy (AP-8)
6  Spell  ni- (2SG.OBJ) + 'adi- (INDF) + l (CL) + zheeh      A; left-headed prefix array;
                                                              Willie's underlying form
7  Output ni'dilzhééh

D2  (1f) bi'distał=yę́ę 'the one who was kicked' (Willie p. 152)
1  Merge  [VP pro(3)  kick ]                                B
2  Merge  [vP INDF[hum,−def,no person] [VP ...] v ]         B; AP-8
3  Check  voice: no 1st/2nd/4th argument → ranking applies   B; AP-10
4  Rank   INDF [hum] vs pro(3) on the animacy scale          B; AP-5, AP-7 (no person read)
5  Move   pro(3) → EXT; INDF stays INT                       B; Willie: "subject of agentive passive"
6  Agree  object slot ← [3] from pro(3)                      A
7  Spell  bi- + 'adi-/'d- + stem                             A; not segmented further
8  Merge  [ clause ] + =yę́ę (REL, binds EXT)               A/B boundary; XA-1 (p. 181)
9  Output bi'distał=yę́ę

Step 4 of D2 shows where the ranking applies, not its outcome. Willie gives no judgment on which voice is preferred when a human 'someone' kicks a non-human patient. That is T70a, and it stays OPEN.

9. The role schema (Focus 2)

This run separates two kinds of information that a verb argument carries. Its role (proto-agent, proto-patient) comes from Σ. Its features ([human], [definite], person) come from the item filling the slot. The ranking that chooses the voice reads only features, and here only animacy (AP-7). That fits a weak schema thesis: roles are assigned schematically and the grammar then consults features of what fills them HYPOTHESIS. INDF is a useful test case because it has animacy but no person and no noun can be added to it (AP-2). It is about as close to a bare role filler as Navajo gets, and it still carries a lexical-type feature ([human]) that the clause uses. So the strong thesis (roles, not lexical items, are the arguments) gets no support from it and stays OPEN. Grant's own revised view, that the strong version does not hold cleanly, is GRANT'S CONJECTURE.

10. Fidelity notes

All Navajo forms are copied from the page images. Willie's "unspec." and "PASS" are rendered INDF.AGT in the Leipzig lines. (23) and the chapter 6 forms are not segmented, because Willie does not segment them. The p. 86 quote is from the PDF's text layer, which has OCR errors elsewhere on the page ("Heirarchy" is Willie's own spelling); the sentences quoted match the image-based quote on ergative-indefinite. No forms or judgments were invented, and D2 step 4 states no outcome.

11. Open questions

IDQuestionWhat would settle it
T75aDoes p. 86 "only third person arguments can be indefinite" give INDF a positive [3], or only say that indefinites are not 1st/2nd (AP-10)?A passage in Willie ch. 3–4 or Willie 1990 that ranks or agrees with INDF as a 3. OPEN
T75bDoes any source gloss an agentive passive with a 3rd patient as "3.PASS" (chapter 2 style), which would confirm AP-6?Young & Morgan 1987 (not opened); other Willie chapters on images. OPEN
T75cIs [human] on INDF a feature the ranking reads, or only an interpretive restriction (AP-7)?T70a: an agentive passive with a non-human patient, with a voice judgment. OPEN
T75dDo the other indefinite prefixes of pp. 50–51 ('a, ho-) also lack person while carrying animacy (p. 38: 'a "marks an indefinite inanimate argument")?Willie §2.3.2 on images (pp. 38–49). OPEN

Still open from earlier: T73b, T73c, T73d (direct evidence), T73e, T74a–c, T74e. Sources: Willie 1991 pp. 35–38, 50–51, 152, 157 (page images, this run); p. 86 (text layer, this run). See also agent-ext, ergative-indefinite, third-voice, indefinite-agent-b, open questions, sources.

Three indefinites outside the person paradigms: 'someone', 'something' and the areal ho- carry a sort feature but no person, and none of them takes a noun (T75a, T75d)

2026-11-11 audit pass 5. Our Leipzig lines for Willie (24) and (26) used IMP for her impersonal subject/object. In Leipzig glossing IMP means imperative, so they now read IMPERS, the label this page already uses for (94). Willie’s own glosses and the forms are unchanged. See the pass 5 report.

2026-11-11 audit pass 6. §8 (rulesets A and B) is now labelled HYPOTHESIS. The A→B changeover it describes is our hypothesis, not a sourced fact. The steps are unchanged. See the pass 6 report.

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

Verbs on this page → dictionary: ditą́ʼ ‘be deep; (with areal ho-) the area is deep’

Restated on the unified structure (2026-11-15). The indefinites (ʼa-, ʼadi-, areal ho-) are restated as personless pronouns in ordinary argument positions (H60): ʼadi- as the silent indefinite agent in Spec,vP that raises to Spec,TP by the default (H10b), ʼa- in VP. The 3i data on this page have not yet been individually restated. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Status on the spine: see unified structure, §4, indefinite and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

T75a asked whether Willie's p. 86 line "only third person arguments can be indefinite" gives the 'someone' prefix ʼadi- a positive third person. T75d asked whether the other indefinite prefixes, ʼa- and ho-, also lack person while carrying an animacy value. This run read Willie's own inventory of the verb's pronouns in her conclusions (pp. 221–222) on page images. It lists the expletive ho-, the indefinite absolutive ʼa and the indefinite ergative ʼadi- "in addition to the person paradigms", and then gives the paradigms as "first, second, third and fourth person" SOURCED. So in her own inventory none of the three is a third person. The p. 86 line sits in a section about nouns linked to ordinary third-person pronouns, which can be read as indefinite with léiʼ (p. 76, (43)–(44)). It is best read as being about those, not about the indefinite prefixes HYPOTHESIS. On T75d: all three are personless, and each has its own sort value. ʼadi- is human (pp. 36, 221) and ho- is "an abstract space, area, or entity" (p. 45) SOURCED. For ʼa- Willie gives conflicting values: an "inanimate argument" in general, but an "animate subject" with intransitives on p. 38, and then an "inanimate subject marker" for that same intransitive (24) on p. 39 OPEN. All three exclude an adjoined noun (p. 51). That supports a tight rule: a noun can be linked at the clause level (ruleset B) only to a slot that has a person value from the word level (ruleset A) HYPOTHESIS.

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 Answer to T75a · 6 Answer to T75d · 7 No person, no noun · 8 Rulesets A and B · 9 Derivations · 10 The role schema · 11 Fidelity notes · 12 Open questions

Notation is the same as on passive-rank and agent-ext. No Syntax · Semantics · Interface conventions file is in the repo or the context folder. A = the word level (prefixal, left-headed). B = the clause level (right-headed). Σ = the role schema, with Dowty's proto-roles as the baseline. INDF = the ʼadi- agent; IMP = the impersonal ʼa-; AR = the pleonastic ("areal") ho-. [π] = a person value (1, 2, 3 or 4). Merge builds structure: it puts two pieces together into a bigger one. Agree builds nothing. It copies feature values such as person and number from one place to another. Link is Willie's "relation of predication" between an adjoined noun and a pronoun (p. 221). It is not Agree, and here it is a ruleset-B operation.

1. Question

passive-rank concluded that INDF has [human] and no person (AP-8) and that the animacy ranking reads only [human] (AP-7). It left one loose end on p. 86 (AP-10): if "only third person arguments can be indefinite" means a positive [3] on every indefinite, INDF would be a third person after all (T75a). It also asked whether this split (sort feature yes, person no) holds for the whole class of indefinite prefixes that Willie lists on pp. 50–51, or only for ʼadi- (T75d). If it holds for the class, "personless" is a property of a natural class and not a fact about one prefix. That matters for Focus 2: these are the slots that come closest to being bare roles.

2. Sources

SOURCED Willie, MaryAnn 1991. Navajo Pronouns and Obviation. PhD dissertation, University of Arizona. Page images rendered from the PDF and read this run: pp. 38–39 (PDF 43–44, printed + 5; (24) and (26) at 600 dpi), p. 45 (PDF 50), p. 51 (PDF 56), p. 76 (PDF 82), p. 102 (PDF 108), p. 126 (PDF 132), pp. 221–222 (PDF 227–228) and p. 225 (PDF 231) (printed + 6 from ch. 3 on). P. 86 (PDF 92), pp. 84–85 and p. 116 (PDF 122) were read from the text layer only.

Already on the site and not new: (24)–(26) and (40) on pleonastic-psych; the p. 221 phrase "an indefinite Ergative pronoun" (EI-1 on ergative-indefinite); the p. 36 "must be understood as human" quote (AP-2). New here: p. 221's "In addition to the person paradigms" frame and the p. 222 list of the paradigms; p. 221's "an anaphor that is not inflected for person with the same morphological shape"; p. 225's "pleonastic ho-, which is not a fourth person form"; the p. 38 vs p. 39 animacy values of ʼa-; and (43)–(44) of p. 76 as the context of p. 86.

Not opened: Young & Morgan 1987, Willie 1990, Hale et al. 1977, Witherspoon 1977.

3. Rules first

PX-1  "In addition to the person paradigms, there is an Expletive subject            SOURCED (Willie p. 221, image)
      pronoun ho-; an indefinite Absolutive pronoun 'a; an indefinite Ergative
      pronoun -'adi- that appears in the Agentive Passive which can be used
      only in reference to humans; and an anaphor that is not inflected for
      person with the same morphological shape."
PX-2  "There are also paradigms of first, second, third and fourth person             SOURCED (Willie p. 222, image)
      pronouns, which typically vary little in shape according to grammatical
      relation".
PX-3  "After setting aside pleonastic ho-, which is not a fourth person form".        SOURCED (Willie p. 225, image; cf. p. 102)
PX-4  'a-: "This pronoun marks an indefinite inanimate argument. When it occurs       SOURCED (Willie p. 38, image)
      with an intransitive verb, it marks an indefinite animate subject; when it
      occurs with a transitive verb, it marks an indefinite inanimate patient.
      In never permits an adjoined coreferent nominal." [sic "In"]
PX-5  Of (24): "an intransitive construction, with an indefinite inanimate            SOURCED (Willie p. 39, image)
      subject marker 'a-, which will not allow a nominal adjunct."
PX-6  ho-: "While the fourth person refers exclusively to human beings,               SOURCED (Willie p. 45, image)
      pleonastic ho- refers to an abstract space, area, or entity."
PX-7  "the definite pronouns permit nominal adjuncts (including free pronouns),       SOURCED (Willie p. 51, image)
      while the indefinite pronouns exclude adjoined nominals." The indefinites
      are "the 'adi that appears in the "Agentive Passive", the
      Impersonal/indefinite -'a, and the pleonastic ho-".
PX-8  P. 76: a noun's preferred reading "is that it is definite unless it is          SOURCED (Willie p. 76, image)
      specifically marked indefinite with such a word as léi'"; "(42 - 44)
      show that either argument in a Direct Transitive can be either definite
      or indefinite."

PX-9  Feature inventory of the verb's argument slots:                                 HYPOTHESIS (from PX-1, PX-2, PX-3)
      PERSON slots  [π ∈ {1,2,3,4}]  (the paradigms, PX-2)
      INDF          [human, −def]      no π
      IMP           [sort: OPEN (PX-4 vs PX-5), −def]  no π
      AR            [sort: area/space/abstract]   no π
      The anaphor is also π-less (PX-1, Willie's own words), but it is
      bound, not indefinite, and is left aside here.
PX-10 P. 86 "only third person arguments can be indefinite" ranges over the           HYPOTHESIS (from PX-1, PX-2, PX-8)
      PERSON slots: among them, only [3] slots can have an indefinite reading,
      through an adjoined noun with léi' ((43), (44)). It says nothing about
      INDF/IMP/AR, which are not in the paradigms. This settles AP-10's
      reading: no positive [3] on INDF.
PX-11 LINK (ruleset B): a bare noun can be linked to a slot only if the slot          HYPOTHESIS (from PX-7, PX-9)
      has a person value [π]. INDF, IMP and AR have none, so they never take
      a noun (PX-4, PX-5, PX-7; p. 36 for INDF; (40d) for AR).
PX-12 LINK is not the only clause-level attachment: AR takes a locative PP           SOURCED ((40e), p. 46) /
      ((40e) táyi' hoditą́'), so PX-11 restricts bare nouns only.                    HYPOTHESIS (the restriction)
PX-13 Sort and person are separate features. The sort value is present on every      HYPOTHESIS (from PX-6, PX-9, AP-2)
      indefinite prefix for which Willie gives one consistently (INDF, AR);
      person is absent on all three.

4. The data

All forms are copied from Willie's page images. Willie's gloss is given first, then a Leipzig line. Willie does not segment (24)–(26); the Leipzig lines use dots for unsegmented material. "PFV" is not added because Willie glosses tense ("ruined") and not aspect.

(24), p. 38ʼííchxǫʼ
Willie: unspec. subj. ruined · Leipzig: IMPERS.SBJ.ruin
'Ruining took place.'
(25), p. 38(hatáál) yíchxǫʼ
Leipzig: (ceremony) 3SBJ.ruin
'(The ceremony) it is ruined.'
(26), p. 39ʼasdzą́ą́ ʼííłchxǫʼ
Leipzig: woman IMPERS.OBJ.3SBJ.TR.ruin (Willie: the classifier marks the verb transitive)
'The woman did the ruining (of it).'
(40c), p. 46hoditą́ʼ
Willie: ho-is deep · Leipzig: AR-deep
'It (the area) is deep.'
(40d), p. 46*tó hoditą́ʼ
Willie: water ho-is deep · Leipzig: water AR-deep
(40e), p. 46táyiʼ hoditą́ʼ
Willie: water: in ho-is deep · Leipzig: water.in AR-deep
'It is deep there in the water.'
(43), p. 76ʼashkii léiʼ ʼatʼééd yiztsʼǫs
Leipzig: boy INDF.DEM girl 3OBJ.3SBJ.kiss
'A boy kissed the girl.'
(44), p. 76ʼashkii ʼatʼééd léiʼ yiztsʼǫs
Leipzig: boy girl INDF.DEM 3OBJ.3SBJ.kiss
'The boy kissed a girl.'

(40c)–(40e) are copied from pleonastic-psych, where they were read on the page images. (43)–(44) carry no gloss line in Willie. Willie calls léiʼ "the indefinite demonstrative" (p. 116, text layer, this run), so INDF.DEM here means that word, not the INDF prefix.

5. Answer to T75a

Willie's inventory. In her conclusions Willie sorts the verb's pronouns into two groups. First come three items "in addition to the person paradigms": expletive ho-, indefinite absolutive ʼa and indefinite ergative ʼadi- (PX-1). Then come "paradigms of first, second, third and fourth person pronouns" (PX-2) SOURCED. Read with pp. 50–51 (PX-7), where the same three are "the prefixes that mark indefinite arguments", this gives no place for INDF in the third-person paradigm. agent-ext §8 got the same answer from pp. 50–51 alone. P. 221 states it in so many words.

What p. 86 is about. The p. 86 sentence closes §3.3, a section about the definiteness of nouns linked to the verb's pronouns. On p. 76 a bare noun is definite by default and indefinite with léiʼ, and either argument of a Direct transitive can be indefinite ((43), (44), PX-8) SOURCED. Both of those arguments are ordinary third-person slots (yiztsʼǫs, Willie's "3.3" type). So "only third person arguments can be indefinite" fits the reading "of the person slots, only [3] ones admit an indefinite reading", with first and second person "always definite" as the contrast she draws (PX-10) HYPOTHESIS. On this reading p. 86 does not mention the indefinite prefixes at all.

Verdict. No source opened gives INDF a positive [3]. Willie's own inventory places it outside the paradigms. AP-10's second reading is now the better-supported one, and the precondition "all arguments are third person" can be restated as "no person slot is 1st, 2nd or 4th" HYPOTHESIS. This is still a reading of Willie and not her stated intent, so T75a is answered as HYPOTHESIS. It does not become SOURCED.

6. Answer to T75d

PrefixPersonSort valueNoun adjunctStatus
INDF ʼadi-none: outside the paradigms (PX-1)human: "must be understood as human" (p. 36); "only in reference to humans" (p. 221)no: "no noun that describes this agent may be adjoined" (p. 36)SOURCED
IMP ʼa-none: outside the paradigms (PX-1)conflicting: "inanimate argument" in general; "animate subject" with intransitives (p. 38); but (24), intransitive, has an "inanimate subject marker" (p. 39)no: "never permits an adjoined coreferent nominal" (p. 38); (24) "will not allow a nominal adjunct" (p. 39)person, adjunct: SOURCED; sort: OPEN
AR ho-none: outside the paradigms (PX-1); "not a fourth person form" (PX-3)"an abstract space, area, or entity", against the 4th person's "exclusively ... human beings" (p. 45)no bare noun: *(40d); a locative PP is fine: (40e)SOURCED

So the answer to T75d is yes for person: all three are personless, and Willie says so herself by listing them apart from the paradigms SOURCED. For the sort feature it is yes for INDF and AR SOURCED. For IMP it cannot be settled from Willie, because she gives "animate" and "inanimate" for the same intransitive use two pages apart OPEN. The English of (24), 'Ruining took place', has no participant at all, which fits neither value well. One possible reading is that intransitive ʼa- is an unspecified subject with no sort restriction. That reading is not tested and is not adopted here.

Generalization: the indefinite prefixes are a natural class defined by [−def] and no [π], and each member adds its own sort value (PX-9, PX-13) HYPOTHESIS. This extends passive-rank's AP-8 from one prefix to the class.

Willie's own generalization is about definiteness: "the definite pronouns permit nominal adjuncts ..., while the indefinite pronouns exclude adjoined nominals" (PX-7) SOURCED. On the definite side, every pronoun that takes a noun is a person-paradigm slot: 3rd ((25), (26), (43)), 1st/2nd through free pronouns (p. 51), and 4th, where on p. 126 "the single nominal present in the utterance can be linked to either an external or an internal argument" SOURCED. On the indefinite side, every pronoun that refuses a noun is outside the paradigms (PX-1). So on all the sourced cases, "definite" and "has [π]" pick out the same slots. PX-11 states the condition on the feature that Agree actually copies in the word, [π], and not on definiteness HYPOTHESIS.

The two versions make different predictions in one place. In the Direct voice a 3rd-person slot can be indefinite with léiʼ and still takes a noun ((43), (44)). A definiteness condition would have to be stated over the slot's lexical class (definite pronoun) and not its reading. A [π] condition covers (43) with no extra clause: the slot is [3], so LINK applies, and the noun brings the indefinite reading. So PX-11 is the tighter statement HYPOTHESIS. A counterexample would be a personless prefix with a linked bare noun, or a person slot that refuses one for no other reason. None has been found in the pages opened. PX-12 limits the claim to bare nouns, because AR takes a locative PP.

8. Rulesets A and B HYPOTHESIS

StepLevelRulesetWhat happens
1vPBΣ assigns roles to slots (proto-agent, proto-patient, or none for AR: Willie introduces it as Navajo's counterpart of expletives, which "fill subject positions that are not assigned a theta role", p. 45).
2WordAAgree copies [π] into the prefix positions from the slots that have one. INDF/IMP/AR have none, so they are spelled out as fixed shapes (ʼadi-, ʼa-, ho-).
3WordASpell-out: prefixes left of the stem (left-headed). The word closes. This is the changeover A→B.
4ClauseBLINK: a noun adjoined to the verb word (right-headed: nouns precede the verb) is linked by predication to a slot that has [π] (PX-11). A locative PP can link to AR (PX-12).

The point for Focus 1: LINK in B depends on a feature that only A makes visible. That is a word-level feature [π], valued by Agree inside the word, and it controls a clause-level operation. This is the same handoff found for the voice choice on passive-rank, except that this time B reads person and not animacy HYPOTHESIS.

9. Derivations

D1  (24) 'ííchxǫ' 'Ruining took place.' (Willie p. 38)
1  Merge  [VP IMP[−def, no π]  ruin ]                       B; head-right VP; intransitive
2  Merge  [vP [VP ...] v ]                                   B; VPISH: the sole argument starts VP-internal
3  Move   IMP → subject position                             B; (Willie: "unspec. subj.")
4  Agree  subject slot: no π to copy                         A; PX-9
5  Spell  'a- (fixed shape) + stem                           A; left-headed; not segmented further
6  Output 'ííchxǫ'
7  LINK   none possible: IMP has no π                        B; PX-11; Willie p. 38 "never permits an
                                                              adjoined coreferent nominal"

D2  (25) (hatáál) yíchxǫ' '(The ceremony) it is ruined.' (Willie p. 38)
1  Merge  [VP pro[3]  ruin ]                                 B
2  Merge  [vP [VP ...] v ]; Move pro → subject                B; VPISH
3  Agree  subject slot ← [3] from pro                        A
4  Spell  yíchxǫ'                                            A; not segmented
5  LINK   hatáál ⟶ pro[3] (optional; adjoined left)           B; PX-11 met

D3  (26) 'asdzą́ą́ 'ííłchxǫ' 'The woman did the ruining (of it).' (Willie p. 39)
1  Merge  [VP IMP[−def, no π]  ruin ]                       B; IMP = patient
2  Merge  [vP pro[3] [VP ...] v(TR) ]                         B; VPISH: agent in Spec,vP;
                                                              Willie: the classifier marks transitive
3  Move   pro[3] → subject                                   B
4  Agree  subject slot ← [3]; object slot: no π               A
5  Spell  'a- + ... + ł (CL) + stem                          A; left-headed; not segmented further
6  LINK   'asdzą́ą́ ⟶ pro[3] (agent), not ⟶ IMP               B; PX-11; Willie p. 39 "permits an adjunct
                                                              to the third person subject pronoun"

D4  (40c) hoditą́', *(40d) tó hoditą́', (40e) táyi' hoditą́' (Willie p. 46)
1  Merge  [VP deep ]; subject position filled by AR           B; AR: pleonastic, no theta role (p. 45)
2  Agree  subject slot: no π (PX-3: not a 4th person)         A
3  Spell  ho- + ditą́'                                          A; Willie segments ho-
4  LINK   *tó ⟶ AR: blocked, AR has no π                      B; PX-11 → *(40d)
5  Adjoin táyi' (PP) ⟶ AR: allowed                           B; PX-12 → (40e)

D5  (43) 'ashkii léi' 'at'ééd yizts'ǫs 'A boy kissed the girl.' (Willie p. 76)
1  Merge  [VP pro[3]  kiss ]; [vP pro[3] [VP ...] v ]          B; VPISH
2  Rank   both 3rd, equal or agent higher → Direct (yi-)       B; (70)
3  Agree  object ← [3]; subject ← [3]                        A
4  Spell  yizts'ǫs                                           A; not segmented
5  LINK   'ashkii léi' ⟶ pro[3] (agent); 'at'ééd ⟶ pro[3] (patient)   B; PX-11 met; léi' supplies the
                                                              indefinite reading (PX-8, PX-10)

In D3 step 6 the choice of the agent as the noun's host is Willie's statement, not a derivation result: with IMP unable to host, PX-11 leaves only the agent. D5 step 2 records Willie's (70) as the rule that applies. The two nouns' order follows her p. 76 examples, and no new order is claimed.

10. The role schema (Focus 2)

The indefinite prefixes are the best Navajo test of "roles, not lexical items, as arguments", because they fill a role and can never be joined by a noun. This run shows that even these slots are not bare roles. Each one carries a sort value (human, area/space), with IMP unclear, and AR is Willie's pleonastic subject, introduced as the Navajo counterpart of expletives that "fill subject positions that are not assigned a theta role" (p. 45) SOURCED. So what the verb combines with is a role plus a feature bundle, and the grammar reads the bundle twice: the ranking reads sort (passive-rank AP-7), and LINK reads person (PX-11) HYPOTHESIS. That is the weak thesis. The strong thesis would need slots with role and nothing else, and none has been found. It stays OPEN, and Grant's revised view that it does not hold cleanly is GRANT'S CONJECTURE. AR also cuts the other way: it is an "incorporated pleonastic subject pronoun" (p. 45), a pronominal slot with no theta role. So "argument" and "role-bearer" come apart in Navajo in both directions HYPOTHESIS.

11. Fidelity notes

All Navajo forms are copied from the page images. (24) and (26) were checked at 600 dpi: (24) has two acute accents, ʼíí-, and (26) has barred ł. P. 38's "In never permits" is Willie's typo for "It". Willie writes ʼa, -ʼa and ʼa- for the same prefix (pp. 38, 39, 51, 221), and all are rendered here as ʼa-. On p. 221 "not inflected for person" is said of the anaphor, the prefix with the same shape as ʼadi-. Willie does not use those words for INDF itself. INDF's personlessness rests on the "in addition to the person paradigms" frame. P. 86 is from the text layer, as before. No forms, judgments or segmentations were invented. Where Willie does not segment a form, the derivations say "not segmented".

12. Open questions

IDQuestionWhat would settle it
T76aIs intransitive ʼa- animate (p. 38), inanimate (p. 39), or unrestricted (the English of (24) names no participant)?More intransitive ʼa- forms with a translation that names the subject; Young & Morgan 1987. OPEN
T76bPX-11 vs Willie's definiteness wording: is there a personless prefix with a linked bare noun, or a person slot that refuses one for no other reason?A search of Willie ch. 3 on images; Willie 1990. OPEN
T76cDoes IMP enter the animacy ranking (70)? An IMP object is "inanimate" (p. 38) and so should rank lowest. Is a bi- form with an IMP object attested or excluded?An Inverse form with ʼa-; Willie ch. 3 or Young & Morgan. OPEN
T76dAlgonquian: are the impersonal actor X and the inanimate "obviative" in a person paradigm, or outside it like the Navajo indefinites (ergative-indefinite: X is ranked 1/2 > X > 3)?Oxford's "Algonquian" chapter (the copy used on ergative-indefinite) re-read for X's agreement. OPEN

T75a answered as HYPOTHESIS (PX-10). T75d answered: person SOURCED, sort SOURCED for INDF and AR, OPEN for IMP (T76a). Still open: T75b, T75c, T73b, T73c, T73d (direct evidence), T73e, T74a–c, T74e. Sources: Willie 1991 pp. 38–39, 45, 51, 76, 102, 126, 221–222, 225 (page images, this run); p. 86 (text layer). See also passive-rank, pleonastic-psych, ergative-indefinite, agent-ext, open questions, sources.

'Something' is the indefinite absolutive: in every example Willie gives, ʼa- stands for the inanimate thing affected, so the "animate subject" clause on p. 38 has no example behind it (T76a)

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

2026-11-11 audit pass 5. Our Leipzig lines for Willie (24) and (26) used IMP for her impersonal subject/object. In Leipzig glossing IMP means imperative, so they now read IMPERS, the label this page already uses for (94). Willie’s own glosses and the forms are unchanged. See the pass 5 report.

2026-11-11 audit pass 6. §8 (rulesets A and B) is now labelled HYPOTHESIS. The A→B changeover it describes is our hypothesis, not a sourced fact. The steps are unchanged. See the pass 6 report.

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

Restated on the unified structure (2026-11-15). The indefinites (ʼa-, ʼadi-, areal ho-) are restated as personless pronouns in ordinary argument positions (H60): ʼadi- as the silent indefinite agent in Spec,vP that raises to Spec,TP by the default (H10b), ʼa- in VP. The 3i data on this page have not yet been individually restated. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §4, indefinite and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

T76a asked whether the impersonal prefix ʼa- is animate when it is the subject of an intransitive verb (Willie p. 38), inanimate (p. 39), or unrestricted. This run re-read Willie's §2.3 introduction (p. 36), §2.3.2 (pp. 38–39), the simple-passive passage of §3.1.3 (pp. 57–58) and pp. 96–100 on page images. Willie calls ʼa- an "indefinite absolutive argument, that is either an intransitive subject or transitive object" (p. 36) SOURCED. Its intransitive example (24) is paired with (25), the same verb with the noun 'ceremony'. That verb is the one she calls a "simple passive" whose one argument "has a patient theta role" (p. 57) SOURCED. Her scenario for (24) is a ceremony that goes wrong, and in the transitive (26) the same prefix is "an unspecified inanimate patient" of the same verb. So in every example the prefix names the thing that gets ruined. Four of Willie's five statements about it say "inanimate", and the one that says "animate" has no example HYPOTHESIS. The rule proposed here: ʼa- is an inanimate, personless indefinite that can fill only the absolutive (patient-side) slot. It mirrors ʼadi-, the human, personless indefinite that fills only the agent slot HYPOTHESIS. For the role schema, this means each indefinite's sort goes with its role. For ʼadi- the sort follows from Dowty's proto-agent list, but for ʼa- it does not, so the strong "roles, not words" thesis stays OPEN.

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 Willie's five statements · 6 Answer to T76a · 7 What (94) can and cannot show · 8 Rulesets A and B · 9 Derivations · 10 The role schema · 11 Fidelity notes · 12 Open questions

Notation is the same as on indef-person and passive-rank. No Syntax · Semantics · Interface conventions file is in the repo or the context folder. A = the word level (prefixal, left-headed). B = the clause level (right-headed). Σ = the role schema, with Dowty's proto-roles as the baseline. IMP = the impersonal ʼa-; INDF = the ʼadi- agent. [π] = a person value (1, 2, 3 or 4). ABS = Willie's "absolutive" (intransitive subject or transitive object, p. 36). Merge builds structure: it joins two pieces into a bigger one. Agree builds nothing. It copies feature values such as person and number from one place to another. Link is Willie's predication relation between an adjoined noun and a pronoun (p. 221). It is not Agree, and here it is a ruleset-B operation.

1. Question

indef-person showed that ʼadi-, ʼa- and ho- sit outside Willie's person paradigms. It also showed that ʼadi- and ho- each carry a fixed sort value (human; area or space). For ʼa- it found a conflict (PX-4 vs PX-5): on p. 38 the intransitive use "marks an indefinite animate subject", while on p. 39 the intransitive example (24) has "an indefinite inanimate subject marker". T76a asks which is right, or whether the prefix has no sort at all. The answer matters for Focus 2. If IMP's sort depends on its role, then the indefinite prefixes are closer to role-restricted pronouns than to bare roles.

2. Sources

SOURCED Willie, MaryAnn 1991. Navajo Pronouns and Obviation. PhD dissertation, University of Arizona. Page images read this run: p. 36 (PDF 41), pp. 38–39 (PDF 43–44) (printed + 5); pp. 57–58 (PDF 63–64), pp. 96–100 (PDF 102–106) (printed + 6). The text layer was used only to find passages (searches for "unspec", "Impersonal", "absolutive", "simple passive").

Already on the site and not new: (24)–(26) and the p. 38/p. 39 quotes (indef-person PX-4, PX-5); (89), (94)–(97) and the p. 97–99 quotes (intransitive-pp); Dowty 1991:572 (27)–(28) as recorded on keep vs bring about. New here: p. 36 "an indefinite absolutive argument, that is either an intransitive subject or transitive object"; pp. 57–58 (12)–(14) and "the subject has a patient theta role" set beside (24)–(25); the p. 38 scenario for (24); the tally of Willie's five statements on IMP's sort.

Not opened: Young & Morgan 1987 (Willie cites p. 120 for "simple passive"), Willie 1990, Hale et al. 1977, Witherspoon 1977.

3. Rules first

IS-1  "In a second construction type, a pronominal prefix marks an indefinite          SOURCED (Willie p. 36, image)
      absolutive argument, that is either an intransitive subject or transitive
      object."
IS-2  "an indefinite Absolutive pronoun 'a" (in the p. 221 inventory; = PX-1)           SOURCED (Willie p. 221, image)
IS-3  "The single argument of an intransitive may have an agent, patient or theme       SOURCED (Willie p. 57, image)
      theta role. There is a kind of intransitive construction in Navajo where the
      subject has a patient theta role." Example (12) yíchxǫ' 'It is ruined';
      "Young and Morgan (1987:120) call this the simple passive."
IS-4  Of (14): "the single argument of the intransitive verb, the subject, which has   SOURCED (Willie p. 58, image)
      a patient theta role."
IS-5  (24) is introduced with "Compare the following intransitive form that permits     SOURCED (Willie p. 38, image)
      an adjoined nominal: (25) (hatáál) yíchxǫ'". Scenario for (24): "when a
      mistake made in a ceremony required it to be terminated or corrected."
IS-6  (26): "a related transitive verb with a referential third person agent and an    SOURCED (Willie p. 39, image)
      unspecified inanimate patient."

IS-7  ROLE: in (24) IMP fills the slot that (25)/(12) fill with a 3rd-person pronoun,   HYPOTHESIS (from IS-3, IS-5)
      the patient subject of the simple passive. In (26) IMP is the patient object.
      In both, IMP = the proto-patient of 'ruin' and is ABS (IS-1).
IS-8  SORT: IMP = [−animate]. Four of Willie's five statements say "inanimate"           HYPOTHESIS (from IS-5..IS-7,
      (§5), and every example has an inanimate undergoer reading. The p. 38             PX-4, PX-5)
      "animate subject" clause has no example in the pages opened.
IS-9  IMP is restricted to the ABS slot: it is never shown as a transitive agent        SOURCED (IS-1: Willie's definition) /
      (that indefinite is INDF 'adi-, p. 36).                                            HYPOTHESIS (as a grammar rule)
IS-10 The two indefinites split by case: INDF = [ERG, human, −def, no π];               HYPOTHESIS (from IS-1, IS-2, IS-8,
      IMP = [ABS, −animate, −def, no π].                                                  PX-1, AP-2)
IS-11 IMP's [−animate] is not a Dowty default: (28) Proto-Patient lists no              HYPOTHESIS (Dowty 1991:572 as
      inanimacy entailment, while INDF's [human] matches (27b) "sentience".             recorded on /maintained-state/)
      So Σ alone does not predict IMP's sort.
IS-12 (94) tsin baah ha'as'na' cannot decide IS-8. *(95) is already excluded by DEF     HYPOTHESIS (from intransitive-pp E4)
      (no definite slot to rank), whatever IMP's animacy.

4. The data

All forms are copied from Willie's page images. Willie's gloss is given first, then a Leipzig line. Willie does not segment the verbs, so the Leipzig lines use dots for unsegmented material. The only hyphen kept is her own, in bi-chidí. "PFV" is not added because Willie glosses tense ("ruined"), not aspect, except "PERF" in (14), which is copied.

(12), p. 57yíchxǫʼ
Willie: 3s.ruined · Leipzig: 3SBJ.ruin
'It is ruined.'
(13), p. 58bi-chidí yíchxǫʼ
Willie: his car 3s.ruined · Leipzig: 3POSS-car 3SBJ.ruin
'His car, it is ruined.'
(14), p. 58[hastiin sání bi-chidí] yíchxǫʼ
Willie: man old 3POSS.car 3s.ruin PERF · Leipzig: man old 3POSS-car 3SBJ.ruin.PERF
'The old man's car is ruined.'
(24), p. 38ʼííchxǫʼ
Willie: unspec. subj. ruined · Leipzig: IMPERS.SBJ.ruin
'Ruining took place.'
(25), p. 38(hatáál) yíchxǫʼ
Leipzig: (ceremony) 3SBJ.ruin
'(The ceremony) it is ruined.'
(26), p. 39ʼasdzą́ą́ ʼííłchxǫʼ
Leipzig: woman IMPERS.OBJ.3SBJ.TR.ruin (Willie: the classifier marks the verb transitive)
'The woman did the ruining (of it).'
(94), p. 99tsin baah haʼasʼnaʼ
Willie: tree 3-on it was climbed · Leipzig: tree 3-on climb.IMPERS
'The tree was climbed.'
*(95), p. 99tsin yaah haʼasʼnaʼ
(no gloss printed)

(12), (24), (25) and (26) contain the same verb, 'ruin'. (12) and (25) are the same form, yíchxǫʼ. The (26) Leipzig line is the one used on indef-person; Willie does not segment the form.

5. Willie's five statements on the sort of ʼa-

#PageWillie's wordsUseSort
S138"This pronoun marks an indefinite inanimate argument."all usesinanimate
S238"When it occurs with an intransitive verb, it marks an indefinite animate subject"intransitiveanimate
S338"when it occurs with a transitive verb, it marks an indefinite inanimate patient"transitiveinanimate
S439"an unspecified inanimate patient" ((26))transitiveinanimate
S539"an indefinite inanimate subject marker 'a-" ((24))intransitiveinanimate

SOURCED S2 and S5 are about the same use (intransitive), and S5 is about Willie's only intransitive example. S1 is a general statement, and S2 contradicts it as written. HYPOTHESIS The simplest reading is that S2's "animate" is a slip: S1, S3, S4 and S5 all say "inanimate", and the next sentence on p. 38 has a slip of its own ("In never permits"). The other reading is that S2 is about a class of intransitives that Willie does not exemplify (for example, verbs whose single argument is an agent). That reading cannot be tested on the pages opened, and it is kept as T77a.

6. Answer to T76a

SOURCED The pieces: (i) IMP is the indefinite absolutive, either an intransitive subject or a transitive object (IS-1, IS-2). (ii) Willie gives (24) as the counterpart of (25). (25) is the verb form of (12), and its subject "has a patient theta role" (IS-3, IS-4). (iii) Every noun she adjoins to that subject is inanimate: hatáál 'ceremony', bi-chidí 'his car', the old man's car. (iv) Her scenario for (24) is a ceremony that has to be stopped or corrected (IS-5). (v) In (26) the same prefix on the same verb is the inanimate patient (IS-6).

HYPOTHESIS So in (24) IMP fills the patient subject slot of the simple passive. That is the slot that hosts 'the ceremony' in (25). 'Ruining took place' means that something got ruined, with no ruiner stated. In both (24) and (26), IMP is the proto-patient of 'ruin' (IS-7). Its sort is [−animate] in both (IS-8). T76a is therefore answered: inanimate, not animate, and not unrestricted as far as the examples go. The status is HYPOTHESIS, because it overrides one of Willie's own sentences (S2) on the strength of the other four and of her examples. "Unrestricted" is not ruled out by the data alone, since the English 'Ruining took place' names no participant. But Willie's scenario, her pairing with (25) and her four "inanimate" statements all point to an inanimate undergoer.

HYPOTHESIS A consequence: in every example the IMP slot is the slot of the affected participant, intransitive or transitive. IMP is never the doer. That is what "absolutive" predicts, and it pairs IMP with INDF, which Willie restricts to the agent of the agentive passive (p. 36) and calls "an indefinite Ergative pronoun" (p. 221). The two indefinites split along the ergative/absolutive line, with a fixed sort on each side (IS-10).

7. What (94) can and cannot show

SOURCED Willie calls the verb of (94) "a simple passive, an Impersonal" (p. 99). With an animate subject, a one-participant verb takes the grammatical yiiʼ ((90), "when the subject is animate", p. 98). With an inanimate one it takes biiʼ ((89)). (94) takes baah and excludes yaah (*(95)).

HYPOTHESIS This looks like a test of IMP's animacy, but it is not a clean one. intransitive-pp E4 already rules out *(95) because the impersonal has no definite slot that could be ranked (DEF). That holds whatever IMP's animacy is. So (94) is consistent with IS-8 (an animate IMP subject would be expected to allow yaah under p. 98 alone), but it does not decide the question (IS-12). Two more caveats. Willie does not segment haʼasʼnaʼ, and that it contains the ʼa- of §2.3.2 rests on its spelling against (96) haasʼnaʼ (as on intransitive-pp). And if IMP is the patient subject of a simple passive, it cannot be the tree, which is the postposition's object. What IMP refers to in (94) is therefore a new question (T77b).

8. Rulesets A and B HYPOTHESIS

StepLevelRulesetWhat happens
1vPΣ / BΣ assigns roles. In the simple passive of 'ruin' there is one role, proto-patient, and no agent slot. In the transitive there are two.
2vPBCase: the proto-patient slot is ABS (the intransitive subject or the transitive object). The agent of a transitive is ERG.
3WordAAgree copies [π] into the prefix positions from the slots that have one. An ABS slot that is [−def] and has no π is spelled out as ʼa- (IS-9, IS-10); an ERG slot of that kind is spelled out as ʼadi-.
4WordASpell-out: prefixes left of the stem (left-headed). The word closes. This is the changeover A→B.
5ClauseBLINK: a noun adjoined on the left is linked to a slot that has [π] (PX-11). The IMP slot has none, so no noun can link to it, which is why (24) has no noun.

The point for Focus 1: the choice between ʼa- and ʼadi- is made in A, but what it reads is the case of the slot (ABS vs ERG), which is set by the role at the vP level. B then reads [π] for LINK. So the word level spells out a clause-level role/case distinction. Here A is the consumer of B's information, the reverse of the LINK handoff on indef-person HYPOTHESIS.

9. Derivations

D1  (24) 'ííchxǫ' 'Ruining took place.' (Willie p. 38)
1  Merge  [VP IMP[−def, −animate, no π]  ruin ]             Σ: IMP = proto-patient (IS-7)
2  Merge  [vP [VP ...] v(simple passive) ]                    B; no agent slot (IS-3); VPISH: the sole
                                                              argument starts VP-internal
3  Case   IMP = ABS (intransitive subject)                    B; IS-1
4  Move   IMP → subject position                             B; Willie: "unspec. subj."
5  Agree  subject slot: no π to copy                         A; PX-9
6  Spell  'a- (fixed shape for [ABS, −def, no π]) + stem     A; left-headed; not segmented further
7  LINK   none: IMP has no π                                 B; PX-11

D2  (25) (hatáál) yíchxǫ' and (12)–(14) (Willie pp. 38, 57–58)
1  Merge  [VP pro[3]  ruin ]                                 Σ: pro = proto-patient (IS-3, IS-4)
2  Merge  [vP [VP ...] v(simple passive) ]; Move pro → subj.  B; VPISH
3  Agree  subject slot ← [3]                                 A
4  Spell  yíchxǫ'                                            A; not segmented
5  LINK   hatáál / bi-chidí / [hastiin sání bi-chidí] ⟶ pro[3] B; optional; adjoined left

D3  (26) 'asdzą́ą́ 'ííłchxǫ' 'The woman did the ruining (of it).' (Willie p. 39)
1  Merge  [VP IMP[−def, −animate, no π]  ruin ]             Σ: IMP = proto-patient (same slot as D1)
2  Merge  [vP pro[3] [VP ...] v(TR) ]                         Σ: pro = proto-agent; B; VPISH: agent in Spec,vP
3  Case   IMP = ABS (transitive object); pro = ERG           B; IS-1, IS-10
4  Move   pro[3] → subject                                   B
5  Agree  subject ← [3]; object slot: no π                   A
6  Spell  'a- + ... + ł (CL) + stem                          A; Willie: the classifier marks transitive
7  LINK   'asdzą́ą́ ⟶ pro[3] only                             B; PX-11

D4  Schematic (no Navajo form): an ERG slot with [−def, no π]
1  Merge  [vP INDF [VP pro[π] V ] v ]                         Σ: INDF = proto-agent
2  Case   INDF = ERG                                         B
3  Spell  'adi- (fixed shape for [ERG, −def, no π])          A; see /passive-rank/ for Willie's forms
          IS-10 predicts that IMP never appears in this ERG slot and INDF never in an ABS slot.
          IS-9 is Willie's own definition for IMP; no counterexample was found in the pages opened.

D1 step 1 gives IMP its [−animate] value from IS-8, which is a HYPOTHESIS. The derivations do not depend on it except at step 1. D4 is schematic on purpose: it states the rule for the ERG side without copying a form. The sourced INDF forms are on passive-rank and third-voice.

10. The role schema (Focus 2)

The two indefinite prefixes are as close as Navajo comes to an argument that is only a role. Each one fills one side of the case split, can never take a noun, and has no person. This run adds that each one's content is predictable from its slot: ERG → human (ʼadi-), ABS → inanimate (ʼa-) HYPOTHESIS. That looks like support for "roles, not words": the verb combines with a role, and the role fixes the filler's sort.

It does not go all the way, for two reasons. (i) Only half of the pattern follows from Dowty's role schema. INDF's [human] matches Dowty's proto-agent property (27b) "sentience (and/or perception)", but Dowty's (28) Proto-Patient list has no inanimacy property, and patients can be human. So IMP's [−animate] is a lexical fact about ʼa-, not a consequence of Σ (IS-11). (ii) The restriction runs through case (ABS/ERG), and case is a grammatical category, not a role. In (24) and (26) the same role (the patient of 'ruin') gets ABS in both, so these examples cannot separate the two. Either way, what the verb combines with is a role plus a feature bundle that includes at least one lexically fixed value. The weak thesis is supported again. The strong thesis stays OPEN, and Grant's revised view that it does not hold cleanly remains GRANT'S CONJECTURE.

11. Fidelity notes

All Navajo forms are copied from the page images. (12) yíchxǫʼ on p. 57 has the hook under the o, as in (25). In (14) Willie has a hyphen in bi-chidí and the gloss "3POSS.car", while (13) glosses the same word "his car". Both are copied. Willie writes ʼa, -ʼa and ʼa- for the same prefix (pp. 38, 39, 221), and all are rendered here as ʼa-. Willie does not segment (24), (26) or (94). The derivations say "not segmented", and the IMP in (94) rests on spelling plus her word "Impersonal" (HYPOTHESIS, as on intransitive-pp). The phrase "affected participant" in §6 is ours, not Willie's. No forms, judgments or segmentations were invented.

12. Open questions

IDQuestionWhat would settle it
T77aDoes any intransitive verb whose single argument is an agent take ʼa- (the use S2 might describe)? If one does, is the unspecified participant animate?Young & Morgan 1987 (around p. 120, which Willie cites for the simple passive); Willie 1990. OPEN
T77bIn (94) tsin baah haʼasʼnaʼ, if IMP is the patient subject of a simple passive, what does it refer to, given that the tree is the postposition's object? Could it be non-referential, like the subject of an impersonal passive?Willie's or Young & Morgan's segmentation of haʼasʼnaʼ; more ʼa- simple passives with a PP. OPEN
T77cIS-10 predicts that IMP never fills an ERG slot and INDF never an ABS slot. Is there any form with an indefinite intransitive subject that is human (and so would need INDF in an ABS slot)?Willie ch. 2 and ch. 4 on images; Young & Morgan 1987. OPEN
T77dIs IMP's [−animate] read by the animacy ranking (70), or only an interpretive restriction? (This sharpens T76c.)A form in which IMP and a definite 3rd-person argument could alternate between Direct and Inverse. OPEN

T76a answered as HYPOTHESIS (IS-7, IS-8: inanimate, the ABS proto-patient). T76c sharpened as T77d, and (94) does not decide it (IS-12). Still open: T76b, T76c, T76d, T75b, T75c, T73b, T73c, T73d (direct evidence), T73e, T74a–c, T74e. Sources: Willie 1991 pp. 36, 38–39, 57–58, 96–100 (page images, this run). See also indef-person, passive-rank, intransitive-pp, ergative-indefinite, keep vs bring about, open questions, sources.

Algonquian's impersonal actor X has its own paradigm cells but no [PERS] feature, like Navajo ʼadi-; but where Algonquian also denies person to inanimates, Navajo gives 'it' a person (T76d)

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

Restated on the unified structure (2026-11-15). The indefinites (ʼa-, ʼadi-, areal ho-) are restated as personless pronouns in ordinary argument positions (H60): ʼadi- as the silent indefinite agent in Spec,vP that raises to Spec,TP by the default (H10b), ʼa- in VP. The 3i data on this page have not yet been individually restated. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §4, indefinite and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

T76d asked whether the Algonquian impersonal actor X ('people, someone') belongs to a person paradigm or sits outside it, as Navajo ʼadi- does (Willie pp. 221–222). This run opened Oxford 2022 (lingbuzz 006486), which the site had not used before, plus two other Oxford papers for side points. The answer has two parts. X has its own cells and its own morphemes in the verb paradigm (Table 3), but in Oxford's feature analysis it lacks [PERS]: 3IMPERS = [ϕ, AN] (23b). So it sits in the paradigm but outside the person category, as ʼadi- does SOURCED (Oxford's analysis) HYPOTHESIS (the parallel). The difference: Oxford's 3INAN 'it' also lacks [PERS], while Willie glosses Navajo 'it' as 3s. So in Algonquian being personless goes with animacy, and in Navajo with indefiniteness HYPOTHESIS. The strong roles-not-words thesis stays OPEN.

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 Answer to T76d · 6 Where the parallel breaks · 7 Ranked high, specified low · 8 Rulesets A and B · 9 Derivations · 10 The role schema (Focus 2) · 11 Fidelity notes · 12 Open questions

Notation is the same as on absolutive-indef and indef-person. No Syntax · Semantics · Interface conventions file is in the repo or the context folder. A = the word level (prefixal, left-headed). B = the clause level (right-headed). Σ = the role schema, with Dowty's proto-roles as the baseline. [π] = a Navajo person value. INDF = ʼadi-. IMP (Navajo) = ʼa-. In Oxford's glosses IMP means the Algonquian impersonal and X:Y means "X acts on Y"; Algonquian forms are in plain italics, not tagged as Navajo. Merge builds structure (puts two pieces together). Agree copies feature values such as person and number from a goal onto a probe. Oxford's version of Agree (after Deal 2015) splits a probe's conditions in two. Its interaction condition says which goals it may copy from. Its satisfaction condition says which goal makes it stop searching.

1. Question

indef-person found that Willie lists ʼadi-, ʼa- and ho- "in addition to the person paradigms" (PX-1, p. 221), so all three are personless. ergative-indefinite §7 had called Algonquian X "a person category on the theme-sign hierarchy", based on its rank in "1/2>X>3>3’>0" (Oxford, Routledge chapter, PDF p. 17). T76d asks which of these X really is: a person value, or a personless sort like ʼadi-.

2. Sources

SOURCED Oxford, Will 2022. "Probe specification and agreement variation: Evidence from the Algonquian inverse." February 2022 manuscript, lingbuzz 006486, free PDF, 45 pages; the PDF page numbers match the printed ones. The text was read from the PDF's text layer. The PDF is born-digital (typeset, not scanned), so its text layer is the publisher's own text. Read this run: p. 18 (Table 3 and fn. 15), p. 19 (Table 4), pp. 20–21 (Table 5 and the cline), p. 23 (22), p. 24 (23), p. 26 (C3), p. 28 (E6), p. 33 (how (22) is derived).

SOURCED Oxford, Will 2018. "Inverse marking and Multiple Agree in Algonquin." NLLT, author's version, p. 17 fn. 13 only. Oxford, Will 2023. "A tale of two inverses." February 2023 manuscript (to appear in Syntax), lingbuzz 006567, p. 15 fn. 8 only. Both from the PDFs' text layers.

Already on the site and not re-opened: Willie 1991 p. 221 (PX-1, from page images on indef-person); p. 37 (21) niʼdilzhééh (from page images on passive-rank); p. 57 (12) yíchxǫʼ (from page images on absolutive-indef); Oxford's Routledge chapter PDF pp. 17–18 (on ergative-indefinite).

Not opened: Goddard 1979 and 2007 (Oxford's sources for Table 3), Goddard 1969, Dryer 1996, Hockett 1996, Wolvengrey 2011, McGinnis 2008, Deal 2015.

3. Rules first

AX-1  Feature bundles (23b): 3INAN [ϕ]; 3IMPERS [ϕ, AN]; 3ANIM [ϕ, AN, PERS];        SOURCED (Oxford 2022 p. 24)
      1 [ϕ, AN, PERS, PART]; 2 [ϕ, AN, PERS, PART, ADDR]; plus [PL] as relevant.
      The geometry (23a) puts [PERSON] under [ANIMATE].
AX-2  "[PERS], a feature that the impersonal agent lacks." Inanimates lack it too.    SOURCED (Oxford 2022 p. 33)
AX-3  X has its own paradigm cells and morphemes: 3IMP→1PL rows in Tables 3–4,        SOURCED (Oxford 2022 pp. 18–19)
      with -en IMP (conjunct) and -ekoː / -ikawi INV.IMP.
AX-4  Implicational hierarchy (22a) "animate > impersonal > inanimate": if             SOURCED (Oxford 2022 p. 23)
      3AN→SAP is inverse, so are 3IMP→SAP and 3IN→SAP; if 3IMP→SAP is inverse,
      so is 3IN→SAP.
AX-5  C3: "If the interaction condition is changed to [PERS], Infl will not interact   SOURCED (Oxford 2022 p. 26)
      with nominals that lack [PERS]—i.e., inanimates and impersonals." So the
      verb goes inverse with those agents.
AX-6  Fn. 15: whether X forms are transitives with an impersonal agent or             SOURCED (Oxford 2022 p. 18)
      "detransitivized passives" is "controversial" among Algonquianists; Oxford
      follows "the most standard view": a transitive "indefinite actor" construction.
AX-7  "First, second, and impersonal third persons do not participate in or trigger   SOURCED (Oxford 2018 p. 17 fn. 13)
      obviation" (Algonquin).
AX-8  Answer to T76d: X is IN the paradigm (cells, morphemes, a rank) but OUTSIDE     SOURCED as Oxford's analysis
      the person category in the feature sense (no [PERS]).                           (AX-1–AX-3)
AX-9  Parallel: Navajo INDF [ERG, human, −def, no π] ≈ Algonquian 3IMPERS [ϕ, AN].     HYPOTHESIS (our comparison)
      Both are animate-sort bundles with no person, and both have their own
      morpheme in the verb.
AX-10 Contrast: Algonquian 3INAN is referential ('it') and has no [PERS]. Navajo     HYPOTHESIS (Willie's 3s gloss SOURCED,
      referential 'it' is glossed 3s (Willie p. 57 (12)). Navajo IMP 'a- is            p. 57; the reading is ours)
      inanimate AND indefinite. So Navajo ties personlessness to indefiniteness,
      Algonquian (in Oxford's analysis) ties it to the animacy scale.
AX-11 Fn. 15's passive-vs-indefinite-actor debate mirrors Willie's two labels for     HYPOTHESIS (both labels SOURCED)
      'adi-: the "Agentive Passive" construction vs an "indefinite Ergative pronoun"
      (p. 221).
AX-12 Tension: X ranks above 3 on the independent theme-sign hierarchy (Routledge,    OPEN (T78a)
      1/2>X>3), but has fewer features than 3ANIM (AX-1) and makes a verb inverse
      more readily than an animate (AX-4).

4. The data

Algonquian forms are copied from Oxford's Tables 3 and 4 with his segmentation and glosses. Table 3 is reconstructed Proto-Algonquian (Oxford: "inflections follow Goddard 1979:88, 2007:267"), not attested speech. The Navajo forms are copied from earlier pages that were checked against Willie's page images.

T3, PA, conj.seːkihinamenki
seːkih -i -en -amenk -i · scare -1OBJ -IMP -3:1PL -IND
'someone scared us' (conjunct, neutral)
T3, PA, ind.neseːkihekoːhmenaː
ne- seːkih -ekoː -ehmenaː · 1- scare -INV.IMP -1PL
'someone scared us' (independent, inverse)
T3, PA, ind.neseːkihekonaːna
ne- seːkih -ekw -wenaːn -a · 1- scare -INV -1PL -3AN.SG
'she scared us' (independent, inverse; animate agent, for comparison)
T3, PA, ind.neseːkihekoneːnaːni
ne- seːkih -ekw -eneːnaːn -i · 1- scare -INV -1PL -3IN.SG
'it scared us' (independent, inverse; inanimate agent)
T4, Plains Cree, conj.seːkihikawiyaːhk
seːkih -ikawi -yaːhk · scare -INV.IMP -1PL
'someone scared us' (conjunct, inverse)
T4, Kickapoo, conj.θeːkihiːnameki
θeːkih -i -en -amek -i · scare -1OBJ -IMP -3:1PL -IND
'someone scared us' (conjunct, neutral)
(21), p. 37niʼdilzhééh (ni-ʼadi-l-zheeh in underlying form)
Willie: 2s.PASS.shaving · Leipzig: 2SG.OBJ-INDF.AGT-CL-shave
'You are being shaved.'
(12), p. 57yíchxǫʼ
Willie: 3s.ruined · Leipzig: 3SBJ.ruin
'It is ruined.'

Oxford places Plains Cree in C5 and Kickapoo in C1 (pp. 20–21, Table 5). C1 has no inverse in any conjunct 3→SAP form; C5 has it with all inanimate and impersonal agents and with plural SAP patients. In a plural cell such as 'us', then, Plains Cree goes inverse by two routes at once (§9 D2).

5. Answer to T76d

  • In the paradigm. X has its own rows in the inflection tables (3IMP→1PL in Tables 3 and 4) and its own exponents: conjunct -en IMP, independent -ekoː INV.IMP, Plains Cree -ikawi INV.IMP (AX-3). It also has a slot on the theme-sign ranking (Routledge; AX-12) SOURCED.
  • Outside the person category. In Oxford's features it is [ϕ, AN], with no [PERS] (AX-1, AX-2). A probe that interacts only with [PERS] skips it, as it skips inanimates (AX-5). Algonquin obviation, which sorts referential third persons, also leaves it out (AX-7) SOURCED (Oxford's analysis).
  • So T76d gets a split answer: in the paradigm as a cell, outside it as a feature. This corrects the wording of ergative-indefinite §7. There X is called a "person category". It is a ranked category, but in Oxford 2022's features it is not a person value. That page is a dated working note and is left as it was.
  • Navajo ʼadi- matches on both points. It has its own fixed prefix (ni-ʼadi-l-zheeh) but stands "in addition to the person paradigms" (PX-1). And Willie gives it a sort value ("only in reference to humans") SOURCED (Willie p. 221). That INDF and X are the same kind of object, an animate sort with no person, is our comparison HYPOTHESIS (AX-9).

6. Where the parallel breaks

In Oxford's geometry, [PERSON] hangs under [ANIMATE] (23a), so nothing inanimate can have person. Algonquian 0 'it' is a referential argument ("a referential inanimate ('it')", §4.1, p. 18), but it is still [ϕ] only. Navajo is different. Willie glosses the referential 'it' of (12) yíchxǫʼ as "3s", a value in the third-person paradigm SOURCED (p. 57). In Navajo the arguments with no person are the indefinites (ʼadi-, ʼa-) and the areal ho-, whatever their animacy. So the two languages draw the line between having person and lacking it in different places: Algonquian along the animacy scale, Navajo between definite and indefinite HYPOTHESIS (AX-10). It follows that Navajo IMP ʼa- ([−animate], no π) matches Oxford's 3INAN only partly. Both are inanimate and personless, but 3INAN is referential and IMP is indefinite.

A second parallel is in the analysts' terms (AX-11). For Algonquian X, Oxford notes a standing dispute: transitive with an impersonal agent, or a detransitivized passive (fn. 15)? Oxford 2023 fn. 8 reports Dahlstrom's description of Plains Cree as having an "agentless impersonal form" SOURCED. Navajo literature has the same split: the construction is the "Agentive Passive", but Willie calls ʼadi- an "indefinite Ergative pronoun" (p. 221) SOURCED. That the two disputes are the same question is HYPOTHESIS. Nothing opened this run decides either one.

7. Ranked high, specified low

The Routledge chapter gives the independent-order theme-sign hierarchy as 1/2>X>3>3’>0, so X outranks animate 3. Oxford 2022 gives X fewer features than 3ANIM, and (22a) shows a 3IMP agent triggering the inverse more readily than a 3AN agent. In the conjunct, X behaves like a weak argument, near the inanimates. Both papers are Oxford's own, and they describe different parts of the verb (independent theme signs vs conjunct Infl agreement), so this is not a contradiction. But the site cannot yet say how a rank above 3 can come from fewer features. Answering that needs X→3 forms, which neither paper gives OPEN (T78a). This bears on EI-8 and T70e, which used the X>3 rank.

8. Rulesets A and B

StepAlgonquian X (Oxford)Navajo INDF (Willie, via passive-rank)
Merge of the agentSpec,VoiceP; clause syntax B[vP INDF …]; clause syntax B
AgreeVoice → patient; Infl probes into VoiceP (the PIC limits it to Voice's head and specifier). Agree as Oxford models it; how the inflected word is then assembled is not discussed on the pages readobject slot ← patient's π; agent slot gets no π A
What the probe reads[AN], [PERS] (sort and person features), not the roleπ for agreement; [human] for the ranking (AP-1–AP-10)
Spell-outsuffixal theme sign and central suffix (-ekoː, -ehmenaː)left-headed prefix array (ni-ʼadi-l-zheeh) A

In both languages the decision about the agent is made before or at the word, and nothing in B (Navajo LINK) can give X or INDF a noun. In Navajo this is PX-11; for Algonquian it is T78c. Algonquian exponents are mostly suffixes, so the site's A = "prefixal, left-headed" label is a Navajo label. The Algonquian column only marks which steps happen inside the word.

9. Derivations

D1  Kickapoo (C1) θeːkihiːnameki 'someone scared us' (Oxford Table 4)
    Infl: [int: ϕ], no SAP satisfaction condition
1  Merge  [VP scare 1PL]                                     B
2  Merge  Voice; Agree Voice ↔ 1PL  → -i 1OBJ                 B (syntax); Oxford's Voice = theme sign
3  Merge  X [ϕ, AN] in Spec,VoiceP                           B
4  Merge  Infl [int: ϕ]; Agree: X has [ϕ] → interacts;       B (syntax)
          Infl copies from both arguments (neutral)
5  Spell  -en IMP (X) + -amek 3:1PL + -i IND                 word: suffixes
6  Output θeːkihiːnameki (neutral)                           matches Table 4 (C1 = Kickapoo, Table 5)

D2  Plains Cree (C5) seːkihikawiyaːhk 'someone scared us' (Oxford Table 4)
    Infl: [int: PERS, sat: PART+PL] (C5 = C3's interaction plus a plural-SAP condition)
1  Merge  [VP scare 1PL]                                     B
2  Merge  Voice; Agree Voice ↔ 1PL                           B
3  Merge  X [ϕ, AN] in Spec,VoiceP                           B
4  Agree  Infl [int: PERS]: X lacks PERS → no interaction    B; AX-2, AX-5
          (and the 1PL patient's [PART, PL] already satisfies
          the probe: the second route)
5  Agree  Infl copies only the patient's features (via Voice)
6  Spell  Voice with no person to show → elsewhere           Oxford (15): -ikw "elsewhere ('inverse')";
          form, here -ikawi INV.IMP; -yaːhk 1PL                 -ikawi is Oxford's INV.IMP gloss
7  Output seːkihikawiyaːhk (inverse)
          Step 4 has two routes, so this cell does not by itself show that
          the [PERS] condition is at work. A C3 language with a singular 'me'
          patient would. Oxford gives no such form this run. Schematic only.

D3  Navajo (21) ni'dilzhééh 'You are being shaved.' (Willie p. 37), read through Oxford
1  Merge  [VP pro(2sg) shave]                                B; head-right VP
2  Merge  [vP INDF[hum, −def, no π] [VP …] v]                B; VPISH
3  Agree  object slot ← [2sg]                                A
          agent slot: INDF has no π, so nothing is copied    A; AX-9 says this matches C3 step 4
4  Spell  ni- + 'adi- + l + zheeh                            A; prefix array
5  Output ni'dilzhééh
          No Direct/Inverse choice here (VC-13: a 2nd-person argument is present).
          So Navajo has no counterpart to the Algonquian inverse for INDF→SAP;
          the parallel is in the features only, not the voice morphology. HYPOTHESIS

D1 and D2 apply Oxford's model to forms he gives, using the settings he assigns to C1 and C5. He gives the model; we wrote out the steps. D3 repeats passive-rank D1 with step 3 restated in Oxford's terms. It adds no Navajo data.

10. The role schema (Focus 2)

Oxford's Table 3 holds the role fixed: the agent of 'scare' acting on 'us'. Only the agent's sort varies across the 3AN, 3IMP and 3IN rows, and the agreement changes with it. So in Algonquian, the thing the probe reads is a feature bundle ([AN], [PERS]), not the theta role SOURCED (the design of Tables 3–5 and C3). The Navajo pages point the same way: π for agreement, [human] for the ranking (passive-rank). This is the weak thesis again, in a second language family. Arguments that are close to a bare role (X, INDF, IMP) are still built from sort features plus a missing person, and the grammar reads those features HYPOTHESIS. It does not show that verbs select roles rather than words. Indefinites have no noun to select in any case, so they cannot decide that question. The strong thesis stays OPEN. Grant's revised view (the claim "does not hold cleanly") stays GRANT'S CONJECTURE.

Algonquian also gives a test the site lacked: a role-only argument that still gets ranked, and ranked high (AX-12). If roles alone drove the ranking, X and 3ANIM agents (both agents) should rank the same. The independent hierarchy ranks them differently, and the conjunct cline (22a) treats them differently SOURCED. What does the ranking, then, is sort and person features, not the role HYPOTHESIS.

11. Fidelity notes

The Algonquian forms are copied from the text layer of Oxford's born-digital PDF. The long vowels are written with ː (U+02D0), as in the PDF. Oxford gives Table 3 as Proto-Algonquian after Goddard, so those forms are reconstructions. The Navajo forms are copied from passive-rank and absolutive-indef, which transcribed them from Willie's page images; they were not re-opened this run. No Navajo form, segmentation or judgment is new on this page. The Navajo glosses for ʼadi- ([human], no π) and ʼa- ([−animate]) carry their earlier statuses: SOURCED for ʼadi- (Willie p. 221) and HYPOTHESIS for ʼa- (IS-8).

12. Open questions

IDQuestionWhat would settle it
T78aX ranks above 3 on the independent theme-sign hierarchy (1/2>X>3) but is [ϕ, AN] without [PERS], and (22a) puts it below animate. Is an X→3 form direct or inverse? (This also bears on EI-8 and T70e.)X→3 forms in Goddard 1979 or an Ojibwe/Cree grammar; Oxford's other papers. OPEN
T78bAX-10: does Navajo [π] really include inanimates as a grammatical value, or is Willie's "3s" for 'it' only a gloss label? Is there a sourced Navajo form where an inanimate agent acts on a 1st/2nd person, to set beside Oxford's 3IN→SAP cells?Willie ch. 2–3 and ch. 6 on images; Young & Morgan 1987. OPEN
T78cCan Algonquian X ever be doubled by an overt nominal (cf. PX-11: in Navajo a noun links only to a slot with a person value)? Oxford 2018 fn. 13 leaves X out of obviation. Does Navajo INDF likewise stay out of the 3rd/4th-person (obviation) system?Algonquian grammars on indefinite-actor forms; Willie ch. 9 (fourth person) on images. OPEN
T78dIs there a Navajo test that tells a present but personless agent from no agent at all? This is the Algonquian fn. 15 dispute (Hockett vs Dryer) applied to Willie's "Agentive Passive".Goddard 1969, Dryer 1996 (for the Algonquian tests); Willie pp. 36–38 on images. OPEN

T76d answered: in the paradigm as a cell, outside it as a feature (AX-8, SOURCED as Oxford's analysis). The Navajo parallel is a HYPOTHESIS (AX-9), with a sourced contrast for inanimates (AX-10, HYPOTHESIS). Still open: T76b, T76c, T77a–d, T75b, T75c, T73b–e, T74a–c, T74e, T70e. Sources: Oxford 2022 pp. 18–28, 33; Oxford 2018 p. 17; Oxford 2023 p. 15 (all from the PDFs' text layers). See also indef-person, absolutive-indef, ergative-indefinite, passive-rank, The inverse.

Young & Morgan's "3i subject" ʼa- is 'someone' as often as 'something': people go, work and dream with it, so the indefinite ʼa- is not only an inanimate patient (T77a, T77c, T77b, T33e)

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

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

Restated on the unified structure (2026-11-15). The indefinites (ʼa-, ʼadi-, areal ho-) are restated as personless pronouns in ordinary argument positions (H60): ʼadi- as the silent indefinite agent in Spec,vP that raises to Spec,TP by the default (H10b), ʼa- in VP. The 3i data on this page have not yet been individually restated. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Position-class template statements below describe the order of prefixes; they are not the derivation (§3). The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §4, indefinite and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

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.

T77a asked whether any Navajo intransitive verb whose single argument is an agent takes ʼa-, and whether that participant is animate. Willie says so once (p. 38, "an indefinite animate subject", S2 on absolutive-indef), but gives no example, and the last page therefore made ʼa- [−animate] (IS-8). This run opened Young & Morgan 1987 (The Navajo Language, dictionary part) for the first time on the site, through the University of Colorado searchable edition. Y&M define ʼa- in Position V as the "3i subj. pron., also with the meaning someone, something indefinite" SOURCED. They give it with human referents on agentive intransitives: 'everyone ... is going to Gallup', 'the passengers got into the bus', 'work is being done = people are busy', 'an indefinite subject dreams' SOURCED. So T77a is answered yes, and IS-8 is too narrow HYPOTHESIS. T77c is answered too: human indefinite intransitive subjects exist, but they are spelled ʼa-, not ʼadi-, which Y&M keep for the agent of the agentive passive. So IS-10's case split survives and its sort split does not HYPOTHESIS. One Y&M sentence, níłtsą́ ʼadeezhjool 'a rainstorm has started moving', puts an overt noun next to a 3i subject. That is contradicting data for PX-11 and Willie's "indefinite pronouns exclude adjoined nominals" OPEN. The strong roles-not-words thesis stays OPEN.

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 Answers to T77a and T77c · 6 (94) again, and the climbing verb (T77b, T33e) · 7 A noun beside a 3i subject (against PX-11) · 8 Rulesets A and B · 9 Derivations · 10 The role schema (Focus 2) · 11 Algonquian X compared · 12 Fidelity notes · 13 Open questions

Notation is the same as on absolutive-indef. No Syntax · Semantics · Interface conventions file is in the repo or the context folder. A = the word level (prefixal, left-headed). B = the clause level (right-headed). Σ = the role schema, with Dowty's proto-roles as the baseline. IMP = Willie's impersonal ʼa-; INDF = the ʼadi- agent. 3i is Young & Morgan's label ("3i subj. pron.", "3i obj. pron."). Pos. IV and Pos. V are their prefix positions (IV = direct object, V = deictic subject). [π] = a person value. ABS = Willie's "absolutive" (intransitive subject or transitive object). Merge builds structure: it joins two pieces into a bigger one. Agree builds nothing. It copies feature values such as person and number from one place to another. Link is Willie's relation between an adjoined noun and a pronoun (a ruleset-B operation, not Agree).

1. Question

absolutive-indef concluded that in every example Willie gives, ʼa- stands for an inanimate thing affected (IS-7, IS-8), and that her p. 38 "animate subject" clause had nothing behind it. It left four questions. T77a: does any intransitive agent verb take ʼa-, and is the participant animate? T77b: in (94) tsin baah haʼasʼnaʼ the tree is the postposition's object, so what does IMP refer to? T77c: is there a human indefinite intransitive subject, which IS-10 ("INDF = ERG, IMP = ABS and −animate") has no place for? T77d: is IMP's [−animate] read by the ranking? The source named for all four was Young & Morgan 1987, which Willie herself cites for the simple passive (p. 57, "Young and Morgan (1987:120)").

2. Sources

SOURCED Young, Robert W. & William Morgan, Sr. 1987. The Navajo Language: A Grammar and Colloquial Dictionary, revised edition. Albuquerque: University of New Mexico Press. Read through the free searchable dictionary at the University of Colorado (verbs.colorado.edu/~mahu0110/ym/, word + definition search). Each entry there carries a page number ("pg") and a definition number ("def#"). These are cited below as "Y&M p. N def M" as the interface gives them. They were not checked against the printed book (§12). Entries used: ʼa- (pg 116 def 5); ʼadi- (p. 21 def 6); ʼadijooł (p. 16 def 15); ʼaleeh (p. 86 def 10); bidiʼni- (p. 178 def 3); biʼdi- (p. 212 def 3); bíʼni- (p. 215 def 12); bił (p. 223 def 1); bíníshchééh (p. 239 def 4); daʼdildééh (p. 303 def 3); daʼiildééh (p. 303 def 21); haashʼnééh (p. 434 def 8); ʼídlį́ (p. 465 def 21); naʼanish (p. 549 def 24); naʼiidzeeł (p. 555 def 17); neishheeł (p. 623 def 9); nááʼa- (p. 581 def 14); naʼniltin (p. 558 def 3); taaʼa- (p. 704 def 5); kíhonii- (p. 493 def 33); and the climbing sentences under dishłeeh (p. 337), hadaashneʼ (p. 380 def 12), yishkʼai (p. 785 def 3) and yishło (p. 788 def 4). The grammar part (including p. 120 on the simple passive and p. 67 on 3i) is not in the interface and was not opened.

SOURCED Hale, Ken 2001. "Navajo verb stem position and the bipartite structure of the Navajo conjunct sector." Linguistic Inquiry 32: 678–693 (text file on the box, as on earlier pages). Re-read for position V: "The prefix glossed 3i is an indefinite subject marker, also construed with the third person. It functions autonomously as a subject but also in the so-called agentive passive (not a true passive (see Neudorf 2000), but rather a fully transitive impersonal subject construction)."

Already on the site and not re-opened: Willie 1991 pp. 36–39, 57–58, 97–100 and the S1–S5 table (absolutive-indef, intransitive-pp, from page images); Willie p. 221 and PX-7/PX-11 (indef-person); Willie p. 124 and IMP-4 (fourth-uses); Oxford 2022 AX-1, AX-9 (impersonal-x).

3. Rules first

TI-1  Y&M 'a-: "Pos. IV, 3i direct obj. pron., something or someone indefinite, and     SOURCED (Y&M pg 116 def 5)
      Pos. V, 3i subj. pron., also with the meaning someone, something indefinite."
      It "takes variable shape" ('ídiz, 'iidiz, 'ohdiz, 'e'estsééh, 'o'oolwod) and
      "occurs as a thematic prefix in some verb bases" ('ashzhish 'dance', 'ashhosh
      'sleep').
TI-2  Y&M 'adi- in bidi'ni-: "'adi-, Pos. V, 3i subj. pron., some person. -'Adi-          SOURCED (Y&M p. 178 def 3;
      represents the subject (agent) in the Agentive Passive (Passive B)." bi'di-        p. 212 def 3)
      "forms Passive B". Elsewhere 'adi- is also 'a- (Pos. V, "something") + di-
      (Pos. VIa) (p. 21 def 6).
TI-3  HUMAN 3i SUBJECTS of agentive intransitives, with Y&M's own breakdown "'a-,        SOURCED (Y&M p. 303 defs 3, 21;
      Pos. V, 3i subj. pron., someone": da'dildééh 'to start (to go) (plural            p. 549 def 24; p. 555 def 17;
      indefinite subjects)', da'iildééh 'to go (plural indefinite actors ...)';         p. 465 def 21; p. 86 def 10)
      also na'anish (work; "3i subject 'a-, someone"), na'iidzeeł ("an indefinite
      subject dreams"), 'ídlį́ (be; "3i subject, someone"), 'áłah 'aleeh (a meeting;
      "indefinite subjects become. Lit. people become together").
TI-4  INANIMATE 3i SUBJECTS: natural-force movers with 'adijooł (Y&M: "One would         SOURCED (Y&M p. 16 def 15;
      expect the subject pronoun to be the ordinary 3rd person ... Ø. However, when       p. 223 def 1; p. 558 def 3;
                                                                                       p. 581 def 14; p. 704 def 5)
      the subject is one of those entities listed above, it is represented in the
      verb by a 3i subject pronoun 'a-, something indefinite"); a vehicle ("The
      vehicle may also be represented by the 3I subject pronoun 'a-, something");
      a horse being ridden (na'niltin, "Lit. something gallops around with him");
      the sun (náá'a-, taa'a-).
TI-5  With na'anish Y&M "note deletion of l-classifier" against naashnish 'to work'.      SOURCED (Y&M p. 549 def 24)
TI-6  'Climb (a tree)' is haash'nééh 'to crawl up out, ... to climb up on hands and      SOURCED (Y&M p. 434 def 8;
      knees', d-classifier, an intransitive with a postpositional object: "Bąąh          pp. 337, 380, 785, 788)
      haash'nééh, to climb it (a tree). Tsin bąąh hasis'na', I climbed the tree."
      With 3rd-person climbers the postposition is y-: tsé yąąh haas'na', ńdíshchíí'
      yąąh haas'na', kin yighą́ą́' haas'na'. No 3i form of this verb was found.

TI-7  T77a: yes. Agentive intransitives take 3i 'a- with human referents (TI-3).      SOURCED (Y&M forms and glosses) /
      Willie's S2 ("an indefinite animate subject") now has examples, from Y&M.       HYPOTHESIS (that Willie's IMP is
                                                                                       Y&M's 3i 'a-; same shape, slot, cite)
TI-8  IS-8 (IMP = [−animate]) is wrong as a general rule. The sort of 'a- is not       HYPOTHESIS (from TI-1, TI-3, TI-4)
      fixed by the prefix ("someone, something"); the verb or the context fixes it.
      IS-7 (IMP = proto-patient in Willie's 'ruin' forms) stands as it was.
TI-9  T77c: human indefinite intransitive subjects exist, and they are 'a-, not        HYPOTHESIS (from TI-2, TI-3)
      'adi-. IS-10 is revised to IS-10′: INDF = [ERG, human, −def, no π];
      IMP = [ABS, −def, no π], with the sort unvalued.
TI-10 IS-9 survives in the entries found: every 3i 'a- subject is on a verb with no    HYPOTHESIS (TI-3, TI-4; bí'nílcheeh
      direct object (at most a postpositional one: biih, shił, bí-). The one possible    is T80d)
      exception, bí'nílcheeh (3i subject of bíníshchééh 'compel him'), is not glossed.
TI-11 T77b: in (94) tsin baah ha'as'na', IMP is the Pos. V 3i subject, the            HYPOTHESIS (from TI-3, TI-6, TI-7,
      (human) climber. Willie's English passive 'The tree was climbed' is a             DEF on intransitive-pp)
      translation of an impersonal, like Y&M's 'work is being done' (TI-3). The verb
      has no direct object to promote (TI-6). b- follows because an indefinite
      subject is not a ranked slot (DEF). This replaces IS-12's "cannot decide".
TI-12 T33e: haas'na' is not a transitive stem. Y&M list it as an intransitive         SOURCED (TI-6) / HYPOTHESIS (that
      with a postpositional object; Willie's "corresponding transitive verb" (p. 99)    Willie means "with a P object")
      means a verb with an object of P.
TI-13 Contradicting data for PX-11: níłtsą́ 'adeezhjool 'a rainstorm has started       SOURCED (form, Y&M p. 16 def 15) /
      moving' has a bare noun beside a 3i subject. The vehicle case goes the other      OPEN (whether níłtsą́ is linked,
      way: an overt chidí or łį́į́' takes the ordinary 3rd person (Chidí shił yílwod),    adjoined or something else)
      and the 3i form has no noun (Shił 'ílwod).
TI-14 Focus 2: with 'adijooł the verb picks the subject's person form (3i 'a- vs      SOURCED (Y&M's statement) /
      ordinary Ø) by the KIND of subject (clouds, smoke, rain ...), not by its role,    HYPOTHESIS (the reading)
      which is the same mover in both. That is item-sensitive selection.

4. The data

All Navajo forms are copied from the Y&M interface. Y&M's translation is given as they print it. The Leipzig line segments only as far as Y&M do (their breakdown of the headword). Material they do not segment is joined with dots. 3i = Y&M's indefinite person. The subordinator =go is glossed SUB; Y&M translate it 'when'.

p. 303 def 3daʼdildééh
Y&M: da- Pos. III plural marker + '(a)- Pos. V 3i subj. pron., someone + di- Pos. VIa inceptive + *ldééh 'to move, to go — 3 or more actors' · Leipzig: PL-3i.SBJ-INCEP-go.PL
'to start (to go), to go (plural indefinite subjects)'. Example (perfective daʼdeesdeeʼ): 'everyone from home is going to Gallup but I'm staying behind.'
p. 303 def 21Bus biih daʼiisdeeʼgo chidí bikeeʼ łaʼ niiłtsǫǫz
Y&M (headword daʼiildééh): da- PL + ʼa- Pos. V 3i subj. pron., someone + yi- = hi- seriative + *ldééh · Leipzig: bus 3.into PL.3i.SBJ.SER.go.PL.PFV=SUB car 3.tire one ...
'when the passengers got into the bus one of the tires went flat.' (Also Taah daʼiisdeeʼgo 'when the people got into the water'.)
p. 549 def 24naʼanish
Y&M: "3i subject ʼa-, someone. See naashnish, to work, but note deletion of l-classifier"; prefix naʼa-, stem *Ønish · Leipzig: 3i.SBJ.work (naʼa- unsegmented)
'to be going on (work)'. Tʼáá ʼákwíí jį́ naʼanish ... 'people are always busy (Lit. everyday work is being done).'
p. 555 def 17naʼiidzeeł
Y&M: "(neishheeł, to dream, naʼiidzeeł, dreaming takes place, an indefinite subject dreams)" · Leipzig: 3i.SBJ.dream (unsegmented)
'dream'; 'dreaming takes place'.
p. 465 def 21Tʼah chʼikę́ę́h ʼídlínę́ędą́ą́ʼ ʼashiiké bitahgóó dzidéezʼį́į́ʼ łeh niʼ
Y&M (headword ʼídlį́): "3i subject, someone"; *d-lį́ 'to be' · Leipzig: still girl 3i.SBJ.be.PST ... (rest unsegmented)
'when one is a young girl one always looks at the boys.'
p. 16 def 15Níléí dziłtsį́į́dę́ę́ʼ nihichʼįʼgo níłtsą́ ʼadeezhjool
Y&M (headword ʼadijooł): ʼa- 3i subject "something indefinite" + di- (prefix 'adi-((1))) + *Øjooł 'to move independently — non-compact matter' · Leipzig: yonder mountain.base.from 1PL.toward=SUB rain 3i.SBJ.INCEP.move.PFV
'a rainstorm has started moving toward us from the base of the mountain.'
p. 223 def 1Chidí shił yílwod · Shił ʼílwod
Y&M: ʼílwod = "'a + (y)ílwod" · Leipzig: car 1SG.with arrive.running · 1SG.with 3i.SBJ-arrive.running
'I went (by car).' · 'I went (by a vehicle that ran with me).' Also Naʼnízhoozhígi kǫʼ naʼałbąąsii shił niiltła 'I stopped in Gallup (Lit. the train stopped with me in Gallup)' vs Naʼnízhoozhígi shił niʼíltła '(Lit. something indefinite — the vehicle — stopped with me in Gallup)'.
p. 434 def 8Tsin bąąh hasisʼnaʼ
Y&M (headword haashʼnééh, d-class, prefix ha-((2)), *d-nééh = ʼnééh 'to crawl') · Leipzig: tree 3-against up.1SG.crawl.PFV (unsegmented)
'I climbed the tree.'
p. 337; p. 788 def 4tsé yąąh haasʼnaʼ
Leipzig: rock 3-against up.3.crawl.PFV (unsegmented)
'he climbed the cliff.'

Willie's (94) tsin baah haʼasʼnaʼ and (96) tsin yaah haasʼnaʼ are printed without nasal hooks on absolutive-indef and intransitive-pp. Y&M write bąąh, yąąh. Willie's spelling is kept when her examples are quoted. "3-against" follows Willie's "3-on" for the same postposition. Y&M gloss bąąh in the entry only through the translation 'climb it'.

5. Answers to T77a and T77c

SOURCED T77a. Y&M's 3i subject ʼa- appears on 'go' (daʼdildééh, daʼiildééh), 'work' (naʼanish), 'dream' (naʼiidzeeł) and 'be' (ʼídlį́), glossed "someone", "people", "everyone", "passengers", "one". These are verbs whose single argument does something (Dowty's volition and movement). The referents are human. So Willie's S2 is right, and the absolutive-indef conclusion that it had nothing behind it was a result of her examples, not of the language (TI-7).

HYPOTHESIS That Willie's IMP and Y&M's 3i ʼa- are the same item rests on three things: the same shape, the same slot (Willie's ABS intransitive subject = Y&M's Pos. V subject of an intransitive), and Willie's own reliance on Y&M (p. 57). Y&M also give the sort as open: "someone, something" (TI-1). So IS-8 does not hold as a rule about the prefix. Willie's 'ruin' forms ((24), (26)) are still inanimate, and IS-7 still describes them (TI-8).

HYPOTHESIS T77c. IS-10 split the two indefinites by case and by sort. The question was whether a human indefinite intransitive subject would need INDF in the ABS slot. Y&M answer it the other way: the human indefinite intransitive subject is ʼa- (TI-3), and ʼadi- is "some person" as "the subject (agent) in the Agentive Passive" (TI-2). Hale 2001 says the same from the other side: 3i "functions autonomously as a subject but also in the so-called agentive passive" SOURCED. So the case split stands and the sort split goes (TI-9):

Case/slotSortPersonStatus
INDF ʼadi-ERG: agent of Passive Bhuman ("some person")noneIS-10 kept
IMP / 3i ʼa-ABS: intransitive subject (TI-3, TI-4); transitive object (Pos. IV, TI-1)unvalued: someone or somethingnoneIS-10′ (TI-9)

Y&M also analyse ʼadi- as ʼa- + di- in some entries (p. 21 def 6, "something"). Whether the INDF agent is historically or synchronically 3i + di- is not decided here. T77d (is IMP's [−animate] read by the ranking?) loses its premise. It is restated as T80c.

6. (94) again, and the climbing verb (T77b, T33e)

SOURCED Y&M list the climbing verb as haashʼnééh 'to crawl up out, to creep up, to climb up on hands and knees' (d-classifier). It takes its tree as a postpositional object: "Bąąh haashʼnééh, to climb it (a tree). Tsin bąąh hasisʼnaʼ, I climbed the tree." In four other entries the climber is 3rd person and the postposition is y-: tsé yąąh haasʼnaʼ 'he climbed the cliff', ńdíshchííʼ yąąh haasʼnaʼ (a little brother climbs a pine), kin yighą́ą́ʼ haasʼnaʼ 'the boy climbed up onto the roof'. This is Willie's pattern for (96) exactly. It also settles T33e: the verb is not a transitive stem (TI-12).

HYPOTHESIS T77b. Since the verb has no direct object, Willie's "simple passive" in (94) cannot have promoted the tree, and the tree is in the PP anyway. With TI-3 and TI-7, the natural reading is that IMP in haʼasʼnaʼ is the Pos. V 3i subject, the indefinite climber: 'someone climbed up on the tree'. 'The tree was climbed' is then an English rendering of an impersonal, the same move Y&M make with 'work is being done' for naʼanish. The b- on the postposition follows from DEF (intransitive-pp E4): an indefinite subject is not a ranked slot, so the default b- appears, as it does with the 1st-person climber in Tsin bąąh hasisʼnaʼ. This replaces IS-12's "cannot decide" (TI-11). Two limits. Y&M give no haʼasʼnaʼ, so the segmentation ha-ʼa-... is still read off the spelling against haasʼnaʼ. And the verb is d-class in the lexicon, so the classifier cannot show whether (94) is a passive in Y&M's sense.

7. A noun beside a 3i subject (against PX-11)

SOURCED PX-11 (indef-person) says a bare noun can be linked only to a slot with a person value, so the indefinites never take a noun. Willie's own words are "the indefinite pronouns exclude adjoined nominals" (PX-7). Y&M's vehicle data fit this. With an overt vehicle noun the verb has the ordinary 3rd person (Chidí shił yílwod; kǫʼ naʼałbąąsii shił niiltła). Without one, the vehicle is 3i ʼa- (Shił ʼílwod; shił niʼíltła, "something indefinite — the vehicle").

OPEN But níłtsą́ ʼadeezhjool 'a rainstorm has started moving' has the noun níłtsą́ 'rain' and a 3i subject in one clause. Y&M say this is the rule for these subjects: one "would expect" Ø, but "when the subject is one of those entities listed above", it is 3i ʼa- (TI-4). Taken at face value this is contradicting data for PX-11 and for Willie's generalization (TI-13). Three readings stay open. (i) níłtsą́ is linked to the 3i slot, and PX-11 is wrong. (ii) It is a different kind of attachment (an appositive or topic, like the PP in PX-12). (iii) In this verb ʼa- is thematic, as Y&M say it is in some bases (TI-1), and not a subject pronoun at all. Y&M's wording ("represented in the verb by a 3i subject pronoun") favours (i) or (ii). This is T80a.

8. Rulesets A and B

Step3i human subject (daʼiisdeeʼgo)3i with a noun (níłtsą́ ʼadeezhjool)Ordinary 3rd (Chidí shił yílwod)
Where the subject's form is decidedin the word: Pos. V ʼa- Ain the word, but chosen by the noun's kind (TI-4, TI-14) Ain the word: Ø 3rd A
Sorthuman, from the verb ('go', 3+ actors) and contextnon-compact matter (Y&M's list)from the noun
What clause syntax receivesa subject slot with no π; no noun; the PP biih Ba slot with no π and a noun B (PX-11 problem)a 3rd-person slot; noun linked B
Spell-outprefix array, then =go Aprefix array Aprefix array A

Focus 1. The indefinite subject is a word-level item: Y&M place it in the prefix template (Pos. V), and Hale treats position V as subject material inside the word. What clause syntax sees is only the result, a subject with no person value. The vehicle pair shows the changeover clearly: one meaning ('I went by vehicle') is built either with a clause-level noun (B) plus an ordinary 3rd-person prefix, or with no noun and the word-level indefinite (A). The rain sentence is the case where both appear at once, which is why it matters HYPOTHESIS.

9. Derivations

D1  Bus biih da'iisdee'go 'when the passengers got into the bus' (Y&M p. 303 def 21)
1  Merge  [PP bus [bi- -ih]] 'into the bus'                   B (head-final PP)
2  Merge  [VP PP go.PL]                                       B (head-right VP)
3  Merge  v/Voice; subject = 3i 'a- in Pos. V                 A decision (word level); VPISH: the
                                                              subject starts inside the verb phrase
4  Agree  subject agreement: 3i has no π → no 1/2/3/4 value   A; plural da- (Pos. III) from 3+ actors
5  Spell  da- + 'a- + yi- (seriative) + stem                  A, Y&M's breakdown; surface da'iisdee'
6  Merge  =go (subordinator)                                  B (right edge, head-final)
7  Output da'iisdee'go; no subject noun anywhere              matches Y&M; fits PX-11

D2  (94) tsin baah ha'as'na' under TI-11 (Willie p. 99; reading is HYPOTHESIS)
1  Merge  [PP tsin [b- -aah]]                                 B
2  Merge  [VP PP crawl.up]  (intransitive + P object, TI-6)    B
3  Merge  subject = 3i 'a- (the climber, sort from context)   A
4  Rank   no definite 3rd subject to outrank the P object     DEF (intransitive-pp E4)
          → P gets the default b-, not y-                     matches (94) and *(95)
5  Spell  ha- + 'a- + ... + 'na' (segmentation from spelling) A; schematic beyond Willie's form
6  Output tsin baah ha'as'na'                                 matches Willie (94)
          Compare (96) tsin yaah haas'na': 3rd subject outranks the tree → y-.

D3  níłtsą́ 'adeezhjool 'a rainstorm has started moving' (Y&M p. 16 def 15)
1  Merge  [VP (rain) move]                                    B
2  Select subject form by the kind of subject: rain ∈ Y&M's list → 3i 'a-, not Ø
                                                              A, lexical (TI-4, TI-14)
3  Agree  3i has no π                                         A
4  Spell  'a- + di- (Pos. VIa) + stem                         A, Y&M's breakdown
5  Link?  níłtsą́ to the 3i slot                               B: blocked by PX-11 → mismatch
          (readings (i)–(iii) in §7)                          OPEN (TI-13)

D4  Shił 'ílwod vs Chidí shił yílwod 'I went (by car)' (Y&M p. 223 def 1)
1  Merge  [PP shi- -ł] 'with me'                              B
2  Merge  [VP PP arrive.running]                              B
3a Merge  subject = 3i 'a- (vehicle, unnamed)                 A → 'a + (y)ílwod = 'ílwod
3b Merge  subject = Ø 3rd; Link chidí to it                   A, then B (PX-11 satisfied)
4  Output Shił 'ílwod / Chidí shił yílwod                     both match Y&M

Merge builds each layer (PP, VP, the subject). Agree is the step that copies a person value onto the verb, and with 3i there is none to copy. The step labels and the vP layering are ours. Y&M give only the prefix breakdowns.

10. The role schema (Focus 2)

HYPOTHESIS The 3i subject is neutral about sort and about role. It can be an agent ('go', 'work'), an experiencer-like dreamer, a mover with no will (rain, a vehicle), and in Willie's forms a patient ('ruining took place', IS-7). So ʼa- is not tied to one proto-role. What limits it is its slot (subject of a verb with no direct object, or Pos. IV object). That is what a role schema predicts if the indefinite simply fills whatever role the verb's single argument has. To that extent the weak thesis (roles are visible in the grammar) is supported again.

HYPOTHESIS The ʼadijooł facts point the other way. The mover has the same role whatever moves, but the verb picks 3i for clouds, smoke, rain and snowstorms and Ø for others (TI-14). That is selection by the kind of item, which a pure role schema does not express. The thematic ʼa- of ʼashzhish 'dance' and ʼashhosh 'sleep' (TI-1) is a further case where an argument marker has been frozen into the lexical item, which is lexicalization at the word level. Neither fact decides the strong thesis, which stays OPEN. Grant's revised view (the claim "does not hold cleanly") stays GRANT'S CONJECTURE.

11. Algonquian X compared

HYPOTHESIS Oxford's Algonquian X is [ϕ, AN], animate only, with no [PERS] (AX-1, impersonal-x). AX-9 compared it with Navajo INDF ʼadi-, which is human, has no person and is an agent. TI-2 keeps that comparison. Navajo's other indefinite, 3i ʼa-, has no Algonquian match in this respect: it has no person, like X, but takes either sort, and it serves as the subject of agentive intransitives (TI-3), which X does as well (Algonquian VAI impersonals), and of inanimate movers (TI-4), which X does not. So Navajo separates indefinites by case (ERG vs ABS) and lets sort float on the ABS one. Algonquian, as far as Oxford goes, fixes X's sort as animate. This is a comparison of two descriptions, not a sourced typological claim. The claim about Algonquian VAI impersonals is not checked here (T80e).

12. Fidelity notes

All Navajo forms on this page were copied from the HTML of the Colorado Y&M interface (word + definition search for "3i subj", "indefinite subject", "climb" and the headwords). That edition is a digitized text of the 1987 dictionary, and its OCR shows in places ("sone verb bases" for "some", "peopłe", "Pos_ VIa"). These were corrected only where obvious, and no Navajo form was corrected. The page numbers are the interface's "pg" field. The ʼa- entry is listed as pg 116, which is out of alphabetical order with ʼadijooł at p. 16 and could be an interface error, so it should be checked against the print. No Navajo form, segmentation or judgment on this page is ours. Leipzig lines segment only where Y&M give a breakdown. Hale's sentence is copied from the LI text file. The search found 35 entries for "3i subj"; all 35 were read (two result pages). None of them contradicts TI-1. One adds to it: Y&M also call ho- a "Pos. V, 3i subj. pron., space, area" (kíhonii-, p. 493 def 33), which fits AR on indef-person (an indefinite with a fixed sort, beside ʼa- with an open one).

13. Open questions

IDQuestionWhat would settle it
T80aníłtsą́ ʼadeezhjool: is the noun linked to the 3i slot (PX-11 wrong), attached some other way, or is ʼa- thematic here (TI-13)? Do other ʼadijooł sentences have nouns?Y&M grammar on 3i subjects (around p. 67); Willie ch. 9; more ʼadijooł examples. OPEN
T80bDoes a 3i subject change the classifier (naʼanish: "deletion of l-classifier"; naʼiidzeeł beside neishheeł)? If so, are Y&M's and Willie's impersonals a derived voice ("simple passive") and not just a subject pronoun?Y&M 1987 grammar p. 120 (simple passive) and the 3i section; Neudorf 2000 (cited by Hale). OPEN
T80c(replaces T77d) Is there a 3i ʼa- object (Pos. IV) with a human referent, as TI-1 allows ("something or someone")? Does the ranking ever read 3i's sort?Y&M entries with "3i obj."; Willie ch. 4 on images. OPEN
T80dbíʼnílcheeh (3i subject of bíníshchééh 'compel him'): is bí- there a postposition (Y&M Pos. O-Ib for bíʼni-) or a direct object? If a direct object, IS-9/TI-10 fails.Y&M paradigm page for bíníshchééh; a sentence with bíʼnílcheeh. OPEN
T80eGeneric 'one': Y&M use 3i for 'when one is a young girl' (ʼídlį́), Willie uses the fourth person (IMP-4). How do 3i and fourth person divide the generic reading? And do Algonquian VAI impersonals match 3i on agentive intransitives (§11)?Willie pp. 118–126 on images; Y&M ji- entries; Wolvengrey 2011 on VAI impersonals. OPEN

T77a answered (TI-7: forms SOURCED, the identification HYPOTHESIS). T77c answered (TI-9, HYPOTHESIS; IS-10′). T77b answered as HYPOTHESIS (TI-11). T33e answered (TI-12). T77d restated as T80c. The rules IS-8, IS-10 and IS-12 on absolutive-indef are left as they were published; TI-8, TI-9 and TI-11 supersede them. Still open: T79a–d, T78b, T76b, T76c, T75b, T75c. Sources: Young & Morgan 1987 (dictionary entries as listed in §2); Hale 2001. See also absolutive-indef, indef-person, intransitive-pp, ergative-indefinite, impersonal-x, fourth-uses.

When Plains Cree's impersonal actor X acts on a third person, the verb looks direct but cannot take an actor at all, so X>3 is not a transitive direct (T78a, T70e)

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

Restated on the unified structure (2026-11-15). The indefinites (ʼa-, ʼadi-, areal ho-) are restated as personless pronouns in ordinary argument positions (H60): ʼadi- as the silent indefinite agent in Spec,vP that raises to Spec,TP by the default (H10b), ʼa- in VP. The 3i data on this page have not yet been individually restated. 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, §4, indefinite and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

T78a asked whether an Algonquian X→3 form ('someone sees him/her') is direct or inverse. The last run could not tell, because Oxford 2022 gives only X→1/2 forms. This run opened Wolvengrey 2011, a Plains Cree grammar-length dissertation the site had not used before, which gives the full X paradigm. The answer has three parts. (1) In the independent order, X→3 has the direct theme -ā and endings "formally identical" to 1→3 and 2→3 direct forms (Table 2.65, (112c)) SOURCED. (2) In the conjunct it has -iht, which is neither the direct nor the inverse theme (Table 2.66) SOURCED. (3) No actor can ever be expressed with an X form, and the same meaning with an overt 'someone' uses the ordinary inverse ((105)–(106), (113)–(114)) SOURCED (Wolvengrey's judgments). So X→3 is direct in form only, and Wolvengrey treats it as syntactically intransitive SOURCED (his analysis). Two things follow. The warning against EI-8 on ergative-indefinite loses its basis (T70e) HYPOTHESIS. And the X>3 rank turns out to be a summary of this same split, not separate evidence (AX-12 dissolved) HYPOTHESIS. The strong roles-not-words thesis stays OPEN.

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 Answer to T78a · 6 Oxford's probe applied to X→3 · 7 What this does to EI-8 (T70e) · 8 No doubling, and a control test (T78c, T78d) · 9 Rulesets A and B · 10 Derivations · 11 The role schema (Focus 2) · 12 Fidelity notes · 13 Open questions

Notation is the same as on impersonal-x. No Syntax · Semantics · Interface conventions file is in the repo or the context folder. A = the word level (prefixal, left-headed in Navajo). B = the clause level (right-headed). Σ = the role schema, with Dowty's proto-roles as the baseline. INDF = Navajo ʼadi-. X = the Algonquian impersonal or "unspecified" actor. Wolvengrey's glosses are kept as he writes them (for example "X-3s" = X acts on 3s, "1/2" = local-person ending, IPV = preverb, CNJ = conjunct). A Leipzig line is added under each; it only changes his labels and keeps his segmentation. Plains Cree forms are in plain italics, not tagged as Navajo. Merge builds structure (puts two pieces together). Agree copies feature values such as person and number from a goal onto a probe. Oxford's Agree splits a probe's conditions into interaction (which goals it may copy from) and satisfaction (which goal stops the search).

1. Question

Oxford's Routledge chapter gives the independent theme-sign hierarchy as 1/2>X>3>3’>0 (PDF p. 17), which predicts that X→3 is direct. Oxford 2022, though, gives X fewer features than an animate third person ([ϕ, AN], no [PERS]; AX-1), and in the conjunct an X agent makes a verb go inverse more readily than an animate one (AX-4). impersonal-x §7 left this as AX-12 and T78a: what does an actual X→3 form look like? The answer also decides T70e (ergative-indefinite), which warned that a direct X>3 would count, by analogy, against EI-8 (Willie's inverse reading of biʼdistał).

2. Sources

SOURCED Wolvengrey, Arok Elessar 2011. Semantic and Pragmatic Functions in Plains Cree Syntax. PhD dissertation, University of Amsterdam. Utrecht: LOT (dissertation series 268). Free PDF from LOT (lotpublications.nl, Documents/268_fulltext.pdf). The PDF is born-digital (made from Word), and printed page = PDF page − 22. Read this run from the text layer, and checked against page images for the forms used here: pp. 158–159 ((105), (106)), pp. 163–164 ((111), Tables 2.65 and 2.66), pp. 165–167 ((112)–(115)), pp. 185–190 (§3.2, (18), (21), (23)–(29)), pp. 191–192 (Figures 3.1, 3.2).

SOURCED Oxford, Will 2022. lingbuzz 006486 (as on impersonal-x). Re-read this run for the mechanism only: pp. 9–10 (VoiceERG), p. 15 ((20) and "featureless Voice is realized as the elsewhere form -iko"), pp. 24–26 ((24), (25), the definition of the inverse pattern on p. 25, C3, C5).

Already on the site and not re-opened: Oxford's Routledge chapter PDF pp. 17–18 (ergative-indefinite, impersonal-x); Willie 1991 p. 152 (1f) biʼdistał=yę́ę and p. 37 (21) niʼdilzhééh (from page images on earlier pages).

Cited by Wolvengrey but not opened: Dahlstrom 1991, Jolley 1982, Déchaine & Reinholtz 1998, Ellis 1970, Hockett 1996. Still not opened: Goddard 1979, Nichols 1980, Valentine 2001.

3. Rules first

XT-1  Plains Cree independent X→3 = Ø (no person prefix) + stem + -ā + -w / -wak /    SOURCED (Wolvengrey Table 2.65 p. 164,
      -ýiwa. These suffixes "are formally identical to those which occur in the VTA       (112c) p. 165)
      Direct Mixed Set with first or second person actors and third person patients."
XT-2  Independent and conjunct X→SAP = stem + -ikawi + local ending (with the person     SOURCED (Table 2.65–2.66 p. 164)
      prefix in the independent). -ikawi is Oxford's INV.IMP.
XT-3  Conjunct X→3 = ē- + stem + -iht (3s), -ihcik (3p), -im-iht (3’). This is neither   SOURCED (Table 2.66 p. 164)
      the direct theme -ā nor the inverse -iko. Wolvengrey: the 3rd-person conjunct
      forms are "truly exceptional" (p. 167).
XT-4  No X form allows an overt actor: (*awiyak) in (106), (*awiya) in (114). With an    SOURCED (pp. 158–159, 165–166;
      overt 'someone' the same meaning uses the ordinary inverse -ikw: 3s→1s (105),        Wolvengrey's judgments)
      3’→3s (113), where awiya 'someone' is itself glossed obviative (PR.3’).
XT-5  The hierarchy 2>1>X>3>3’>0 is "suggested" to analysts by exactly the split in   SOURCED (pp. 163, 185–186;
      XT-1 vs XT-2 (local forms ≈ inverse, third-person forms ≈ direct).                  after Jolley 1982, Déchaine &
                                                                                          Reinholtz 1998)
XT-6  With kakwē- 'try', the one who tries may be the patient ((23)–(25)) or the         SOURCED (pp. 188–190)
      unspecified actor ((26)–(28)), by context; one form, (29), has both readings.
      "The actor is unspecified, but not pragmatically absent."
XT-7  Wolvengrey's analysis: the X construction does not remove the semantic argument     SOURCED as his analysis
      but makes it "obligatorily unspecified and unindexed"; X forms are                  (pp. 158, 165–167, 190)
      syntactically intransitive, with no grammaticalized subject or object.
XT-8  Answer to T78a: X→3 is DIRECT IN FORM in the independent (XT-1), NEITHER in the    SOURCED (forms and judgments) /
      conjunct (XT-3), and not a transitive direct in either, because no actor can be     the summary is ours
      expressed (XT-4).
XT-9  Oxford's C5 Infl [int: PERS, sat: PART,PL], run on a TRANSITIVE X→3 structure      HYPOTHESIS (our application;
      (X in Spec,VoiceP), copies only the patient's features via Voice, so Infl and        Oxford's paper covers 3→SAP only)
      Voice match, Voice is impoverished, and the elsewhere -iko is predicted. The
      attested forms have -ā or -iht. So either the model does not carry over to X→3,
      or X→3 has no X argument in Spec,VoiceP (XT-7).
XT-10 AX-12 dissolved: the X>3 rank restates the XT-1/XT-2 split (XT-5). Under XT-7,  HYPOTHESIS
      an X→3 form contains no second argument to be ranked against the 3, so the rank
      is not evidence about X's features, and it does not conflict with AX-1.
XT-11 T70e: the Algonquian argument against EI-8 by analogy loses its basis. X→3 is not  HYPOTHESIS
      a transitive direct (XT-8). The nearest transitive analogue of Navajo bi'distał,
      an overt indefinite 'someone' acting on a 3rd person, is inverse in Plains Cree (113).
      This is consistent with EI-8′ but does not decide it.
XT-12 T78c, Algonquian half: X cannot be doubled by a nominal (XT-4). Parallel with      SOURCED (XT-4) / HYPOTHESIS (the parallel)
      PX-11 (a Navajo noun links only to a slot with a person value). The overt
      'someone' is a person-bearing obviative 3rd and enters the ordinary
      direct/inverse system. X does not (cf. AX-7).
XT-13 T78d, Algonquian half: XT-6 is a test that tells a present but unindexed agent     SOURCED (the test) / OPEN (Navajo)
      from no agent (the actor can be the one who tries). A Navajo version needs an
      INDF form under a 'try'-type predicate; none has been opened.

4. The data

Plains Cree forms are copied from Wolvengrey's examples and tables, with his segmentation and glosses, and checked against the page images. The Leipzig line is ours. It changes only his labels.

(105), p. 159awiyak nikī-wāpamik.
awiyak ni- kī- wāpam -ikw -Ø · W: PR.3s 1 IPV VTA INV 3s / someone PST see 3s-1s
Leipzig: someone 1-PST-see-INV-3SG
'Someone saw me.' (overt 'someone', ordinary inverse)
(106), p. 159(*awiyak) nikī-wāpamikawin.
(awiyak) ni- kī- wāpam -ikawi -n · W: PR.3s 1 IPV VTA X 1/2 / someone PST see
Leipzig: (someone) 1-PST-see-X-LOC
'I was seen.' *'Someone saw me.' (X form; an overt actor is ungrammatical)
(112a), p. 165niwāpamāw.
ni- wāpam -ā -w · W: 1 VTA DIR 3s / see 1s-3s
Leipzig: 1-see-DIR-3SG
'I see him/her.'
(112c), p. 165wāpamāw.
Ø- wāpam -ā -w · W: X VTA DIR 3s / see X-3s
Leipzig: Ø-see-DIR-3SG
'(Someone) sees him/her.' / 'S/he is seen (by someone).'
(113), p. 166awiya kī-wāpamik cān.
awiya kī- wāpam -ikw -Ø cān · W: PR.3’ IPV VTA INV 3s NA.3s / someone PST see 3’-3s John
Leipzig: someone.OBV PST-see-INV-3SG John
'Someone saw John.' / 'John was seen by someone.'
(114), p. 166(*awiya) kī-wāpamāw cān.
(awiya) Ø- kī- wāpam -ā -w cān · W: (PR.3’) X IPV VTA DIR 3s NA.3s / (someone) PST see X-3s John
Leipzig: (someone.OBV) Ø-PST-see-DIR-3SG John
'John was seen.' *'Someone saw John.' *'John was seen by someone.'
T2.66, p. 164ē-wāpamiht · ē-wāpāmihcik · ē-wāpamimiht
W: ē- + stem + -iht (3s) / -ihcik (3p) / -im-iht (3’)
Leipzig: CNJ-see-X.3SG · CNJ-see-X.3PL · CNJ-see-OBV-X.3SG
'(as) s/he is seen' · '(as) they are seen' · '(as) (the other(s)) is/are seen'
(29), p. 190nikī-kakwē-wāpamikawin.
ni- kī- kakwē- wāpam -ikawi -n · W: 1 IPV IPV VTA XAct 1/2 / PST try see (X-)1s
Leipzig: 1-PST-try-see-X-LOC
'Someone tried to see me.' / 'I tried to be seen.'

Table 2.66 prints the 3p form as ē-wāpāmihcik, with a long ā in the stem, unlike every other cell. It is copied as printed. Glossing -iht as "X.3SG" is our label. Wolvengrey's table leaves the X column empty for these rows and puts -iht under "endings" spanning both columns. Navajo comparison forms, not re-opened: biʼdistał=yę́ę (Willie p. 152 (1f), source gloss "PASS 3.3.kicked", 'the one who was kicked') and niʼdilzhééh (p. 37 (21), 'You are being shaved'), both as transcribed from page images on ergative-indefinite and passive-rank.

5. Answer to T78a

In the independent, (112c) wāpamāw is built exactly like (112a) niwāpamāw 'I see him/her' minus the person prefix: same -ā, same -w SOURCED. So the 1/2>X>3 prediction ("X→3 is direct") is met in form. In the conjunct it is not met: ē-wāpamiht has neither -ā nor -iko SOURCED. And in neither order does the form act like a transitive. An overt actor is starred in (114), while the inverse (113) with an overt 'someone' is fine SOURCED. Wolvengrey's conclusion is that the X→3 form is "formally identical" to a direct (p. 165) but is intransitive: his analysis is "a bit of a cross-breed" between the traditional true-actor analysis and a syntax in which "the third person is the sole participant allowed" (p. 166), and the form is "equivalent to a VAI form" (p. 167) SOURCED (his analysis). He adds that the independent form is homophonous with VAI forms, and that this homophony gave rise to a new VAI stem, ayamihā- 'pray' ((115), p. 167). So the answer is: direct in form only (XT-8).

6. Oxford's probe applied to X→3

Oxford's paper derives only 3→SAP forms. The extension below is ours HYPOTHESIS. In his model the inverse pattern is the case where "Infl agrees only with the patient's features on Voice" (p. 25). When that happens, Infl and Voice carry identical features, Voice's copy is deleted, and "featureless Voice is realized as the elsewhere form -iko" (p. 15) SOURCED. Plains Cree is C5: Infl [int: PERS, sat: PART,PL] (p. 26) SOURCED. Suppose X→3 were transitive, with X in Spec,VoiceP. Infl could not interact with X, which lacks [PERS] (AX-1), so it would copy only the 3AN patient's features from Voice. The identity condition would then hold, and the model predicts -iko. The language has -ā (independent) and -iht (conjunct) instead (XT-9). There are two ways out, and the sources opened do not choose between them OPEN:

  • (i) No X argument. Wolvengrey's analysis: Voice has no external argument. It still agrees with the patient (-ā = 3AN object, which is Rhodes's reading of -aa cited in the Routledge chapter fn. 4). Infl then has a single goal, and Oxford's inverse is defined only for a transitive Voice. Nothing is mispredicted.
  • (ii) A different probe in 3-patient forms. Oxford derives the third-person inverse with a distinct head, VoiceERG (ergative alignment, an [EPP] that attracts the patient; pp. 9–10). If X gets no such case, X→3 might never enter that system at all. This would need Oxford's own treatment of X→3, which has not been found.

Either way, the X>3 rank does no work. Wolvengrey reports that the extended hierarchy (111)/(18) was "suggested" to analysts by the very split under discussion, local -ikawi ≈ inverse versus third-person -ā ≈ direct (p. 163) SOURCED. So the rank restates the morphology and is not separate evidence that X has more features than a 3. That removes the tension in AX-12 (XT-10) HYPOTHESIS.

7. What this does to EI-8 (T70e)

ergative-indefinite §7 warned: if Algonquian X>3 is direct, then by analogy Navajo INDF>3 (biʼdistał) should not be an inverse, against EI-8. The warning assumed a transitive direct. Plains Cree X→3 is not one (XT-8), so the analogy no longer goes through HYPOTHESIS. The closer transitive analogue is (113). There an overt indefinite 'someone', inflected obviative, acts on a proximate 'John', and the verb is inverse SOURCED. In Willie's account biʼdistał is an inverse-type form with an indefinite agent and a 3rd-person patient that becomes the "subject of agentive passive" (p. 152). The two look alike: an indefinite agent ranked below a 3rd-person patient, with inverse morphology HYPOTHESIS. They also differ. Plains Cree (113) is still transitive (it allows the overt agent), while third-voice VC-14 made the Navajo agentive passive intransitive (EI-8′). The Plains Cree X form (114) is the intransitive one, and it is direct-looking. So Algonquian splits the two properties that EI-8′ puts together, inverse marking and intransitivity, across two constructions. The analogy is consistent with EI-8′ but cannot confirm it OPEN. T70e is answered in the sense that the Algonquian fact is now known (XT-8). What it means for Navajo stays a HYPOTHESIS (XT-11).

8. No doubling, and a control test (T78c, T78d)

Doubling. X can never be doubled by a nominal, not even by the indefinite pronoun awiyak/awiya 'someone' ((106), (114)) SOURCED. The pronoun is a full 3rd person: it is glossed PR.3s or PR.3’ (obviative), and it takes the ordinary inverse ((105), (113)). So Plains Cree has two 'someones'. One is lexical, carries person and obviation, and lives in clause syntax. The other is X: personless, unindexed, and never a noun. That fits AX-7 (X "do[es] not participate in or trigger obviation", Oxford 2018 fn. 13) and PX-11 for Navajo (a noun links only to a slot with a person value) HYPOTHESIS (the parallel). The Navajo half of T78c, whether INDF also stays out of the 3rd/4th-person system, needs Willie ch. 9 and stays OPEN (T79b).

Control. Wolvengrey's kakwē- 'try' data give a test that tells a present but unindexed agent apart from no agent at all. In (26)–(28) the unspecified actor is the one who tries ('someone's always trying to see me'). In (23)–(25) the patient is ('try to be seen!'). (29) is ambiguous SOURCED. If X were simply absent, the (26)–(28) readings should be impossible. Wolvengrey takes the choice to be pragmatic, not grammatical (p. 190) SOURCED (his analysis). For Navajo, T78d can now be stated as a schema with no Navajo forms: [V TRY [V INDF-V]]. Can the unexpressed INDF agent be the one who tries? No sourced Navajo INDF form under a 'try'-type predicate has been opened. The Navajo half of T78d is OPEN (T79c).

9. Rulesets A and B

StepPlains Cree X→3 (Wolvengrey)Plains Cree overt 'someone' (113)Navajo INDF (Willie, via earlier pages)
Where the agent is decidedin the word: the X form makes the actor unspecified and unindexed (Wolvengrey says -ikawi "essentially mirrors" -ikātē, which derives stems, and weighs treating it "as stem derivation", p. 165) Ain the clause: awiya Merges as an argument noun Bin the word: ʼadi- in the prefix array A
What clause syntax receivesa one-participant verb; no actor noun can Merge B (XT-4)a two-participant verb; obviation and topicality pick the inverse Ba verb whose agent slot has no π; no noun links to it (PX-11) B
AgreeVoice ↔ 3AN patient (-ā); Infl has only the patient as a goalInfl ↔ patient only → inverse -ikw (Oxford's pattern)object slot ← patient's π; agent slot gets no π A
Spell-outwāpam-ā-w, ē-wāpam-iht (suffixes)kī-wāpam-ikw-Øprefix array, e.g. ni-ʼadi-l-zheeh A

This is the site's Focus 1 point in a second language: the word is where "an agent without a noun" is decided, and the clause only sees the result. A clause-level indefinite (the noun awiya) behaves like any 3rd person. A word-level one (X, INDF) is invisible to clause-level linking HYPOTHESIS. Algonquian exponents are suffixes, so A = "prefixal, left-headed" is a Navajo label. In the Cree columns A only marks "inside the word".

10. Derivations

D1  (112c) wāpamāw '(someone) sees him/her', reading (i): no X argument (Wolvengrey)
1  Merge  [VP see 3AN]                                       B (clause syntax)
2  Merge  Voice, no specifier (X form: actor unspecified)     A decision (word level), XT-7
3  Agree  Voice ↔ 3AN patient → -ā                            Rhodes's 3AN-object reading of -ā
4  Merge  Infl; its only goal is the patient                  B
          (no second goal, so no "agrees only with the patient"
          contrast and no impoverishment)
5  Spell  Ø- + wāpam + -ā + -w                                word: suffixes
6  Output wāpamāw                                             matches (112c)
          An overt actor at step 2 would need a specifier the
          X form does not have → *(114) with awiya            matches XT-4

D2  The same form, reading (ii'): X in Spec,VoiceP with Oxford's C5 Infl (our test)
1  Merge  [VP see 3AN]                                       B
2  Merge  Voice; Agree Voice ↔ 3AN                            B
3  Merge  X [ϕ, AN] in Spec,VoiceP                            B
4  Agree  Infl [int: PERS]: X lacks PERS → no interaction     AX-1, AX-5
          Infl copies the patient's [ϕ, AN, PERS] from Voice
5  Rule   Infl = Voice → delete Voice's copy                  Oxford p. 15
6  Spell  featureless Voice → elsewhere -iko                  predicted: an -iko form (not attested)
7  Output does NOT match (112c) -ā or Table 2.66 -iht         XT-9: the model as extended mispredicts.
          (The shape of the predicted form is schematic; no such form is in the sources.)

D3  (113) awiya kī-wāpamik cān 'Someone saw John.'
1  Merge  [VP see cān(3, PROX)]                              B
2  Merge  Voice; Merge awiya(3’, OBV) in Spec,VoiceP          B; a full 3rd person noun
3  Rank   PROX patient outranks OBV agent → inverse          B (obviation; Oxford's VoiceERG, pp. 9–10)
4  Agree  Infl ↔ patient only                                B
5  Spell  kī- + wāpam + -ikw + -Ø (3s)                        word
6  Output kī-wāpamik cān                                      matches (113)

D4  Navajo (6) bi'distał=yę́ę 'the one who was kicked' (Willie p. 152 (1f)), set beside D1 and D3
1  Merge  [VP 3rd patient kick]                               B; head-right VP
2  Merge  [vP INDF[hum, −def, no π] [VP …] v]                 A decision (prefix 'adi-), VPISH
3  Agree  object slot ← 3rd patient (bi-); INDF gets no π     A
4  Rank   EI-8′ (Willie): inverse-type bi-, patient = "subject of agentive passive"
                                                              HYPOTHESIS (as on ergative-indefinite)
5  Spell  bi- + 'a(di)- + d- + stem, =yę́ę                     A, then enclitic
          Like D1 in step 2 (an agent decided in the word, with no noun).
          Like D3 in step 4 (inverse-type marking with a 3rd-person patient on top).
          Plains Cree keeps these two apart (D1 ≠ D3); Navajo, under EI-8′, has both
          in one form. HYPOTHESIS. No new Navajo data.

D1 and D3 put Wolvengrey's forms into the site's step format. The step labels are ours. D2 applies Oxford's C5 settings to a configuration his paper does not treat. D4 repeats ergative-indefinite D1 in outline. Merge builds each layer. Agree copies person values. The question in every case is which head gets features from which goal.

11. The role schema (Focus 2)

X is the clearest case so far of a theta role that is present without any argument item. Wolvengrey says the construction "does not remove the semantic argument from the understanding of the state of affairs, but it does render it obligatorily unspecified and unindexed on the verb" (p. 158) SOURCED. The control data show that this role can still be active: it can be the one who tries (XT-6) SOURCED. In site terms, Σ contains a proto-agent, while the syntax and agreement contain nothing for it. So the role schema and the argument list come apart HYPOTHESIS. On its face this is the closest thing yet to Grant's original picture, a verb that takes a role rather than a lexical item. But it is a special construction marked in the word (X, like INDF), and Wolvengrey argues the control is pragmatic, not syntactic. So it shows that roles can exist without items. It does not show that ordinary verbs select roles instead of items. The weak thesis gains a second, cleaner case. The strong thesis stays OPEN. Grant's revised view (the claim "does not hold cleanly") stays GRANT'S CONJECTURE.

One more point. The overt 'someone' awiya has the same role as X in (113) vs (114): proto-agent of 'see'. The grammar still treats the two completely differently (inverse and transitive vs direct-looking and intransitive). As on impersonal-x §10, what the agreement reads is the item's features (person, obviation), not the role SOURCED (the contrast) HYPOTHESIS (the reading).

12. Fidelity notes

All Plains Cree forms were copied from the text layer of Wolvengrey's born-digital PDF and checked against page images of PDF pp. 181, 186, 187 and 188 (printed pp. 159, 164, 165 and 166). The macrons (ā, ē, ī) and the ý in -ýiwa are as printed. On p. 190 Wolvengrey writes "(as in (36)-(38))" for the actor-control examples. In context these are (26)–(28), and they are cited that way here. Wolvengrey's own gloss labels (VTA, IPV, PR, NA) are kept in the source lines. No Navajo form, segmentation or judgment on this page is new. The Navajo forms are copied from earlier pages that transcribed them from Willie's page images.

13. Open questions

IDQuestionWhat would settle it
T79aDoes Oxford (or anyone using his probe model) treat X→3? Does reading (i) (no X argument) or (ii) (no VoiceERG for X) rescue the model (XT-9)? Is independent X→3 in Ojibwe (the Routledge chapter's other main template source, Nichols 1980) also built on -aa?Oxford 2017/2019 on impoverishment; Nichols 1980; Valentine 2001. OPEN
T79b(= T78c, Navajo half) Does Navajo INDF ʼadi- stay outside the 3rd/4th-person system, as X stays out of obviation and only the overt 'someone' enters it (XT-12)?Willie ch. 9 on images. OPEN
T79c(= T78d, Navajo half) Is there a sourced Navajo INDF form under a 'try'-type or other control predicate, and can the unexpressed agent be the one who tries (XT-13)?Young & Morgan 1987; Willie pp. 36–38 and ch. 6 on images. OPEN
T79dIs the conjunct X→3 -iht segmentable? Does it contain a cognate of Kickapoo/PA -en IMP (Oxford Tables 3–4), which would make the conjunct X a real morpheme in the word and not just a missing actor?Wolfart 1973; Dahlstrom 1991; Goddard 1979. OPEN

T78a answered: direct in form only (XT-8; forms SOURCED, the summary ours). AX-12 is dissolved as HYPOTHESIS (XT-10). T70e: the Algonquian fact is now known, and the analogy against EI-8 loses its basis (XT-11, HYPOTHESIS). The Algonquian halves of T78c and T78d are answered (XT-12, XT-13). Their Navajo halves continue as T79b and T79c. Still open: T78b, T76b, T76c, T77a–d, T75b, T75c, T73b–e, T74a–c, T74e. Sources: Wolvengrey 2011 pp. 158–167, 185–192; Oxford 2022 pp. 9–10, 15, 24–26. See also impersonal-x, ergative-indefinite, third-voice, passive-rank, The inverse.

Smoke, cloud and rain take the 3i prefix with their nouns: in the *Øjooł family the verb picks ʼa- or Ø by what kind of stuff moves, so PX-11 fails for one verb class (T80a)

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

2026-11-11 audit pass 3. Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference. See the pass 3 report.

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

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

Restated on the unified structure (2026-11-15). The indefinites (ʼa-, ʼadi-, areal ho-) are restated as personless pronouns in ordinary argument positions (H60): ʼadi- as the silent indefinite agent in Spec,vP that raises to Spec,TP by the default (H10b), ʼa- in VP. The 3i data on this page have not yet been individually restated. Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §4, indefinite and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

2026-11-11 audit pass 1. D1 called a subject merged directly inside VP “VPISH”, but the standing implementation first-merges the subject DP in left Spec,vP. That particular D1 is withdrawn as our completed derivation rather than patched by moving an unaccusative/theme arbitrarily. D1–D4 now count as historical hypotheses; noun class agreement, a moving light d, classifier/v identity and primary V→v→T versus PF realization need re-analysis. Dictionary classifications are not thereby invalidated. OPEN: an event LF and a complete subject-base derivation for this family.

T80a asked about one sentence: níłtsą́ ʼadeezhjool 'a rainstorm has started moving' (Young & Morgan 1987, p. 16 def 15). It has a bare noun beside a 3i subject ʼa-. PX-11 (indef-person) says that cannot happen, because a noun links only to a slot with a person value. This run read all 106 Y&M entries whose headword ends in -jooł (the stem for moving or handling non-compact matter), plus ʼoojoł and naanáʼoojoł. The rain sentence is not a stray. In 9 entries for clouds, smoke, fog, rain, snowstorms, dust and a stench moving, all 19 example sentences have an overt noun, and every verb has the 3i ʼa- SOURCED. Twice Y&M add a note: here the 3i "does represent the substance functioning as the subject", and "Note apparent use of 3i ʼa- to represent a definite substance" SOURCED. The same stem with wool, hay, hair or a wig as the mover has the ordinary Ø 3rd person. That gives five minimal pairs (naajooł / naʼajooł, chʼéjooł / chʼéʼéjooł, …) SOURCED. So T80a is answered: the noun is linked to the 3i slot in this verb class, and PX-11 needs a restriction HYPOTHESIS. Plain words: in this family the verb's subject prefix is chosen by what kind of thing moves (gas vs. loose solid), not by the mover's role, which is the same in both. That counts against the strong "roles, not words" thesis. The strong thesis stays OPEN.

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 Minimal pairs · 6 Answer to T80a · 7 Rulesets A and B · 8 Derivations · 9 The role schema (Focus 2) · 10 Fidelity notes · 11 Open questions

Notation is the same as on indef-subject. No Syntax · Semantics · Interface conventions file is in the repo or the context folder. A = the word level (prefixal, left-headed). B = the clause level (right-headed). Σ = the role schema, with Dowty's proto-roles as the baseline. 3i is Young & Morgan's indefinite person ("3i subj. pron."). NC = Y&M's "non-compact matter". Merge builds structure (puts two pieces together into a bigger one). Agree values features: a head with an unvalued feature such as person, number or here noun class copies the value from a matching item. LINK is the site's name for the clause-level (B) association of a noun with a word-internal argument slot (PX-11).

1. Question

indef-subject §7 left three readings of níłtsą́ ʼadeezhjool. (i) The noun níłtsą́ 'rain' is linked to the 3i slot, so PX-11 is wrong. (ii) It is attached some other way (an appositive or topic, like the PP in PX-12). (iii) In this verb ʼa- is a thematic prefix, not a subject pronoun. T80a also asked: "Do other ʼadijooł sentences have nouns?" One sentence cannot settle this, so the test is the whole stem family. If (i) is right, nouns should appear with the 3i regularly, and Y&M should treat the 3i as standing for the noun's referent. If (iii) is right, Y&M should tag the ʼa- as thematic. If the 3i is an ordinary indefinite pronoun, as with the vehicle verbs on indef-subject, nouns should be absent.

2. Sources

SOURCED Young, Robert W. & William Morgan, Sr. 1987. The Navajo Language: A Grammar and Colloquial Dictionary, revised edition. Albuquerque: UNM Press. Read through the free University of Colorado searchable edition (verbs.colorado.edu/~mahu0110/ym/). Searches: (a) headword regex jo[oó]ł$, verbs only: 106 entries, all read (five result pages); (b) word + definition adijooł: 10 entries (adds ʼoojoł p. 681 def 18 and naanáʼoojoł p. 585 def 4); (c) definition regex 3i.{0,40}(represent|definite|specific): 5 entries; (d) headword ^'a-: the five ʼa- prefix entries (p. 116 def 5; p. 119 def 7; p. 681 def 12; p. 1 def 2; p. 3 def 32); (e) headword ^níyol$ (p. 665 def 1; p. 664 def 19). Cited as "Y&M p. N def M" as the interface gives them. They were not checked against the print (§10).

Already on the site and not re-opened: PX-7, PX-11, PX-12 (Willie 1991 p. 51 and (40d, e); indef-person); TI-1–TI-14 and IS-10′ (indef-subject).

3. Rules first

GM-1  Y&M's 'a- prefixes: "Pos. IV, 3i direct obj. pron. ... and Pos. V, 3i subj.      SOURCED (Y&M p. 116 def 5; p. 119
      pron." (p. 116); "Pos. IV-V, 3i obj. subj. pron." (p. 119); "('a- and           def 7; p. 681 def 12; p. 3 def 32)
      Progressive Mode yi- combine to produce 'oo- in the 3rd person.)" (p. 681);
      and a distinct 'a-, "Pos. Ib, away out of sight", which before a second 'a-
      "takes the shape 'e-" (p. 3; its own text calls it 'A-((1))).
GM-2  The stem *Øjooł: "to move independently --- non-compact matter" (wording         SOURCED (stem notes in every entry
      varies: "gaseous or non-compact", "through the air"). Transitive *łjooł "to       of §4)
      handle --- non-compact matter" (wool, loose hay, a wig).
GM-3  GAS SET. Nine intransitive *Øjooł entries for cloud, smoke, fog, rain,           SOURCED (§4, Table 1)
      snowstorm, dust, a stench: 'adijooł, 'adiijooł, 'e'ejooł, ha'ajooł, 'íjooł,
      na'ajooł, naaná'oojoł, 'oojoł, ch'é'éjooł. All have 'a- in the prefix string
      (Y&M's prefix labels 'adi-, 'adin-, 'e'e-, ha'a-, 'a-((2)), na'a-((1)),
      naaná'a-, 'a-((2)), ch'é'e-). Their 19 example sentences ALL have an overt
      noun for the mover (k'os, łid, 'áhí, níłtsą́, chxííl, łeezh, bilid, bitł'id).
GM-4  Y&M say what the 'a- is in the gas set: "One would expect ... Ø. However,      SOURCED (p. 16 def 15; p. 362 def 8;
      when the subject is one of those entities listed above, it is represented in    p. 401 def 9)
      the verb by a 3i subject pronoun 'a-" (p. 16); "'a-, Pos. IV-V is the 3i
      pronoun, and it does not usually represent a specific entity or thing. Here
      it does represent the substance functioning as the subject of the verb"
      (p. 362); "Note apparent use of 3i 'a- to represent a definite substance"
      (p. 401).
GM-5  LOOSE-SOLID SET. The same intransitive stem with wool, hay, hair, a wig,        SOURCED (§4, Table 2)
      tumbleweeds or clothes as mover: dijooł, diijooł (two entries), naajooł,
      ch'éjooł, yijooł, yíjooł, hijooł. Prefix labels di-, dii-, na-((2)), ch'í-,
      Ø, Ø, hi-. No 3i. Overt nouns ('aghaa', tł'oh, tsiigháshchíín, ...) with the
      ordinary Ø 3rd person, as PX-11 expects.
GM-6  Five minimal pairs differ by the 'a- alone, with Y&M cross-referencing        SOURCED (§5)
      them ("Cf. 'adijooł", "Cf. dah 'adiijooł"): dijooł/'adijooł,
      dah diijooł/dah 'adiijooł, naajooł/na'ajooł, ch'éjooł/ch'é'éjooł,
      yíjooł (P nijool)/'íjooł (P 'aníjool).
GM-7  Two smoke verbs without a 3i label: łid binák'e 'ajooł 'get into one's eyes    SOURCED (p. 76 def 1; p. 474 def 17)
      (one puff)' (Y&M: 'a-((1)), Pos. Ib) and łid binák'ee 'iijooł (puff after
      puff; 'ii-((1))). One 'a- verb with wool, 'iijooł 'move away out of sight'
      (p. 474 def 15), is also Pos. Ib.

GM-8  T80a: reading (i). In the gas set the noun is linked to the 3i slot.           HYPOTHESIS (from GM-3, GM-4)
      Y&M's own analysis is that the 3i "represents the substance" that the noun
      names, and the noun co-occurs in 19/19 sentences, so this is the normal case,
      not an appositive afterthought. Reading (iii) (thematic) is not Y&M's
      analysis: they reserve "thematic" for 'ashzhish, 'ashhosh (TI-1) and label
      the gas-set 'a- 3i. Reading (ii) cannot be tested without judgments (T81e).
GM-9  Two uses of 3i 'a-.                                                             HYPOTHESIS (from GM-4, TI-4, TI-13)
      3i-PRO: indefinite pronoun, no noun, "does not usually represent a specific
      entity" (vehicles: Shił 'ílwod; people: da'iisdee'go).
      3i-CL: subject marker for one noun class (gaseous/free-moving NC matter) on
      *Øjooł; takes a noun; can be definite ('Áhí yę́ę ... 'i'ííjool 'the fog').
GM-10 PX-11 is restated as PX-11′: a bare noun can be linked to a slot only if the    HYPOTHESIS (replaces PX-11 for B-LINK)
      slot has a person value [π] OR a valued noun-class feature [cl]. 3i-PRO has
      neither (no noun: vehicles, people, Willie's IMP); 3i-CL has [cl: gas].
      PX-11 is left unchanged on indef-person as published.
GM-11 IS-10′ (IMP = [ABS, −def, no π], sort unvalued) holds for 3i-PRO only.         HYPOTHESIS (from GM-9; the definite
      3i-CL is [ABS, no π, cl: gas] and has no [−def]: 'áhí yę́ę, "the            readings are Y&M's translations)
      rainstorm", "a definite substance".
GM-12 Focus 2: in GM-6 the mover is the same proto-role (Dowty: movement, not        HYPOTHESIS (the reading; data GM-6)
      volition; causally independent: "to move independently"). The subject
      marker varies with the lexical class of the item filling the role (gas vs
      loose solid). The stem already selects NC matter (GM-2), so there are two
      layers of selection by kind of item. A role schema alone predicts one
      subject form for one role. This is evidence against the strong thesis.
GM-13 Not every gas mover takes 3i: níyol 'the wind blows' (Y&M: "(Ø)" prefix) has  SOURCED (p. 665 def 1)
      Ø, and with snow or dust the substance is a bił PP (Yas bił níyol, Łeezh bił
      níyol), not a subject. So 3i-CL is a property of *Øjooł bases, not of "gas"
      in general, as far as checked.

4. The data

Navajo forms are copied from the Y&M interface. Translations are Y&M's. The Leipzig line segments only as far as Y&M's prefix label does. Where it segments further it says so. NC = non-compact. Interface OCR in the translations ("movíng", "na·ajooł") is corrected only in the English and in that one headword spelling (§10).

Table 1. Gas set (3i). All 19 sentences in the nine entries.

Entry (Y&M prefix)Mover nouns in the example sentencesn
ʼadijooł p. 16 def 15 (ʼadi-)níłtsą́ rain1
dah ʼadiijooł p. 28 def 13 (ʼadin-)łeezh dust; kʼos cloud; łid smoke3
ʼeʼejooł p. 362 def 8 (ʼeʼe-)kʼos; ʼáhí yę́ę 'the fog'; chxííl snowstorm; łééchąąʼí ... bitłʼid 'a stink from the dog'4
haʼajooł p. 401 def 9 (haʼa-)łid1
ʼíjooł p. 466 def 9 ('a-((2)))níłtsą́ 'the rainstorm'1
naʼajooł p. 548 def 11 (na'a-((1)))kʼos; sigháala bilid 'cigar smoke'2
naanáʼoojoł p. 585 def 4 (naanáʼa-)kʼos1
ʼoojoł p. 681 def 18 ('a-((2)))chxííl; ʼáhí (×2); kʼos; łid5
chʼéʼéjooł p. 277 def 15 (chʼéʼe-)łid1
Sentences with no mover noun0

Table 2. Loose-solid set (Ø).

Entry (Y&M prefix)Mover nouns in the example sentences
dijooł p. 317 def 13 (di-((1)))ʼaghaaʼ wool (×2); tłʼoh hay
dah diijooł p. 341 def 13 (dii-((1)))wool (ʼaghaaʼ in the clause before; łaʼ 'some')
baa diijooł p. 341 def 14 (dii-((2)))tsiigháshchíín wig; bidághashchíín 'his false beard'
naajooł p. 576 def 6 (na-((2)))beʼaghaaʼ 'his wool'
chʼéjooł p. 277 def 9 (chʼí-)bitsiiʼ 'her hair'; ʼaghaaʼ léiʼ 'some wool'
yijooł p. 760 def 11 (Ø)ʼaghaaʼ (×3)
yíjooł p. 760 def 12 (Ø)tłʼoh; golchóón quilt; ʼayání bighaaʼ buffalo wool
hijooł p. 444 def 3 (hi-((1)))ʼééʼ clothes; chʼil deenínii tumbleweeds

Four examples with Leipzig lines:

(G1) p. 548 def 11Sigháala bilid tʼahdii kóneʼé naʼajooł
cigar 3.smoke still in.here about.3i.SBJ.move.NC (Y&M: prefix na'a-((1)) + *Øjooł; splitting naʼa- into na- 'about' + ʼa- 3i is ours, by comparison with naajooł, HYPOTHESIS)
'some cigar smoke is still floating around in here.'
(G2) p. 576 def 6Háíshąʼ beʼaghaaʼgo ʼakǫ́ǫ́ honibąąhgóó naajooł
who.Q 3.wool.be=SUB there fire.edge.along about.move.NC (Y&M: na-((2)) + *Øjooł; Ø 3rd subject; the gloss of the first two words is ours, from Y&M's translation)
'whose wool is it that's drifting (tumbling) about near the fire?'
(G3) p. 362 def 8ʼÁhí yę́ę ńléí dził báátis góʼąą ʼiʼííjool
fog =yę́ę yonder mountain 3.over.across beyond away.3i.SBJ.PFV.move.NC (Y&M: prefix ʼeʼe- = Pos. Ib ʼa- + 3i ʼa-, by GM-1 and the note in the entry; =yę́ę left unglossed beyond Y&M's 'the')
'the fog has drifted away over the mountain.'
(G4) p. 466 def 9 / p. 760 def 12Ńléí dziłtsį́į́dę́ę́ʼ níłtsą́ nihaaʼníjool (= nihaa ʼaníjool) · Dííshąʼ tłʼoh háádę́ę́ʼ ńjool
yonder mountain.base.from rain 1PL.to 3i.SBJ.arrive.NC.PFV · this.Q hay where.from arrive.NC.PFV (Y&M: 'a-((2)) vs Ø)
'the rainstorm reached us from the foot of the mountain.' · 'where did this hay come from?'

5. Minimal pairs

Loose solid (Ø)Gas (3i ʼa-)Meaning of the pairY&M link
dijooł (P deezhjool)ʼadijooł (P ʼadeezhjool)start to move along"Cf. ʼadijooł" (p. 317)
dah diijooł (P diijool)dah ʼadiijooł (P ʼadiijool)start off, billow up"Cf. dah ʼadiijooł" (p. 341)
naajooł (P naazhjool)naʼajooł (P naʼazhjool)drift aboutboth "*Øjooł"
chʼéjooł (P chʼíníjool)chʼéʼéjooł (P chʼíʼníjool)drift out horizontallyadjacent entries, p. 277
yíjooł (P nijool)ʼíjooł (P ʼaníjool)arrive driftingboth "*Øjooł"; Y&M "(= nihaa ʼaníjool)"

The perfectives are copied from the paradigm lines in each entry. The only difference in each pair is the ʼa- segment, which is what Y&M's prefix labels say. That the two members of a pair are one stem in two subject forms, and not two lexemes, follows Y&M's cross-references and identical stem notes HYPOTHESIS.

6. Answer to T80a

SOURCED The rain sentence is typical, not exceptional. Every gas-set example has a noun (19/19), and Y&M say the 3i "represents the substance functioning as the subject" and can stand for "a definite substance" (GM-3, GM-4).

HYPOTHESIS Of the three readings left by indef-subject, (i) fits best (GM-8). The noun is linked to the 3i slot. Reading (iii) goes against Y&M's labels. Reading (ii) (the noun as an appositive or topic) cannot be excluded from a dictionary. The word order gives nothing decisive: in (G4) the noun comes right before the verb, after a locative phrase, and in (G3) it comes first. That is T81e.

HYPOTHESIS The vehicle verbs on indef-subject go the other way: a noun takes Ø, no noun takes 3i (Chidí shił yílwod / Shił ʼílwod). So the 3i prefix has two uses (GM-9). One is 3i-PRO, an indefinite pronoun with no noun, where PX-11 holds. The other is 3i-CL, the subject marker that *Øjooł bases use for gaseous matter, which takes a noun. PX-11 survives only as PX-11′ (GM-10). The personless slot can be linked if it carries a valued class feature. PX-11 on indef-person is not edited. GM-10 supersedes it for this verb class. The definite readings also show that IS-10′'s [−def] belongs to 3i-PRO only (GM-11).

OPEN GM-7 is the loose end. The smoke-in-the-eyes verbs ʼajooł and ʼiijooł have a gas mover (łid), but Y&M label their ʼa-/ʼii- as the Pos. Ib 'away out of sight' prefix, not as 3i. If GM-1 is right that Ib ʼa- + 3i ʼa- gives ʼeʼe-, these verbs should show ʼeʼe-, and they don't. So either 3i-CL is not obligatory for all gas movers, or the 3i is hidden there (T81c).

7. Rulesets A and B

StepGas set (naʼajooł)Loose-solid set (naajooł)3i-PRO (Shił ʼílwod)
Stem selects the kind of itemNC matter (GM-2) ANC matter Avehicle (verb of running) A
Subject form decided3i, by the class of the mover (gas) AØ 3rd A3i, by indefiniteness A
Features clause syntax receives[no π, cl: gas] B[π: 3] B[no π, −def] B
LINK of a bare nounallowed (PX-11′ via [cl]) Ballowed (via [π]) Bblocked B

Focus 1. This is the clearest case so far of the word level (A) having to "know" a property of a clause-level (B) noun. The choice between ʼa- and Ø inside the left-headed prefix string depends on the kind of noun that the right-headed clause supplies. That is possible in two ways. (D1 below) Under VPISH the noun is merged inside the verb phrase, so word-internal Agree can see its class before the word is closed. (D2) Under the pronominal-argument view (Jelinek; Willie) the class is listed with the verb base, and B-level LINK checks that the noun matches it. The data here do not choose between them OPEN. Either way the changeover is not a clean wall. Information about the lexical item passes from B to A (D1) or is matched at B against something A fixed (D2) HYPOTHESIS.

8. Derivations

D1  Sigháala bilid ... na'ajooł 'cigar smoke is floating around' (G1), VPISH version
1  Merge  [NP sigháala bilid] 'cigar smoke'                      B (head-final NP)
2  Merge  [VP NP move.NC]  (*Øjooł; NP = mover, inside VP)       B; HISTORICAL HYPOTHESIS, not current VPISH; subject-base derivation OPEN
3  Merge  v (non-causative, "move independently")                A (the verb word begins here)
4  Merge  Pos. V subject head with [π:_ , cl:_]                  A
5  Agree  probe finds NP: [cl: gas] valued; no π valued          A; GM-9, GM-10 (3i-CL)
6  Spell  na- + 'a- + *Øjooł → na'ajooł                          A (Y&M prefix na'a-((1)))
7  LINK   noun ⟶ 3i-CL slot (here: same chain, trivially)        B; PX-11′ met
8  Output Sigháala bilid t'ahdii kóne'é na'ajooł                 matches Y&M p. 548 def 11

D2  The same, pronominal-argument version
1  Lex    base na'a-+*Øjooł listed with [cl: gas] on Pos. V       A (lexical, GM-6)
2  Merge  stem + Pos. V 3i-CL                                    A
3  Spell  na'ajooł                                               A
4  Adjoin [NP sigháala bilid] at clause level                    B
5  LINK   NP ⟶ Pos. V: check NP's class = gas                    B; PX-11′ via [cl]
          (a wool NP here would mismatch and need naajooł instead)

D3  Háíshą' be'aghaa'go ... naajooł 'whose wool is drifting about' (G2)
1  Merge  [VP (wool) move.NC]                                    B
2  Merge  Pos. V subject head [π:_, cl:_]                        A
3  Agree  wool: [cl: loose-solid] → no 3i; [π: 3] → Ø            A; GM-5
4  Spell  na- + *Øjooł → naajooł                                 A (Y&M na-((2)))
5  LINK   be'aghaa' (in the =go clause) ⟶ 3rd subject             B; PX-11 met
          (exactly how the cleft-like =go clause attaches is not modelled here)

D4  Shił 'ílwod 'I went (by a vehicle)' (indef-subject D4), for contrast
1  Merge  [PP shi- -ł]; [VP PP arrive.running]                   B
2  Merge  subject = 3i-PRO [no π, −def]                          A
3  Spell  'a + (y)ílwod → 'ílwod                                 A
4  LINK   none possible (no π, no cl)                            B; PX-11 / PX-11′

Merge builds the NP, the VP and the subject head. Agree is step 5 of D1 and step 3 of D3: the subject head copies a class value from the noun, and a person value if there is one. With 3i-CL it copies a class but no person. The step labels are ours. Y&M give only the prefix labels.

9. The role schema (Focus 2)

HYPOTHESIS In each minimal pair the mover has the same role. In Dowty's terms it is a proto-agent by movement and independent existence, with no volition, and Y&M's stem note "to move independently" is the same for both members. If the verb's arguments were roles only, the subject form would be fixed by the role, and both members would take the same prefix. They don't. The prefix tracks the class of the lexical item filling the role (GM-12). And the stem itself was already chosen by item kind (non-compact matter, GM-2), so the item enters twice. This is the strongest data yet on the site against the strong thesis "Navajo verbs take theta roles rather than lexical items as arguments". It does not touch the weak thesis that roles are visible: the 3i still fills the mover role.

OPEN The strong thesis is not refuted. A defender could put class features into Σ, for example a "gaseous theme" sub-role. That is a lexical feature renamed as a role, so it gives up the strong thesis in substance. Grant's revised view (the claim "does not hold cleanly") is consistent with this data. It stays GRANT'S CONJECTURE, since one verb family is not the language. The theta tracker gets a row (T-11): class-conditioned subject marking with role held constant. Chemistry is flagged only. Any correlate belongs to the Doping×Theta push.

10. Fidelity notes

All forms were copied from the interface HTML of the Colorado Y&M edition (saved on the box in [local working copy, not in the repository]). Its OCR shows in places: "movíng", "na·ajooł" in the naʼajooł sentence (written here with the headword's apostrophe), "nocompact", "independentIy", and "Ńleí" beside "Ńléí". No Navajo form was corrected except that one interpunct, which is noted. The page and def numbers are the interface's. They were not checked against the printed book. The -jooł regex also returned transitive *łjooł handling verbs and unrelated -jooł verbs ('cower', 'tip over', 'buckle'). Those were read and left out of Tables 1–2. The counts in Table 1 cover the example sentences in the nine gas entries. Y&M's example selection is not a corpus, so "0 sentences without a noun" is a fact about the dictionary, not a judgment that such sentences are ungrammatical (T81a). ʼahijooł (p. 59 def 4, wool, prefix 'ahi-((1))) and hadahajooł (p. 374 def 16, wool, hadahi-) have "gas" only in the generic stem note and are counted with neither set. The gloss "3i.SBJ" on ʼa- inside naʼa-, haʼa-, ʼeʼe-, chʼéʼe- and naanáʼa- is our segmentation of Y&M's prefix labels, guided by GM-1 and the minimal pairs. Y&M state it outright only for ʼadijooł, ʼeʼejooł, haʼajooł, and for ʼíjooł/ʼoojoł through 'a-((2)) = the p. 119/p. 681 3i entries.

11. Open questions

IDQuestionWhat would settle it
T81aCan a gas-set verb occur with no noun? Does it then mean 'something (gaseous) drifted' (3i-PRO) or still 'smoke/cloud' (3i-CL)?Texts (Y&M grammar section on 3i; Navajo texts in sources already on the box). OPEN
T81bIs 3i-CL limited to *Øjooł, or do other stems use it for natural forces or substances (fire, light, sound, water)?Y&M definition searches for "3i" + natural-force nouns. OPEN
T81cWhy do the smoke-in-the-eyes verbs ʼajooł/ʼiijooł (GM-7) have no 3i (expected ʼeʼe-)?Y&M paradigm pages; the Pos. Ib + 3i combinations listed under ʼeʼe-. OPEN
T81dIs 3i-CL [cl: gas] a noun class like the classificatory-verb classes, or a sort value ([−animate, −compact])? Does the ranking (Willie (70)) treat a 3i-CL subject as definite?A gas-set verb with a P object: is the P y- or b-? (cf. nihaa in (G4), a 1PL object, which does not decide it.) OPEN
T81eIs the gas noun clause-internal (LINK) or a topic or appositive (reading (ii))? Does D1 (VPISH, Agree with the noun) or D2 (lexical class plus B matching) hold?Word-order data (focus particles, ʼéí placement as in the p. 548 def 11 sentence Ńleí dziłghą́adi ʼéí kʼos ʼáłtsʼíísígo naʼajooł 'there's a little cloud drifting around on the mountaintop'), or a source's judgment. OPEN

T80a answered (GM-8, HYPOTHESIS; data GM-3–GM-6 SOURCED). PX-11 is restated as PX-11′ (GM-10) and IS-10′ is narrowed (GM-11). The published rules on indef-person and indef-subject are left as they were. Still open: T80b–e, T79a–d, T78b, T76b, T76c. TI-13 is resolved as HYPOTHESIS by GM-8–GM-10. Sources: Young & Morgan 1987 (entries as listed in §2 and §4). See also indef-subject, indef-person, absolutive-indef, nominal-adjunction.

The sun, sunlight, melting snow and lightning also take the 3i prefix, but without a noun: 3i-CL is not a "gas" class, and the 3i follows the light whether it moves or is moved (T81b)

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

2026-11-11 audit pass 3. Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference. See the pass 3 report.

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

Restated on the unified structure (2026-11-15). The indefinites (ʼa-, ʼadi-, areal ho-) are restated as personless pronouns in ordinary argument positions (H60): ʼadi- as the silent indefinite agent in Spec,vP that raises to Spec,TP by the default (H10b), ʼa- in VP. The 3i data on this page have not yet been individually restated. Status on the spine: see unified structure, §4, indefinite and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

T81b asked whether the 3i ʼa- that gas-subject found on smoke, cloud and rain movers (3i-CL) is limited to the stem *Øjooł. This run searched the Young & Morgan 1987 dictionary (the Colorado interface) for every entry with "3i" (385 entries), and for "sun", "the sun" and "light" (528 more). It then read the natural-force entries in full. The answer is no. Four more natural entities take the 3i SOURCED. (1) The sun, with the handling stem *Øʼaah 'solid roundish object': Y&M gloss the 3i as "something (i.e. the sun)" and add that the stem "does not usually refer to the independent movement of an object. The movement of the sun is an exception." (2) Light (*dlaad; also *łdíín 'shine', where the 3i is our segmentation). (3) Snow and ice melting (*dlííd). (4) A lightning bolt striking (*lchʼííł). Two facts change the gas-subject analysis. First, the sun is a "solid roundish object", not gas, so the class feature [cl: gas] is wrong. 3i-CL is not a classificatory noun class HYPOTHESIS. Second, the gas movers had a noun in 19 of 19 sentences, but the sun-set entries have it in 0 of 28, melting snow in 0 of 4, and lightning in 0 of 4 SOURCED. Plain words: the 3i marks a natural entity that acts on its own. The noun shows up only when the verb hasn't already fixed which entity it is HYPOTHESIS. One more result bears on Focus 2. The light keeps its 3i when it is the thing that moves ('i'didlaad) and when a person directs it ('i'dishdlaad). On Dowty's schema those are different proto-roles, so the marker follows the item, not the role HYPOTHESIS. The strong thesis stays OPEN.

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 Examples · 6 Answer to T81b · 7 Rulesets A and B · 8 Derivations · 9 The role schema (Focus 2) · 10 Fidelity notes · 11 Open questions

Notation is the same as on gas-subject. No Syntax · Semantics · Interface conventions file is in the repo or the context folder. A = the word level (prefixal, left-headed). B = the clause level (right-headed). Σ = the role schema, with Dowty's proto-roles as the baseline. 3i is Y&M's indefinite person. SRO = Y&M's "solid roundish object" (the classificatory stem class of *Øʼaah). 3i-PRO and 3i-CL are from GM-9. Merge builds structure: it puts two pieces together into a bigger one. Agree values features: a head with an unvalued feature, such as person, number or a sort or class feature, copies the value from a matching item. LINK is the clause-level (B) association of a noun with a word-internal argument slot (PX-11, PX-11′).

1. Question

gas-subject split the 3i ʼa- into two uses (GM-9). 3i-PRO is an indefinite pronoun with no noun ('someone/something'). 3i-CL marks "gaseous/free-moving non-compact matter" on *Øjooł, takes a noun, and can be definite. It gave 3i-CL the feature [cl: gas] (GM-10) and left T81b open: is 3i-CL limited to *Øjooł, or do other stems use it for natural forces or substances? The answer also bears on T81d (is [cl: gas] a classificatory noun class or a sort value?) and on T81a (can 3i-CL occur without a noun?). If 3i-CL is a class of gaseous matter, only gases should take it. If it is something wider, other natural entities should take it too, including ones that are not gases.

2. Sources

SOURCED Young, Robert W. & William Morgan, Sr. 1987. The Navajo Language: A Grammar and Colloquial Dictionary, revised edition. Albuquerque: UNM Press. Read through the free University of Colorado searchable edition (verbs.colorado.edu/~mahu0110/ym/), word + definition search. Searches, with every result page fetched: (a) 3i: 385 entries (16 pages), all parsed and filtered by natural-force words (fire, light, sound, water, rain, snow, wind, smoke, cloud, sun, lightning, thunder, ice, …); (b) sun: 231 entries; (c) the sun: 81; (d) light: 216; (e) ná'oodlííd: 3; (f) prefix headwords ná'a-, náo-, 'i'di-, ch'í'di-, ha'a-, 'e'e-, 'oo-, 'adii-, bik'í'-, náá'a-, łe'di-, ch'é'e-, 'ałní-. Entries used are cited "Y&M p. N def M" as the interface gives them, listed in §4. They were not checked against the print (§10).

Already on the site and not re-opened: GM-1–GM-13 (gas-subject); PX-11 (indef-person); IS-10′ (indef-subject); Willie's inanimate absolutive ʼa- (absolutive-indef).

3. Rules first

NF-1  SUN SET. The handling stem *Ø'aah 'handle, move --- a solid roundish object'    SOURCED (p. 362 def 10; p. 581 def 14;
      is used with the sun as the mover, and Y&M say why this is odd: "The stem      p. 704 def 5; p. 362 def 9)
      'ááh does not usually refer to the independent movement of an object. The
      movement of the sun is an exception." Two prefix entries gloss the 'a- as
      "Pos. V, 3i subj. pron., something (i.e. the sun)": náá'a- (p. 581) and
      taa'a- (p. 704). 'e'e'aah 'west' is "A construction comparable to 'e'ejooł".
NF-2  The sun-set entries: ha'a'aah 'sun rises' (p. 401 def 12; ha'a- = ha- + 3i,     SOURCED (entries and prefix entries
      p. 401 def 5); 'e'e'aah 'sun sets, evening' (p. 362 def 10; 'e'e- = Ib 'a- +   as cited; Table 1)
      3i, p. 362 def 2); ch'é'é'aah 'sun comes out' (p. 277 def 16); taa'a'a' 'it
      moves among the clouds (the sun)' (p. 704 def 6); 'oo'aah 'spend a day'
      (p. 681 def 21; 'oo- = 3i + yí-, p. 681 def 20); bik'í'ooz'ą́ 'use up the
      day' (p. 207 def 19; bik'í'- = bik'í- + 3i, p. 207 def 14); dah 'adii'aah
      'mid-morning' (p. 28 def 15); yaa 'adi'aah 'mid-afternoon' (p. 17 def 5);
      'ałné'é'aah 'noon' (p. 100 def 8). 28 example sentences; none has a noun
      for the sun (jóhonaa'éí, shá).
NF-3  The sun as an ordinary goer: 'iighááh 'to set (a star, moon, sun)', stem        SOURCED (p. 474 def 1)
      *Øghááh 'walk, go -- one actor', prefix 'a-((1)) (Pos. Ib, GM-1), no 3i:
      Jóhonaa'éí ha'ąą 'ííyáá ... with an overt noun; likewise 'ooljéé' 'the moon'.
NF-4  LIGHT. The prefix strings 'i'di- (all three entries), ch'í'di-, ch'éé'di-       SOURCED (p. 466 def 10; p. 467 defs 2, 4;
      contain "'(a)-, Pos. IV-V, 3i obj.-subj. pron., something" + di- "relates to   p. 286 def 13; p. 280 def 15)
      fire". With the Ø classifier the light moves on its own: 'i'didlaad 'to       (p. 466 def 11; p. 466 def 17)
      move (light -- as sunlight ...)'. With the ł classifier a person moves it:
      'i'dishdlaad 'to reflect light, ... direct a beam (of light)', 1sg -sh-.
      Both entries give the prefix as 'i'di-((1)).
NF-5  The sun beside a light verb: ch'í'deełdlaad 'to come out, break through       SOURCED (p. 286 def 14; p. 280 def 16)
      (the clouds) (the sun)' and ch'éé'deełdlaad 'come back out (the sun)', both
      ł classifier (Y&M: "*łdlaad: to tear, rip" / "to cause to rip,
      tear"), both with the overt noun jóhonaa'éí / jóhonaa'áí (OCR as given),
      both with 3i in the prefix (NF-4).
NF-6  MELTING. ná'oodlííd 'melt and run off (snow and ice)' (prefix ná'a-((2)); every  SOURCED (p. 559 def 17; ná'a- p. 548
      ná'a- entry is ná- + "'a-, Pos. IV-V, 3i obj.-subj. pron."): 4 sentences, no   def 15, p. 549 def 11, p. 551 def 27,
      noun, Y&M translate "it (the snow)". náodlííd 'melt away, melt off (snow,     p. 553 def 4; p. 567 defs 9–10;
      ice)', náo- = ná- + yi-ni-, no 3i: Yas 'ałtso náoosdlid, with the noun yas.     p. 351 def 4)
      Y&M cross-reference the two ("Cf. ná'oodlííd"). Dook'o'oosłííd 'San
      Francisco Peak' has "'(a), Pos. V, 3i subj. pron., something" + ł-lííd 'melt
      off --- ice and snow'.
NF-7  LIGHTNING STRIKE. łe'dilch'ííł 'to strike the ground (a bolt of lightning)';     SOURCED (p. 516 defs 8–9)
      łe'di- = łe- 'into the ground' + 3i obj.-subj. + di- 'fire'. 4 sentences, no
      noun for the lightning ('ii'ni' is never used); the struck object is a bił PP.
NF-8  Thunder: 'adoo'nííł 'a rumble of thunder is rolling by', Y&M: "'a-, Pos. V,   SOURCED (p. 33 def 13), with Y&M's
      3i obj. pron., something ? + di-, Pos. VI, noise?". Y&M mark their own         own question marks
      segmentation as uncertain.
NF-9  Not 3i, as far as checked: níyol 'wind blows' (GM-13); 'iighááh (NF-3);        SOURCED (p. 665 def 1; p. 474 def 1;
      lightning flashing jiilgęsh (jii-) and hadahachih (hadahi-).                  p. 490 def 7; p. 374 def 14)

NF-10 T81b answered: 3i-CL is not limited to *Øjooł. The sun (*Ø'aah), light         HYPOTHESIS (from NF-1–NF-8; the
      (*dlaad), melting snow and ice (*dlííd) and a lightning strike (*lch'ííł) take  segmentations in NF-2, NF-6 rely on
      it. Wind, the sun as a "goer", and lightning flashes don't (NF-9).               Y&M's prefix entries)
NF-11 [cl: gas] (GM-10) is wrong as a value. The sun is an SRO in its own stem         HYPOTHESIS (from NF-1; bears on T81d)
      (NF-1), so 3i-CL cannot be a classificatory noun class. A sort value is the
      better fit. Working label [nat]: a natural entity or force that moves or
      changes on its own. The sorts of the members are gas, light, the sun, snow,
      lightning. PX-11′ is restated with [nat] in place of [cl: gas].
NF-12 Noun co-occurrence is not a property of 3i-CL. It goes with how much the      HYPOTHESIS (counts SOURCED: GM-3
      stem fixes the referent. *Øjooł covers cloud, smoke, fog, rain, dust,          19/19; NF-2 0/28; NF-6 0/4; NF-7 0/4)
      snowstorm and stench, and a noun says which (19/19). *Ø'aah + 3i means the
      sun ("something (i.e. the sun)"), *dlííd means snow or ice, and *lch'ííł
      means lightning, and no noun appears (0/36). Recoverability, not class,
      decides whether a noun appears.
NF-13 3i-CL without a noun exists (the sun set), so GM-9's "takes a noun" is         HYPOTHESIS (from NF-12; T81a for the
      dropped. T81a for the *Øjooł set itself stays open.                            gas set stays OPEN)
NF-14 The light keeps its 3i across valency: subject of Ø-classifier 'i'didlaad,     HYPOTHESIS (data NF-4, NF-5)
      object of ł-classifier 'i'dishdlaad (1sg agent). In ch'í'deełdlaad the 3i is
      the light (object), and jóhonaa'éí is the causer, linked to the Ø 3rd subject
      slot. (Y&M's "obj.-subj." label allows either slot. That the 3i is the
      object here is our reading.)
NF-15 Focus 2: in NF-14 the light's proto-role profile changes (Dowty: moving, sole  HYPOTHESIS
      argument vs. causally affected, moved by an agent), and the marker stays. In
      GM-6 the role stays and the marker changes with the item. Together they
      dissociate the marker from the Dowty role in both directions. On a
      Gruber/Jackendoff schema the light is Theme in both, so the dissociation
      holds for Dowty's schema only.

4. The data

Navajo forms are copied from the Y&M interface, and translations are Y&M's. The 3i segmentation of each prefix string comes from Y&M's prefix entries as cited in NF-2, NF-4, NF-6 and NF-7. Where Y&M give no prefix label, the table says so.

Table 1. Sun set (*Øʼaah with 3i), all example sentences.

Entry (Y&M prefix; 3i source)MeaningSentenceswith a sun noun
haʼaʼaah p. 401 def 12 (haʼa-; p. 401 def 5)sun rises; dawn overtakes30
chʼéʼéʼaah p. 277 def 16 (chʼéʼe-; treated as 3i on gas-subject, GM-3; that prefix entry was not re-opened)the sun comes out20
taaʼaʼaʼ p. 704 def 6 (taaʼa-; p. 704 def 5 "i.e. the sun")sun moves among clouds10
ʼeʼeʼaah p. 362 def 10 (no label in the entry; ʼeʼe- p. 362 def 2)sun sets; evening; spend the day; expire120
ʼooʼaah p. 681 def 21 (ʼoo-; p. 681 def 20)spend an entire day20
bikʼíʼoozʼą́ p. 207 def 19 (bikʼíʼ-; p. 207 def 14)use up the whole day10
dah ʼadiiʼaah p. 28 def 15 ('adii-((1)); ʼadii- entries are "3i obj. pron.")become mid-morning20
yaa ʼadiʼaah p. 17 def 5 ('adi-((1)); which ʼadi- entry is ((1)) is not shown)become mid-afternoon20
ʼałnéʼéʼaah p. 100 def 8 (ʼałní- only; the 3i inside -ʼéʼe- is our segmentation)become noon30
Total280

Most of these are time-of-day or day-span expressions. The sun's motion in them is lexicalized as time. That is a reason to expect no noun independent of NF-12 OPEN. The literal sun-motion sentences (haʼaʼaah, chʼéʼéʼaah, taaʼaʼaʼ; 6) also have no noun. The derived nouns haʼaʼaah 'east' (p. 401 def 10) and ʼeʼeʼaah 'west' (p. 362 def 9) are not counted.

Table 2. Other natural-force entries with and without 3i.

Entity3i form (noun?)Non-3i form (noun?)
the sunNF-2: 28 sentences, 0 nounsʼiighááh p. 474 def 1: jóhonaaʼéí overt (1/1)
lightʼiʼdidlaad p. 466 def 11: 2 sentences, no noun for the light; chʼíʼdeełdlaad, chʼééʼdeełdlaad: the sun overt as causer (NF-5)none found
snow, icenáʼoodlííd p. 559 def 17: 0/4 nouns ("it (the snow)")náodlííd p. 567 def 10: yas overt (1/1)
lightning strikełeʼdilchʼííł p. 516 def 9: 0/4 nounsflashing: jiilgęsh p. 490 def 7 (no 3i, no noun)
thunderʼadooʼnííł p. 33 def 13 (Y&M "3i ... ?")—
gas (for comparison)GM-3: 19/19 nounswind níyol (GM-13)

5. Examples

The Leipzig lines segment only as far as Y&M's prefix entries do. Glosses beyond Y&M's translation are ours and are marked. 3i.SBJ/3i.OBJ follows our reading of Y&M's "obj.-subj." label in each verb (NF-14).

(S1) p. 401 def 12Shį́įgo ʼahbíínígo hanáʼátʼááh
summer=SUB morning=SUB up.out-again-3i.SBJ-move.SRO.REP (Y&M: R form hanáʼátʼááh of ha'a'aah; ha- 'up out' + 3i ʼa- per p. 401 def 5; the ná- 'again' gloss is ours)
'the sun comes up early in the summertime.'
(S2) p. 474 def 1Jóhonaaʼéí haʼąą ʼííyáá dóó chahóółhéél
sun beyond.hill away.out.of.sight-PFV-go.SG=and darkness.came (Y&M: 'a-((1)) Pos. Ib + *Øghááh 'one actor goes'; Ø 3rd subject; the =and and 'darkness came' glosses are ours, from the translation)
'the sun sank behind a hill, and darkness fell.'
(S3) p. 559 def 17 · p. 567 def 10Tsé bikááʼgóó náʼoosdlid lá · Yas ʼałtso náoosdlid
rock 3.on.surface.along melt.off-3i.SBJ-PFV EVID · snow all melt.off-PFV (Y&M: náʼa- = ná- + 3i vs náo- = ná- + yi-ni-; the EVID gloss of lá is ours, from "I find that")
'I find that it (the snow) has already melted off on the rock.' · 'the snow is all gone, the snow has all melted off.'
(S4) p. 466 def 11Kʼosgo biniinaa tsésǫʼdę́ę́ʼ tʼáadoo nihił yah ʼiʼdidlaadí díkwíí shį́į́ yiską́
cloud=SUB because window.from NEG 1PL.with in away-3i.SBJ-fire-move.light=NEG several perhaps nights.passed (Y&M: ʼiʼdi- = Ib + 3i + di- 'fire'; *Ødlaad 'move -- light'; the gloss of the rest is ours, from the translation)
'the sun hasn't shone in through the window for several days because of the clouds.'
(S5) p. 466 def 17Dziłghą́ą́ʼdóó ńléí halgaijįʼ shizhéʼé bichʼįʼ ʼiʼdííłdláád, biiʼ adéestʼį́į́ʼ bee
mountain.top.from yonder plain.toward 1SG.father 3.toward away-3i.OBJ-fire-PFV.1SG-CAUS-rip.light, mirror 3.with (Y&M: 'i'di-((1)) + *łdlaad; the 1SG subject is from Y&M's translation 'I reflected'; 3i.OBJ is our reading)
'I reflected a beam of light from the mountaintop to my father on the flat, with a mirror.'
(S6) p. 286 def 14Kʼadę́ę ʼałnéʼéʼaah, ndi jóhonaaʼéí tʼah doo chʼíʼdeełdlaad da
almost become.noon, but sun still NEG out-3i.OBJ-fire-CAUS-rip.light NEG (Y&M: chʼíʼdi- = chʼí- + 3i + di- 'fire'; *łdlaad; 3i.OBJ and the causer reading of the sun are ours, NF-14)
'it's nearly noon but the sun hasn't come out yet.'
(S7) p. 516 def 9ʼAtiin bąąhgo tʼiis ʼííʼáhą́ą bił łeʼdoolchʼil
road 3.beside cottonwood stood=REL 3.with into.ground-3i.SBJ-fire-PFV-flash (Y&M: łeʼdi- = łe- + 3i + di-; *lchʼííł 'flash -- lightning'; the glosses of the relative are ours)
'lightning struck the cottonwood tree that stood by the road.'

6. Answer to T81b

SOURCED Y&M use the 3i for the sun moving (with the gloss "something (i.e. the sun)", twice), for light moving or being directed, for snow and ice melting off, and for a lightning bolt striking (NF-1–NF-7). They compare the sun construction to the gas one ("comparable to ʼeʼejooł") and flag the sun as the one case where *Øʼaah means independent movement.

HYPOTHESIS So T81b is answered: 3i-CL is not a property of *Øjooł (NF-10). Its members are natural entities that move or change on their own, and the sun is one of them. The sun is an SRO in its own stem, so the shared property cannot be a classificatory class. GM-10's [cl: gas] becomes a sort value [nat] (NF-11). That also answers the first half of T81d. It is a sort value, not a noun class, as far as these entries go. The definiteness half of T81d stays open.

HYPOTHESIS The noun facts split by stem, not by 3i-CL (NF-12). The gas stem is general, and a noun picks out the substance. The sun, melting and lightning stems pin down their entity, and no noun appears. The two minimal contrasts point the same way: the sun with a "goer" stem (S2), and snow with the non-3i náodlííd (S3). Both drop the 3i and add the noun. So among the natural-entity verbs, 3i with no noun and Ø with a noun are both attested, and so is 3i with a noun (gas). 0/36 is a fact about Y&M's examples, not a judgment that a noun is impossible (§10).

OPEN Wind (níyol) is a natural force that moves on its own and takes no 3i (GM-13, NF-9). So [nat] is necessary but not sufficient, or wind is special. Rain falling (nahałtin) and water verbs were not checked (T82c).

7. Rulesets A and B

StepSun (haʼaʼaah)Light, moved (ʼiʼdííłdláád)Sun overt (jóhonaaʼéí ... chʼíʼdeełdlaad)
Stem selects the kind of itemSRO handling stem, read as the sun (NF-1) Alight (*dlaad) Alight A
3i placedon the mover (no agent) Aon the light, as object Aon the light, as object A
Subjectthe 3i itself1sg -sh- AØ 3rd (causer) A
Features B receives[no π, nat] B[π: 1] subject; [no π, nat] object B[π: 3] subject; [no π, nat] object B
LINKnone (no noun) Bnone Bjóhonaaʼéí ⟶ [π: 3] (PX-11 as published) B

Focus 1. For the sun set the information about the item sits wholly on the A side. The base itself means "the sun", and B supplies nothing. That matters for T81e. Under D1 of gas-subject (VPISH, word-internal Agree with a merged noun), the probe on the 3i head needs a goal. With the sun there is no overt noun to agree with. D1 then needs a silent noun in the VP. D2 (the base is listed with its entity, and B checks any noun) needs nothing extra. So the sun set favors D2 or a silent-noun version of D1. It does not decide between them OPEN. In S6 the clause-level noun links to the ordinary person slot, not to the 3i, so S6 is consistent with plain PX-11 and is no new evidence for PX-11′ HYPOTHESIS.

8. Derivations

D1  Shį́įgo 'ahbíínígo haná'át'ááh 'the sun comes up early in summer' (S1)
1  Merge  [VP (mover) move.SRO]  (*Ø'aah, read as independent movement; NF-1)  B (head-final VP)
          mover position: no overt NP. VPISH needs a silent N[sun] here; D2-style
          listing needs nothing (§7)
2  Merge  v (non-causative; no agent: "independent movement")               A (verb word begins)
3  Merge  3i head [π:_, sort:_]                                             A
4  Agree  sort valued [nat] (from silent N, or listed with the base ha'a-   A; NF-11
          + *Ø'aah); π stays unvalued → 3i
5  Merge  ná- 'again' (Pos. Id/Ib, repetitive), ha- 'up out' (Pos. Ib)       A (left-headed prefix string)
6  Spell  ha- + ná- + 'a- + t'ááh (R stem) → haná'át'ááh                       A (Y&M R form)
7  Merge  [Shį́įgo] [ʼahbíínígo] adverbial clauses, left of the verb          B (right-headed clause)
8  LINK   none: no noun; the sun is recovered from the base                  B; NF-12

D2  'i'didlaad (S4) vs 'i'dííłdláád (S5): same light, two valencies
     S4: [VP light move.light] → v Ø-classifier (no cause) → 3i head:          A
         Agree [nat]; no π → 3i in subject function; Spell 'i- + 'a- + di- +
         dlaad → 'i'didlaad
     S5: [VP light move.light] → v ł-classifier (CAUSE) introduces an agent    A
         → 3i head on the light: Agree [nat] → 3i in object function
         → subject head: Agree with agent [π: 1] → -sh-/1sg (PFV 'i'dííł-)
         Spell 'i- + 'a- + di- + (PFV) + ł + dláád → 'i'dííłdláád
         [bii' adéest'į́į́' bee] instrument PP, right edge                      B
     The 3i stays on the light in both. Only the classifier and the agent are
     added (NF-14).

D3  jóhonaa'éí t'ah doo ch'í'deełdlaad da (S6)
1  Merge  [VP light move.light] + v ł (CAUSE, causer = sun)                   A
2  Agree  3i head on the light: [nat], no π → 3i (object reading)            A; NF-14 (our reading)
3  Agree  subject head on the causer: [π: 3] → Ø                             A
4  Spell  ch'í- + 'a- + di- + ł + dlaad → ch'í'deełdlaad                     A (Y&M ch'í'di-)
5  Merge  NEG doo ... da around the verb; t'ah 'still'                       B
6  LINK   jóhonaa'éí ⟶ Ø 3rd subject [π: 3]                                   B; PX-11 met

D4  Tsé bikáá'góó ná'oosdlid lá (S3) vs Yas 'ałtso náoosdlid
     3i version: [VP (snow) melt] → 3i head: Agree [nat] (snow recovered      A
     from *dlííd) → ná- + 'a- + PFV + dlid; no noun; LINK none                 B
     Ø version:  [VP yas melt] → 3rd head: Agree [π: 3] → ná- + yi-ni- (PFV)   A
     → náoosdlid; LINK yas ⟶ [π: 3]                                            B; PX-11 met

Merge builds the VP, the v layer and the person heads. Agree is where a person head copies values. With a 3i head it copies a sort value but no person. With an ordinary head it copies [π]. The step labels and the v/classifier analysis are ours (the classifier as a voice/valency head is a standard Athabaskan analysis, not opened this run). Y&M give only the prefix labels and the stem notes.

9. The role schema (Focus 2)

HYPOTHESIS gas-subject showed one direction: the role stays the same, the item changes, and the marker changes (GM-12). The light verbs show the other direction (NF-14, NF-15). In S4 the light is the only argument and moves on its own. In Dowty's terms it leans Proto-Agent: movement, independent existence. In S5 a person directs it, and it is causally affected and moved: Proto-Patient. The marker stays 3i. So under Dowty's schema the 3i is dissociated from the role both ways and follows the item (light, gas, sun, snow, lightning). That is more evidence against the strong thesis "Navajo verbs take theta roles rather than lexical items as arguments".

OPEN Two caveats keep the strong thesis open. (1) On a Gruber/Jackendoff schema the light is Theme in both S4 and S5, so a defender with a Theme role can say the 3i marks "a natural-entity Theme". That again builds an item sort into the role, which concedes the point in substance, as on gas-subject §9. (2) Willie's ʼa- is the inanimate absolutive patient (absolutive-indef). The 3i in S4 (intransitive subject) and S5 (object) is absolutive in both. So an absolutive role-like account also covers NF-14 if [nat] is treated as a sort. The data do not separate an "absolutive + sort" story from an "item class" story. Grant's revised view (the strong claim "does not hold cleanly") fits this data and stays GRANT'S CONJECTURE. Theta tracker: row T-12, a marker held constant across a proto-role change (light). Chemistry: flag only.

HYPOTHESIS Y&M's note on *Øʼaah adds an item-side fact. The sun is put into a classificatory handling stem (SRO) used for objects that someone carries, and it is the one SRO that moves "independently". So the item chooses the stem (the sun counts as roundish), and the verb then lacks the handler that the stem normally has. That is the selection-by-item layer of GM-12 again, here at the stem.

10. Fidelity notes

All forms were copied from the interface HTML (saved on the box in [local working copy, not in the repository]). The interface's OCR shows in "shiníng", "eveníng", "light ning", "somethíng", "chʼíʼdeełdłaad" (with ł for l in a cross-reference) and the stray interpuncts in the chʼíʼdi- prefix entry ("·(a)-") and in the R form "chʼíń·díłdlaʼ·". None of these was used in a Navajo form here, except jóhonaaʼáí (p. 280 def 16), quoted as given in NF-5 beside the usual jóhonaaʼéí. One sentence has "ʼiʼííʼ,ą́" (comma inside), which this page does not quote. Page and def numbers are the interface's. The "((1))/((2))" homograph numbers Y&M give in entries (for example náʼa-((2)), 'i'di-((1))) could not be matched to a specific prefix entry, because the interface lists homographs without the numbers. NF-4 and NF-6 rely on the fact that every entry for that prefix string contains the 3i. The sentence counts in Table 1 count full example sentences, not phrase glosses (ʼEʼeʼaahjí, 'on the west'). Y&M's examples are a selection, not a corpus, so 0/28 and 0/4 are facts about the dictionary, not ungrammaticality judgments. The search terms were English words in Y&M's definitions, so natural-force verbs defined without those words may be missed.

11. Open questions

IDQuestionWhat would settle it
T82aIs the 3i in chʼíʼdeełdlaad / ʼiʼdííłdláád really in object position (NF-14)? Does the paradigm of a ł light verb put the 3i in Pos. IV with a full subject series?Y&M paradigm pages for ʼiʼdishdlaad (the interface gives no full paradigm for p. 466 def 17); Y&M grammar on Pos. IV vs V. OPEN
T82bWhy does wind (níyol) take no 3i if [nat] licenses 3i-CL? Is [nat] tied to "moves as a body" (sun, cloud, light) vs. "blows"?Y&M wind entries (*yol, níłchʼi) with prefix labels. OPEN
T82cDo rain-falling (nahałtin), water and fire verbs use the 3i? The "3i" search returned water prefixes (tóʼii-, tódaʼdii-, tanaʼa-, p. 701–707) not read here.Read those prefix entries' verbs; Y&M fire entries with ʼadi- 'fire'. OPEN
T82dIs the noun-less sun set noun-less because the sun is recoverable (NF-12) or because the forms are lexicalized time expressions? Does any literal sun-motion 3i verb occur with jóhonaaʼéí?Texts with sunrise/sunset descriptions; a source's judgment. OPEN
T82eThe thunder verbs (ʼadooʼnííł, ʼadiʼniih): is the ʼa- the 3i (Y&M's "?") or the noise-thematic 'a-+di-?Y&M grammar on thematic ʼadi-; Young, Morgan & Midgette 1992 if a free copy exists. OPEN

T81b answered (NF-10, HYPOTHESIS; data NF-1–NF-9 SOURCED). T81d first half answered as HYPOTHESIS (NF-11: sort, not noun class); the definiteness half stays open. GM-9's "takes a noun" is dropped and GM-10's [cl: gas] becomes [nat] (NF-11–NF-13). The rules on gas-subject are left as published. Still open: T81a (gas set), T81c, T81d (definiteness), T81e, T80b–e, T79a–d, T78b, T76b, T76c. Sources: Young & Morgan 1987 (entries as listed in §3–§4). See also gas-subject, indef-subject, indef-person, absolutive-indef.

Rain, water and fire: the rainstorm takes the areal ho-, fire and flowing water take a plain third person, and the 3i appears only in drizzle and water-blister verbs, so the verb base, not the kind of thing, picks the subject prefix (T82c)

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

2026-11-11 audit pass 3. Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference. See the pass 3 report.

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

Verbs on this page → dictionary: yidzaa ‘happen to’ · hoołtį́į́ł ‘rain (progressing)’

Restated on the unified structure (2026-11-15). The indefinites (ʼa-, ʼadi-, areal ho-) are restated as personless pronouns in ordinary argument positions (H60): ʼadi- as the silent indefinite agent in Spec,vP that raises to Spec,TP by the default (H10b), ʼa- in VP. The 3i data on this page have not yet been individually restated. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Status on the spine: see unified structure, §4, indefinite and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

T82c asked whether rain, water and fire verbs use the 3i ʼa- that gas-subject and sun-subject found on gas, the sun, light, melting snow and lightning (3i-CL, sort [nat], NF-11). This run searched the Young & Morgan 1987 dictionary (the Colorado interface) for rain, water, fire, burn, flow, snow and tó, and for 20 prefix strings. It read the verbs and their prefix entries in full. The answer is mostly no SOURCED. (1) Rain as a storm, on the stem *łtįįh/*łtin 'to rain', takes the areal ho-/ha- ("3s ... space, area, things, impersonal it") in seven entries, never the 3i, and none of their 16 example sentences names the rain. (2) Falling rain and hail (naałtin), flowing water (*Ølį́, *Øgeeh) and burning fire (*Økʼą́ą́h, *ltłáád, *Ønood, *Øtséés) take a plain third person, usually with the noun (tó 'water', ńló 'hail', kǫʼ 'fire'). (3) The 3i does appear in two small sets: light rain (nʼdizhoł 'drizzle', nikidaʼdiłtsįʼ 'sprinkle') and water blisters (tóʼiiłtįįh, tóʼdiisóół). So nature verbs split three ways: Ø, 3i and ho-. The same rainstorm, in the same role as the mover, takes the 3i with *Øjooł and ho- with *łtį́į́ł. The fire-prefix di- takes the 3i for light and not for fire HYPOTHESIS. Neither the role nor the kind of thing fixes the prefix. The verb base lists it, as Y&M say of ʼa- in "some verb bases" HYPOTHESIS. The strong thesis stays OPEN.

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 Examples · 6 Answer to T82c · 7 Rulesets A and B · 8 Derivations · 9 The role schema (Focus 2) · 10 Fidelity notes · 11 Open questions

Notation is the same as on sun-subject. No Syntax · Semantics · Interface conventions file is in the repo or the context folder. A = the word level (prefixal, left-headed). B = the clause level (right-headed). Σ = the role schema, with Dowty's proto-roles as the baseline. 3i is Y&M's indefinite person ʼa-. 3s is Y&M's label for ho-/ha- "space, area, things, impersonal it", which is Willie's pleonastic or areal ho- (AR, PX-6 on indef-person). 3i-PRO, 3i-CL and [nat] are from GM-9 and NF-11. Merge builds structure: it puts two pieces together into a bigger one. Agree builds nothing. It values features: a head with an unvalued feature, such as person, number or a sort feature, copies a value from a goal. LINK is the ruleset-B relation between a bare noun and a person slot (PX-11/PX-11′). It is not Agree.

1. Question

sun-subject restated 3i-CL as a sort value [nat], "a natural entity or force that moves or changes on its own" (NF-11). It also noted that wind (níyol) takes no 3i (NF-9), so [nat] may be necessary but not sufficient. T82c asked about the other big natural forces: rain falling (nahałtin), water and fire. The "3i" search had also returned water prefixes (tóʼii-, tódaʼdii-, tanaʼa-) that were not read. There are three outcomes. If [nat] is enough to license the 3i, rain, water and fire verbs should take it. If they don't, something else picks the 3i: the role, the item, or the verb base. The answer bears on Focus 2. If the prefix follows the base and not the role, the role schema does not choose the prefix.

2. Sources

SOURCED Young, Robert W. & William Morgan, Sr. 1987. The Navajo Language: A Grammar and Colloquial Dictionary, revised edition. Albuquerque: UNM Press. Read through the free University of Colorado searchable edition (verbs.colorado.edu/~mahu0110/ym/), with every result page fetched. Word + definition searches: rain (382 entries), to (169), nahaltin (49), burn (126), fire (517). Definition-only searches: water (1076), flow (261), snow (272). Prefix entries, read in full: naha-, haha-, ʼaha-, ho-, niho-, nikiho-, ʼahoʼnii-, nikidaʼdi-, nʼdi-, na-, ʼaho-, ʼadi-, hadi-, chʼídi-, ndee-, dii-, hodi-, ʼa-, ʼiʼnii-, tóʼii-/tóʼdii-/tódaʼdii-. Entries are cited as "p. N def M", the interface's page and definition number.

Already on the site and not re-opened: NF-1–NF-15 (sun-subject); GM-1–GM-13, including níłtsą́ ʼadeezhjool (p. 16 def 15) and níyol (GM-13) (gas-subject); PX-6 and PX-11 (indef-person); TI-13 (indef-subject).

3. Rules first

RW-1  RAINSTORM SET. Seven verbs on *łtįįh/*łtin/*łtį́į́ł 'to rain' have the areal     SOURCED (p. 527 def 17, p. 526 defs 9, 12,
      3s ho-/ha- in their prefix string and no 3i: nahałtin 'to rain, be a           p. 528 def 6; p. 390 defs 10–11; p. 36
      rainstorm' (naha- = na- + ha-/ho- 3s), hahałtįįh 'start to rain' (haha-),     def 2, p. 38 def 12; p. 461 def 10;
      'ahałtįįh 'rain away' ('aha- = 'a- Ib 'away' + ha-/ho- 3s), hoołtį́į́ł 'be     p. 630 def 13; p. 635 def 18; p. 72
      raining along, be moving along (a rainstorm)' (ho-), nihoołtįįh 'rain as      defs 9–10)
      far as a point' (niho-), nikihoołtįįh 'start to rain' (nikiho-),
      'aho'niiłtįįh 'start to rain' ('aho'nii- = 'a- Ib 'away out of sight' +
      ho'níi- inchoative).
RW-2  No noun for the rain in the rainstorm set: 0 of 16 example sentences name       SOURCED (count; Table 1)
      the rain (níłtsą́ never appears).
RW-3  FALLING STUFF. naałtin 'to fall (rain), to rain' (na-((1)), *łtin "to fall --   SOURCED (p. 583 def 28)
      water, hail"), plain 3rd, no ho-, no 3i; 3/3 sentences have a noun: tó
      nááłtánígíí 'the rains that fell', ńló nihee nááłtą́ 'we had a hailstorm'.
RW-4  LIGHT RAIN with 3i. n'dizhoł 'to drizzle ... rain lightly' (n'di-((2)));      SOURCED (p. 674 def 3; p. 672 def 16,
      five of the six n'di- entries contain "'(a)-, Pos. IV-V, 3i obj.-subj.        p. 673 defs 1, 4, 5, 6, p. 674 def 2;
      pron."; one labels the '(a)- Pos. VIa. nikida'diłtsį' 'to sprinkle, rain       p. 632 defs 4–5)
      lightly' (nikida'di- = niki- + da- + 3i '(a)- + di- inceptive). 0/3
      sentences have a noun.
RW-5  WATER FLOWING. Twelve *Ølį́/*Øgeeh 'flow' verbs have no 3i: ch'ínílį́, háálį́,    SOURCED (p. 290 def 3, p. 427 def 15,
      nílį́, yílį́, binázlį́, 'ahéénílį́, ninílį́, ndaazlį́, haageeh, hadiigeeh,        p. 643 def 8, p. 762 def 7, p. 230 def 6,
      niigeeh, yígeeh. Where a sentence is given, the liquid is named: tó 'water',  p. 52 def 1, p. 647 def 5, p. 604 def 18,
      dił 'blood', tókǫ́'í 'oil', and once níłtsą́ 'rain' (yígeeh, p. 759 def 18:     p. 424 def 12, p. 386 def 9, p. 667 def 4,
      "the rain washes one into the ground"). 'íílį́ and 'iigeeh 'flow away' have
      'a-((1)) (by meaning Pos. Ib 'away').
                                                                                    p. 759 def 18; p. 475 def 14, p. 473 def 19)
RW-6  WATER BLISTERS with 3i. tó'iiłtįįh 'to break out (a water blister)'           SOURCED (p. 707 defs 23–24, 19–20;
      (tó'ii- = tó- 'water' + 'a- 3i + yi-; stem *łtįįh "to rain, sprinkle");        p. 706 defs 39–40)
      tó'diisóół 'form a water blister' (tó'dii- = tó- + 3i '(a)- + dí- + yi-);
      tóda'diisóół 'form blisters, pox'.
RW-7  FIRE BURNING, no 3i. ch'ídeek'ą́ą́h 'burn out horizontally (a fire)' (ch'ídi-    SOURCED (p. 282 defs 3–4; p. 381 defs
      = ch'í- + di- "relates to fire"); hadik'ą́ą́h 'burn up out (a fire)' (hadi- =     22–23; p. 605 defs 5–6; p. 342 def 13;
      ha- + di- fire); ndeek'ą́ą́h 'burn as far as a point' (ndee- = ni- + di- fire);   p. 664 def 2; p. 384 def 3)
      diiltłáád 'catch fire', nitséés 'go out (a fire)', hadinood 'burst into
      flame'. kǫ' 'fire' is overt in 6 of the 7 fire sentences of the last three.
RW-8  AREA BURNING with ho-. hodiłid 'to burn off (an area)' (hodi- = ho- 3s + di-):  SOURCED (p. 450 def 25)
      Dziłdi ńdaniheeshį́hę́ęgi hodíílid lá 'our summer range burned off in the
      mountains'.
RW-9  SNOW, no 3i: yíchííł 'to snow' (no prefix label), dichííł, niichííł, yidzaas,  SOURCED (p. 757 def 15, p. 313 def 3,
      yiisááł 'fall (a sparse snow)' (yas overt); 'i'niichííł 'start to snow'         p. 665 def 17, p. 759 def 11, p. 797
      has 'i'nii-, whose 'i- Y&M call "3i ... used as a thematic prefix in some     def 1; p. 469 defs 4–5)
      bases".

RW-10 T82c answered: rain as a storm takes ho- (AR), not 3i; falling rain and hail,   HYPOTHESIS (from RW-1–RW-9; the
      flowing water, fire and snowfall take a plain third person; the 3i appears     segmentations rely on Y&M's
      only in light-rain and water-blister bases.                                    prefix entries)
RW-11 [nat] is not sufficient for 3i-CL. Natural-entity subjects split three ways:    HYPOTHESIS (from RW-10, NF-9, NF-10)
      Ø (fire, water, hail, snow, wind), 3i (gas, sun, light, snow melting,
      lightning strike, drizzle, blisters), and ho- (rainstorm, an area burning).
RW-12 Same entity, same role, different prefix: the rainstorm as the mover is 3i     HYPOTHESIS (data SOURCED: p. 16 def 15;
      with *Øjooł (níłtsą́ 'adeezhjool) and ho- with *łtį́į́ł (hoołtį́į́ł). Same thematic  p. 461 def 10; p. 286 def 13; p. 282
      di- 'fire' and the same Ib ch'í- 'out', different entity: 3i for light        def 3)
      (ch'í'di-), none for fire (ch'ídi-).
RW-13 SELECTION. The subject marker of a nature verb (Ø, 3i or ho-) is listed with    HYPOTHESIS (from RW-11–RW-12; cf. Y&M
      the verb base (stem + its prefix string), in ruleset A. It is not computed     p. 116 def 5, 'a- "occurs as a thematic
      from Σ (the role) or from [nat] (the item) alone.                              prefix in some verb bases")
RW-14 PX-11 holds for ho-: the areal ho- has no [π], and no noun is linked to it     HYPOTHESIS (RW-2 is consistent with
      (RW-2, 0/16). The noun appears with the plain 3rd (RW-3, RW-5, RW-7), and    PX-11 and PX-6; not a judgment)
      níłtsą́ 'rain' itself is overt with the Ø-3rd flow verb yígeeh (RW-5).

4. The data

Navajo forms are copied from the Y&M interface, and translations are Y&M's. Segmentations of the prefix strings come from Y&M's prefix entries as cited in §3. "Sentences" counts full example sentences in the entry, not phrase glosses.

Table 1. Rain on *łtįįh/*łtin: subject marker and nouns.

Entry (prefix; marker)MeaningSentenceswith a rain noun
nahałtin p. 527 def 17 (naha-((2)); ho- 3s)to rain, be a rainstorm10
hahałtįįh p. 390 def 11 (haha-; ho- 3s)start to rain10
ʼahałtįįh p. 38 def 12 ('aha-((1)); Ib + ho- 3s)rain away, stop raining10
hoołtį́į́ł p. 461 def 10 (ho-; 3s)be raining along (a rainstorm)30
nihoołtįįh p. 630 def 13 (niho-; 3s)rain as far as a point30
nikihoołtįįh p. 635 def 18 (nikiho-; labelled "3i", see §10)start to rain40
ʼahoʼniiłtįįh p. 72 def 10 (ʼahoʼnii-; Ib + hoʼníi-)start to rain30
Subtotal, ho- set160
naałtin p. 583 def 28 (na-((1)); Ø 3rd)to fall (rain, hail)33 (tó, ńló)
nʼdizhoł p. 674 def 3 (n'di-((2)); 3i, RW-4)drizzle20
nikidaʼdiłtsįʼ p. 632 def 5 (nikidaʼdi-; 3i)sprinkle10
tóʼiiłtįįh p. 707 def 24 (tóʼii-; 3i)break out (a water blister)1tó written apart (§10)

neestin 'it hangs (a cloud of mist ...)' (p. 625 def 27) is on *łtin with no 3i and a noun (tózhool 'mist'); it is not a rain verb and is not counted.

Table 2. Natural-entity subjects: three markers.

EntityØ 3rd (noun?)3i ʼa- (noun?)ho- (noun?)
rain, stormníłtsą́ with yígeeh (RW-5, noun)níłtsą́ ʼadeezhjool (GM, noun)RW-1 (0/16)
rain, hail as falling stuffnaałtin (3/3)light rain RW-4 (0/3)—
waterflowing, RW-5 (the liquid named in every sentence given)blisters, RW-6ho- 'be deep' (Willie, PX-6 page)
fireRW-7 (kǫʼ 6/7)none foundhodiłid (an area burns), RW-8
snowfalling, RW-9melting, NF-6—
light, sun, lightningsun as goer (NF-3)NF-2, NF-4, NF-7—
windníyol (GM-13)——

5. Examples

The Leipzig lines segment only as far as Y&M's prefix entries do. AR = the areal 3s ho-. Glosses beyond Y&M's translation are ours and are marked.

(R1) p. 461 def 10ʼAadę́ę́ʼ hoołtį́į́ł
from.there AR-rain.PROG (Y&M: prefix ho-, stem *łtį́į́ł 'to rain'; the PROG gloss is ours, from "to be raining along")
'here comes the rain.'
(R2) p. 527 def 17Naʼnishkaadgo shee nahóółtą́ą́ ńtʼééʼ dikos shidoolnaʼ
1SG.herd=SUB 1SG.with AR-rain.PFV and.then cough 1SG-caught (Y&M: naha- = na- + ha-/ho- 3s; P form nahóółtą́; word glosses other than the verb are ours, from the translation)
'it rained on me while I was out herding so I caught cold.'
(R3) p. 583 def 28ʼAdą́ą́dą́ą́ʼ ńló nihee nááłtą́
yesterday hail 1PL.with fall.PFV (Y&M: na-((1)), Ø 3rd subject; "Ńló naałtin, to hail")
'we had a hailstorm yesterday.'
(R4) p. 674 def 3ʼAdą́ą́dą́ą́ʼ tʼóó kónízahájįʼ nihee nʼdíízhoł
yesterday just a.short.while 1PL.with drizzle.PFV (Y&M: n'di-((2)), containing 3i '(a)- per RW-4; not segmented further)
'yesterday we had a brief and gentle shower.'
(R5) p. 632 def 5nihí ʼéí tʼóoʼ nihee nikidaʼdiłtsįʼ
1PL TOP just 1PL.with niki-da-3i-di-ł-rain.lightly (Y&M: nikidaʼdi- = niki- + da- + '(a)- 3i + di- inceptive; niki- and da- left unglossed; the TOP gloss of ʼéí is ours)
'(... it's raining quite hard but) we're just getting a sprinkle.'
(R6) p. 707 def 24... shílaʼ tó ʼiiłtą́
1SG.hand water 3i-PFV-ł-rain.PFV (Y&M: tóʼii- = tó- + ʼa- 3i + yi-; stem *łtįįh 'rain, sprinkle'; the sentence writes tó apart)
'(hoeing yesterday) I got a blister on my hand.'
(R7) p. 290 def 3Tséyiʼdę́ę́ʼ tó chʼínílį́
canyon.from water out-flow (Y&M: chʼíni-((2)), *Ølį́; Ø 3rd)
'water flows out of the canyon.'
(R8) p. 342 def 13... kǫʼ doo diiltłáad da ... ʼínda diiltła
fire NEG catch.fire NEG ... then catch.fire.PFV (Y&M: dii-((4)), *ltłáád 'to burn'; Ø 3rd)
'(because of the rain) the fire started slowly ... and then it burned.'
(R9) p. 450 def 25Dziłdi ńdaniheeshį́hę́ęgi hodíílid lá
mountain.at our.summer.range.at AR-fire-burn.PFV EVID (Y&M: hodi- = ho- 3s + di-; the EVID gloss of lá is ours)
'our summer range burned off in the mountains (last year).'

6. Answer to T82c

SOURCED Rain as a storm takes the areal ho-. Seven entries on the rain stem have ho-/ha- in their prefix string, and Y&M label it 3s, "space, area, things, impersonal it" (RW-1). Falling rain and hail, flowing water, fire and falling snow take a plain third person, and the noun is usually there (RW-3, RW-5, RW-7, RW-9). The 3i turns up only in light rain (RW-4) and water blisters (RW-6).

HYPOTHESIS So [nat] does not license the 3i by itself (RW-11). Wind was the first counterexample (NF-9). Fire, water, hail and the rainstorm on *łtįįh make it general. Two contrasts point to the verb base as the selector (RW-12, RW-13). (a) One entity in one role, the rainstorm as the thing that moves, takes the 3i with *Øjooł and ho- with *łtį́į́ł. The marker changes with the stem while the item and the role stay the same. (b) One thematic prefix, di- "relates to fire", with the same Ib chʼí- 'out', takes the 3i for light (chʼíʼdi-, p. 286 def 13) and not for fire (chʼídi-, p. 282 def 3). Y&M already say that ʼa- "occurs as a thematic prefix in some verb bases" (p. 116 def 5). The simplest statement is that 3i-CL is one of those thematic uses. It is listed in the base, and [nat] describes the bases that have it but does not predict which ones.

HYPOTHESIS RW-14. The areal ho- rain verbs never name the rain (0/16), as PX-11 predicts for a slot with no person value. The noun shows up on the Ø-3rd verbs (tó, ńló, kǫʼ), and the rain noun níłtsą́ itself appears with the Ø-3rd flow verb yígeeh (p. 759 def 18). The gas set's 3i with a noun (PX-11′) is still the one exception.

HYPOTHESIS Side results. T39d (is the ʼa- of ʼahoʼnii- the indefinite object or 'away'?): the 1987 dictionary's own prefix entry gives "Pos. Ib, away out of sight" (p. 72 def 9) SOURCED. That is Y&M's analysis, not a test against HMP's, so T39d stays OPEN. T39e (are weather verbs zero-place?): Y&M give the rain verbs an areal 3s subject ho-, so in their analysis they are not zero-place in the morphology. Whether ho- is an argument or an expletive is the open part.

7. Rulesets A and B

StepRainstorm, ho- (hoołtį́į́ł)Rainstorm, 3i (níłtsą́ ʼadeezhjool)Fire, Ø (kǫʼ ... diiltła)
Base listed with its markerho- + *łtį́į́ł A3i + *Øjooł (gas set) Adii- + *ltłáád, no marker A
Subject headAR, [no π, sort: area] A3i, [no π, nat] Aordinary 3rd, Agree gives [π: 3] A
Features B receives[no π] B[no π, nat] B[π: 3] B
LINKnone (0/16) Bníłtsą́ ⟶ 3i (PX-11′, the exception) Bkǫʼ ⟶ [π: 3] (PX-11) B

Focus 1. The choice of Ø, 3i or ho- is made wholly inside the word, by listing (ruleset A). Ruleset B never sees the choice itself. It sees only what the word sends up, a slot with or without [π], and it links a noun or not. The word is the changeover point: what A has fixed, B can only check. That is the same split as on sun-subject §7, now with three markers instead of two HYPOTHESIS.

8. Derivations

D1  'Aadę́ę́' hoołtį́į́ł 'here comes the rain' (R1)
1  Merge  [VP (mover) move.rain]  *łtį́į́ł, ł-classifier listed with the base        B (head-final VP)
          No overt noun. The base is listed with ho- (RW-13).
2  Merge  v (ł; no external agent: the rainstorm moves "along")               A (verb word begins)
3  Merge  subject head = AR ho- [π: none, sort: area]                          A; PX-6
4  Agree  nothing to value: AR has no [π] and its sort is listed               A (Agree is vacuous here)
5  Spell  ho- + (PROG) + ł + tį́į́ł → hoołtį́į́ł                                     A (Y&M form; PROG vowel not segmented)
6  Merge  'Aadę́ę́' 'from there', left of the verb                               B (right-headed clause)
7  LINK   none: AR has no [π] (PX-11); no rain noun in 16/16 (RW-14)            B

D2  níłtsą́ 'adeezhjool 'a rainstorm has started ...' (p. 16 def 15; gas-subject)
1  Merge  [VP (mover) move.gas]  *Øjooł                                          B
2  Merge  v (Ø; no agent)                                                       A
3  Merge  subject head = 3i [π: none, sort: _]                                  A
4  Agree  sort valued [nat] (from the noun or listed with the base; D1/D2 of    A; NF-11
          gas-subject, still OPEN)
5  Spell  'a- + di- + (PFV) + jool → 'adeezhjool                                 A (Y&M form)
6  Merge  níłtsą́ to the left of the verb                                        B
7  LINK   níłtsą́ ⟶ 3i slot (PX-11′)                                             B
   D1 vs D2: same item, same role (mover); only step 1's stem and step 3's
   listed marker differ (RW-12).

D3  ch'ídiník'ą́ą́' (P of ch'ídeek'ą́ą́h, sentence at p. 282 def 4) vs ch'í'deełdlaad (NF-5)
     fire:  [VP fire burn] → v Ø → ordinary 3rd head: Agree [π: 3] → Ø          A
            Spell ch'í- + di- (fire) + (PFV) + k'ą́ą́' → ch'ídiník'ą́ą́'             A
            no fire noun; the burning is given by the clause dibé bichaan diiltłago  B
     light: [VP light move.light] → v ł (CAUSE) → 3i head on the light:          A
            Agree [nat], no π → Spell ch'í- + 'a- + di- + ł + dlaad             A
            jóhonaa'éí ⟶ [π: 3] causer                                          B
   Same Ib ch'í- and same di- 'fire'; the 3i is present only in the light base.

D4  'Adą́ą́dą́ą́' ńló nihee nááłtą́ (R3) vs nahóółtą́ą́ (R2): falling stuff vs a storm
     R3: [VP hail fall] → v ł → ordinary 3rd head: Agree [π: 3] → Ø              A
         Spell na- + (PFV) + ł + tą́ → nááłtą́; LINK ńló ⟶ [π: 3]                  A / B
     R2: [VP (storm) rain] → v ł → AR head ho- (listed) → Spell na- + ho- +       A
         (PFV) + ł + tą́ą́ → nahóółtą́ą́; LINK none                                  B
   One stem family (*łtin), two bases: the falling stuff is a Ø 3rd with a
   noun; the storm over the area is AR with no noun.

Merge builds the VP, the v layer and the subject head. Agree is where a head copies values. With an ordinary third-person head it copies [π: 3]. With AR or a listed 3i there is no person to copy, so Agree values at most a sort. The v/classifier layer and the step labels are ours. Y&M give only the prefix labels, the stems and the forms. The vowel and the perfective pieces are not segmented here.

9. The role schema (Focus 2)

HYPOTHESIS RW-12(a) is the cleanest test so far of "the role picks the marker". The rainstorm moving toward or along a place is the only argument in both verbs. In Dowty's terms it leans Proto-Agent in both (movement, independent existence), and on a Gruber/Jackendoff schema it is Theme in both. The marker is 3i in one and ho- in the other. So neither proto-role schema predicts the marker. gas-subject showed the marker changing with the item when the role stays the same (GM-12). sun-subject showed it staying the same when the role changes (NF-14). Here the role and the item both stay the same, and the marker still changes. What is left is the verb base. That is more evidence against the strong thesis "Navajo verbs take theta roles rather than lexical items as arguments". In these verbs the lexical base, not the role, carries the choice.

OPEN Two caveats. (1) A defender can say ho- marks an "area" argument, so hoołtį́į́ł means 'it rains over an area' and its argument is the area, not the storm (compare hodiłid, where the area burns). Then the two verbs differ in which participant is the argument, not only in their marker. Y&M's gloss "to be moving along (a rainstorm)" is against that reading, but a gloss is not a test. (2) The 3i in RW-4 and RW-6 may be the 3i-PRO 'something' (as in taʼashnííł 'make mush', where the 3i is an object) rather than 3i-CL. The data do not decide. Grant's revised view (the strong claim "does not hold cleanly") fits this data and stays GRANT'S CONJECTURE. Theta tracker: row T-13, same item and same proto-role but a different marker, chosen by the stem (rainstorm). Chemistry: flag only.

10. Fidelity notes

All forms were copied from the interface HTML (saved on the box in [local working copy, not in the repository]). The OCR shows in "burn ing", "precípítate", "raín", "binídaaz1į́" (digit 1 for l), "niqht" and "to- ward". None of these was used in a Navajo form here. One headword displays as "hoołtį́į́" in the interface list, but the example sentences and the stem note have hoołtį́į́ł, which is used here. The nikiho- prefix entry (p. 635 def 17) labels its ho- "3i obj. pron., space, area". Every other ho- entry read here says 3s. This page treats it as a label slip and counts it with the ho- set; if it is not, RW-1 has six entries, not seven. The "((1))/((2))" homograph numbers (n'di-((2)), 'aha-((1)), naha-((2))) could not be matched to a single prefix entry, because the interface lists homographs without numbers. RW-1 and RW-4 rely on all or most entries for the prefix string sharing the marker; for nʼdi- one of six entries labels the ʼ(a)- Pos. VIa. ʼadikʼą́ą́h 'fire spreads, burns up' (p. 17 def 1) has 'adi-((2)); the ʼadi- entries include "3i obj. + di- fire" (p. 14 def 8) and "ʼa- Ib away" (p. 22 def 4, p. 24 def 12), so this verb is left out of RW-7 OPEN. In R6 the sentence writes tó as a separate word while the prefix entry makes it part of tóʼii-. The interface was not checked against the print edition. Y&M's examples are a selection, not a corpus, so 0/16 is a fact about the dictionary, not a judgment that a rain noun is impossible.

11. Open questions

IDQuestionWhat would settle it
T83aIs the areal ho- of the rain verbs the storm (the mover) or the area rained on (§9 caveat 1)? Does hoołtį́į́ł ever take a níłtsą́ noun or a locative that names the area as an argument?Texts with rain descriptions; Willie on pleonastic ho- with weather; a source's judgment. OPEN
T83bIs the 3i in the light-rain and blister bases (RW-4, RW-6) 3i-CL or the 3i-PRO 'something'? Does tó in tó ʼiiłtą́ behave like the gas set's noun (PX-11′)?Y&M grammar on the tó- prefixes; paradigms of tóʼiiłtįįh. OPEN
T83cDoes any rain verb on *łtįįh have a 3i, and does any gas-set verb on *Øjooł have ho-? A base with both would undo RW-13.A full stem search for *łtįįh/*Øjooł in the interface's stem index. OPEN
T83dWhich ʼadi- is ʼadikʼą́ą́h 'fire spreads' (p. 17 def 1): 3i + di- fire, or Ib 'away' + di-? If the first, fire has a 3i base after all.Y&M grammar on ʼadi- homographs; the print edition's numbered prefix entries. OPEN
T83eIf the marker is listed per base (RW-13), is there any productive rule left? For example, do all bases with thematic di- 'fire' + 3i denote light, and all without it denote fire?All ʼ(a)- + di- entries in Y&M, sorted by meaning. OPEN

T82c answered (RW-10, HYPOTHESIS; data RW-1–RW-9 SOURCED). NF-11's [nat] is kept as a description but is not sufficient (RW-11). T82b (wind) is partly bearing: wind joins fire, water and hail as a Ø nature verb, but why is still open. T39d and T39e get partial answers (§6). The rules on sun-subject and gas-subject are left as published. Still open: T82a, T82b, T82d, T82e, T81a, T81c, T81d (definiteness), T81e, T80b–e, T79a–d, T78b, T76b, T76c. Sources: Young & Morgan 1987 (entries as listed in §3–§4). See also gas-subject, sun-subject, indef-person, indef-subject, inchoative-hmp.

Hanging smoke, dust, cloud and fog: the same thing in the same role takes the 3i with two stems and a plain third person with three others, so the verb base, not the role and not the item, lists the subject prefix (T83c)

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

2026-11-11 audit pass 3. Local box file paths were removed from this page; files inside the repository are given by repo-relative path, and the source citation next to each removed path stays as the reference. See the pass 3 report.

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

Verbs on this page → dictionary: hoołtį́į́ł ‘rain (progressing)’

Restated on the unified structure (2026-11-15). The indefinites (ʼa-, ʼadi-, areal ho-) are restated as personless pronouns in ordinary argument positions (H60): ʼadi- as the silent indefinite agent in Spec,vP that raises to Spec,TP by the default (H10b), ʼa- in VP. The 3i data on this page have not yet been individually restated. Status on the spine: see unified structure, §4, indefinite and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

T83c asked whether any rain verb on *łtįįh has the 3i ʼa-, and whether any gas-set verb on *Øjooł has the areal ho-. One base with both would undo RW-13 (the marker is listed with the verb base). This run read every *Øjooł entry again and searched the Young & Morgan 1987 dictionary (Colorado interface) for the rain stems, *Øgoh 'flow', *Økał 'thin flat' and 'hangs'. SOURCED No base has both markers. 0 of 106 *Øjooł entries have ho-, and no rainstorm base on *łtįįh has a 3i. SOURCED The search also turned up a cleaner test. Y&M give 'it hangs (smoke, dust, a cloud, fog)' with five different stems. Two of them, dah ʼoojoł and dah ʼoogoh, carry the 3i. Three, dah yikał, dah shoogish and dah sikaad, carry none. The same nouns (łid 'smoke', łeezh 'dust', kʼos 'cloud', ʼáhí 'fog') appear on both sides. HYPOTHESIS With the role and the item both held fixed, the marker changes with the stem, so it is listed per base (HN-9). OPEN The strong thesis (theta roles, not lexical items, are the verb's arguments) is still not established. This page is more evidence against "the role picks the marker", which is only one way of reading it.

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 Examples · 6 Answer to T83c · 7 Rulesets A and B · 8 Derivations · 9 The role schema (Focus 2) · 10 Fidelity notes · 11 Open questions

Notation is the same as on rain-fire-subject. No Syntax · Semantics · Interface conventions file is in the repo or the context folder. A = the word level (prefixal, left-headed). B = the clause level (right-headed). Σ = the role schema, with Dowty's proto-roles as the baseline. 3i = Y&M's indefinite person ʼa-. AR = the areal 3s ho-/ha-. Ø = Y&M's plain third person, with no prefix. LINK = ruleset B's pairing of a clause noun with a person slot (PX-11, PX-11′). LINK is not Agree. Merge vs Agree, briefly: Merge builds structure, putting two pieces together into a bigger one (a stem and its theme, a prefix and a verb, a noun and a clause). Agree builds nothing. It lets a head copy feature values such as person [π] and number from another element. A choice of prefix that is listed in the lexicon is neither: it is a lexical fact, and Merge only puts it in place.

1. Question

rain-fire-subject found three subject markers on natural forces: AR ho- for the rainstorm, Ø for falling rain, hail, flowing water and fire, and 3i for drizzle and water blisters. It concluded that the marker is listed with the verb base (RW-13). Its §9 left a caveat. In RW-12 the 3i/AR contrast (níłtsą́ ʼadeezhjool vs hoołtį́į́ł) might not be the same role, because ho- could mark an area rather than the storm. T83c asked for the check that would undo RW-13: one base with both markers. T83e asked whether any productive rule is left.

2. Sources

SOURCED Young, Robert W. & William Morgan, Sr. 1987. The Navajo Language: A Grammar and Colloquial Dictionary, revised edition. Albuquerque: UNM Press. Read through the free University of Colorado searchable edition (verbs.colorado.edu/~mahu0110/ym/), with every result page fetched and saved on the box ([local working copy, not in the repository]). Stem and word searches: łtįįh (16 and 9 entries), łtin (61), łtį́į́ł (26), łtą́ (133, mostly unrelated stems), hodiłtįįh (0), neestin (2), goh (136), kał (135) and the definition search hangs. The 106-entry *Øjooł set from gas-subject ([local working copy, not in the repository]) was re-read for ho-/ha-. Entries used: p. 681 defs 12, 17, 18; p. 760 def 15; p. 695 def 26; p. 686 def 15; p. 625 def 27; p. 643 def 29; p. 708 def 35; p. 759 def 20; p. 782 def 4; p. 804 def 6; p. 707 def 24; p. 583 def 28.

SOURCED Dowty, David. 1991. Thematic proto-roles and argument selection. Language 67(3): 547–619. Used as on earlier pages for the proto-role lists. Not re-opened in this run.

Already on the site and not re-opened: RW-1–RW-14 (rain-fire-subject); NF-1, NF-10, NF-11 (sun-subject); GM-6 (gas-subject); PX-6, PX-11 and PX-11′ (indef-person, gas-subject).

3. Rules first

HN-1  T83c, GAS SIDE. None of the 106 entries on *Øjooł has the areal ho- in its      SOURCED (count over [local working copy, not in the repository];
      prefix string. The gas-set bases with 3i are 'adijooł (p. 16 def 15),          entries as listed)
      'adiijooł (p. 28 def 13), ch'é'éjooł (p. 277 def 15), 'e'ejooł (p. 362
      def 8), ha'ajooł (p. 401 def 9), na'ajooł (p. 548 def 11), 'íjooł (p. 466
      def 9) and 'oojoł (p. 681 def 18). The ha- in ha'ajooł is read as Ib 'up      HYPOTHESIS (the ha- prefix entry
      out' from the gloss, not the areal ha-.                                        was not opened)
HN-2  T83c, RAIN SIDE. No rainstorm base on *łtįįh/*łtin/*łtį́į́ł has a 3i. The 7       SOURCED (RW-1 re-checked; searches
      rainstorm bases have AR ho-/ha- (RW-1). The only 3i on the stem family is     in §2)
      tó'iiłtįįh 'break out (a water blister)' (p. 707 def 24), which is not a
      rain verb.
HN-3  The stem family *łtin/*łtįįh carries all three markers on different bases:     SOURCED (forms) / HYPOTHESIS (the Ø
      AR (RW-1), Ø on naałtin 'fall (rain, hail)' (p. 583 def 28) and neestin       of neestin; no prefix label, §10)
      'it hangs (a cloud of mist, smoke, dust, gas, an odor)' (p. 625 def 27;
      Y&M give the stem as *łtin 'to rain'), and 3i on tó'iiłtįįh.
HN-4  HANGING SET, 3i. Two bases mean 'hang, lie static (cloud, smoke, fog)' and      SOURCED (p. 681 defs 17, 18; p. 681
      carry 3i: dah 'oojoł (*Øjoł 'non-compact or gaseous', 'a-((2))) and dah 'oogoh (*Øgoh "to move in a    def 12 for 'oo-)
      flowing manner --- as water", 'a-((2))). Y&M p. 681 def 12: "'a- and
      Progressive Mode yi- combine to produce 'oo- in the 3rd person". Nouns:
      'áhí, k'os, łid (with 'oojoł); 'áhí, łid, łeezh, chííl (with 'oogoh).
HN-5  HANGING SET, Ø. Three bases mean 'hang' for the same kinds of thing and have    SOURCED (p. 760 def 15; p. 695 def
      no 3i: dah yikał (*Økał "to move --- something thin and flat", prefix Ø,      26; p. 686 def 15)
      "Dah + Progressive Mode = a static connotation"), dah shoogish (*Øgish,
      si-((4))) and dah sikaad (*Økaad "to spread, be expansive", si-((2))).
      Nouns: łid, łeezh (yikał); nát'oh 'cigarette smoke', łid, k'os
      (shoogish); k'os, 'áhí (sikaad). Add neestin with tózhool 'mist' (HN-3).
HN-6  SAME NOUN ON BOTH SIDES. łid 'smoke' 2 sentences with 3i, 2 with Ø; łeezh       SOURCED (counts, Table 1)
      'dust' 1 and 1; k'os 'cloud' 1 and 2; 'áhí 'fog' 2 and 1. Near-minimal
      pair: Dziłghą́ą́'dí k'os dah 'oojoł (3i) vs Dził bílátahdi k'os dah shoogish
      (Ø), both 'a cloud hangs over the mountaintop'.
HN-7  SAME ASPECT, SAME NOUN. The tightest pair is dah + Progressive on both sides:   SOURCED (forms and Y&M's mode notes)
      łid dah 'oojoł / łid dah 'oogoh (3i + yi- PROG → 'oo-) vs łid dah yikał (Ø +
      yi- PROG). The si- forms (shoogish, sikaad) differ in aspect too (§10).
HN-8  SAME STEM, DIFFERENT ITEM. On *Øgoh, flowing water is Ø (tó yigoh, p. 759       SOURCED (Y&M's own "Cf." at p. 681
      def 20; p. 782 def 4) and hanging fog, smoke, dust and snow are 3i ('oogoh).  def 17)
      Y&M cross-reference the two: "Cf. tó yigoh, water is flowing along". This
      is the GM-6 pattern (wool Ø vs gas 3i on *Øjooł) on a second stem.
HN-9  T83c ANSWER. No base has two markers, so RW-13 stands, and it can be made       HYPOTHESIS (from HN-1–HN-8)
      tighter: the unit that lists the marker is the base (stem + prefix string
      + the class of thing the base takes), not the stem (HN-3, HN-8) and not the
      role (HN-6, HN-7).
HN-10 RAIN, THREE WAYS. The noun níłtsą́ 'rain' occurs with 3i bases (Níłtsą́ dah      SOURCED (p. 643 def 29; p. 16 def 15;
      'oogoh, Níłtsą́ dah 'oojoł, p. 643 def 29; níłtsą́ 'adeezhjool) and with Ø      RW-3, RW-5)
      bases (tó/ńló naałtin; níłtsą́ with yígeeh, RW-5). The AR bases name no rain
      at all (RW-2).
HN-11 T83e, ONE CANDIDATE RULE TESTED. "Stems for formless or flowing stuff host 3i;   HYPOTHESIS (refutation from HN-8 and
      stems for a shape (thin flat, spread, slender) do not" fits HN-4/HN-5. It    NF-10; the rule was ours, not Y&M's)
      fails twice: water on the flowing stem *Øgoh is Ø (HN-8), and the sun on the
      solid-roundish stem *Ø'aah is 3i (NF-1, NF-10). The marker is a function of
      (base, item class) with no productive rule found yet.
HN-12 Smoke also shows up with a slender-line stem and no 3i: łid yót'ááh yít'i'      SOURCED (p. 804 def 6) / HYPOTHESIS
      'a column of smoke rising into the sky' (p. 804 def 6). The verb is cited    (the Ø; the yít'i' entry was not
      inside the entry for yót'ááh 'in the sky'.                                     opened)
HN-13 Focus 1. The choice between 3i and Ø is made inside the word (ruleset A),        HYPOTHESIS
      listed with the base. Ruleset B sees only whether the slot has [π]. It
      links the same noun in both cases, to the 3i slot by PX-11′ and to the [π:
      3] slot by PX-11. B never sees which stem was chosen.
HN-14 Focus 2. With role (a static located theme) and item fixed, the marker          HYPOTHESIS (the reading); strong thesis
      varies. So the marker does not read the role, and RW-12's caveat (ho- may     OPEN
      be an area) does not arise here, since no ho- is involved.

4. The data

Navajo forms are copied from the Y&M interface, and translations are Y&M's. "Sentences" counts the static (dah) example sentences in the entry.

Table 1. 'It hangs': five stems, two markers, the same nouns.

Entry (stem; prefix label; marker)Y&M glossSentencesNouns
ʼoojoł p. 681 def 18 (*Øjooł; 'a-((2)); 3i)"to drift along ... (smoke, fog, mist, vapor, dust, gas)"; "Dah ʼoojoł, to hang, to lie static (a cloud, fog bank, smoke)"3ʼáhí, kʼos, łid
ʼoogoh p. 681 def 17 (*Øgoh; 'a-((2)); 3i)dah ---, it hangs static (cloud, smoke, fog bank)4ʼáhí, łid, łeezh, chííl
níłtsą́ p. 643 def 29 (noun entry; 3i verbs)rain flows / hangs stationary2níłtsą́
Subtotal, 3i9
yikał p. 760 def 15 (*Økał; Ø; Ø)dah ---, it floats, it hangs (a layer of smoke, dust, scum)2 (+1 scum)łid, łeezh (tátłʼid)
shoogish p. 695 def 26 (*Øgish; si-((4)); Ø)"dah --, to hang (as a cloud of smoke)"3nátʼoh, łid, kʼos
sikaad p. 686 def 15 (*Økaad; si-((2)); Ø)"it sits, it lies spread out ... (... cloud of smoke, fog)"; "Dah sikaad, to sit in the form of a tuft, bunch or clump"2 (+1 mustache tuft)kʼos, ʼáhí
neestin p. 625 def 27 (*łtin; no label; Ø, HYPOTHESIS)it hangs (a cloud of mist, smoke, dust, gas, an odor)1tózhool
Subtotal, Ø8

The ʼoojoł entry also has two mover sentences, which are not counted: chxííl ʼoojoł 'a snowstorm is moving along' and ʼáhí ʼoojoł 'the fog is drifting our way'.

Table 2. Same stem, different item (*Øgoh) and same item, different stem (rain).

ContrastØ3iAR ho-
*Øgoh, by itemwater: tó yigoh (p. 759 def 20; p. 782 def 4)fog, smoke, dust, snow: dah ʼoogoh (p. 681 def 17)—
*Øjooł, by item (GM-6)wool, hay, wigsgas, cloud, smoke, rainnone (0/106)
rain, by basenaałtin (tó, ńló); yígeehdah ʼoogoh, dah ʼoojoł, ʼadeezhjool7 rainstorm bases, no noun
smoke, by baseyikał, shoogish, yítʼiʼ (HN-12)ʼoojoł, ʼoogoh—
*łtin, by base (HN-3)naałtin, neestintóʼiiłtįįhRW-1

5. Examples

The Leipzig lines segment only as far as Y&M's entries do. 3i.PROG = Y&M's ʼa- + yi- → ʼoo- (p. 681 def 12). UP = dah "up, static" (Y&M p. 681 def 17). Word glosses for the locatives and the stem labels (NC = non-compact matter, FLOW, TF = thin flat) are ours, taken from Y&M's translations and stem notes.

(H1) p. 681 def 18Halgaidi łid dah ʼoojoł
flat.at smoke UP 3i.PROG-move.NC (*Øjooł)
'a cloud of smoke hangs (lies) down on the flat.'
(H2) p. 681 def 17Łid dah ʼoogoh
smoke UP 3i.PROG-move.FLOW (*Øgoh)
'a cloud of smoke hangs.'
(H3) p. 760 def 15Ńléí halgai góyaa łid dah yikał
yonder plain down.there smoke UP PROG-move.TF (*Økał; prefix Ø)
'there's a layer of smoke hanging down on the plain.'
(H4) p. 681 def 17Łeezh dah ʼoogoh
dust UP 3i.PROG-move.FLOW
'a cloud of dust hangs.'
(H5) p. 760 def 15Ńléí tééh góyaa łeezh dah yikał
yonder valley down.there dust UP PROG-move.TF
'there's a layer of dust hanging down in the valley.'
(H6) p. 681 def 18Dziłghą́ą́ʼdí kʼos dah ʼoojoł
mountaintop.at cloud UP 3i.PROG-move.NC
'there is a cloud hanging over the mountaintop.'
(H7) p. 695 def 26Dził bílátahdi kʼos dah shoogish
mountain 3.tip.at cloud UP si-hang (*Øgish; si-((4)), no 3i)
'a cloud hangs over the mountaintop.'
(H8) p. 686 def 15Kʼos dah sikaad
cloud UP si-spread (*Økaad; si-((2)), no 3i)
'a bunched up cloud hangs.'
(H9) p. 643 def 29Níłtsą́ dah ʼoojoł
rain UP 3i.PROG-move.NC
'there is rain in progress (Lit. rain hangs stationary).'
(H10) p. 782 def 4Kʼad bikooh góyaa tó yigoh lá
now wash down.along water PROG-move.FLOW EVID (*Øgoh; no 3i; the EVID gloss of lá is ours)
'I see that there is water running in the wash now.'
(H11) p. 625 def 27Tózhool shikʼijįʼ neestingo shichidí tánásgis
mist 1SG.on.toward hang=SUB 1SG.car wash.PFV (*łtin; no prefix label; word glosses other than the verb are ours, from the translation)
'a cloud of mist hangs over me as I wash my car.'

6. Answer to T83c

SOURCED No. None of the 106 *Øjooł entries has ho- (HN-1). None of the seven rainstorm bases on *łtįįh has a 3i, and the stem family's only 3i is the water-blister verb (HN-2). RW-13 is not undone.

HYPOTHESIS RW-13 can be made tighter (HN-9). The stem *łtin/*łtįįh carries all three markers on different bases (HN-3). The stem *Øgoh carries Ø or 3i depending on the item (HN-8). So the stem is too small a unit. What is listed is the base: the stem, its prefix string and the class of thing it takes.

HYPOTHESIS The hanging set (HN-4–HN-7) removes the caveat left on rain-fire-subject §9. There, a defender could say that ho- names an area, so the 3i verb and the ho- verb did not have the same argument. Here there is no ho-. The single argument is smoke, dust, a cloud or fog lying still, which is the same role. In the tightest pair the aspect is also the same (dah + Progressive, HN-7). The 3i is still present with *Øjooł and *Øgoh and absent with *Økał.

HYPOTHESIS T83e is partly answered (HN-11). The one productive rule that fits the hanging set, "formless or flowing stems take the 3i", fails on water with *Øgoh and on the sun with *Øʼaah. No productive rule has been found, but other candidates have not been ruled out.

7. Rulesets A and B

StepSmoke, 3i (łid dah ʼoojoł)Smoke, Ø (łid dah yikał)Water, Ø (tó yigoh)
Base listed with its marker3i + *Øjooł AØ + *Økał AØ + *Øgoh (water) A
Subject head3i, [no π, nat] Aordinary 3rd; Agree gives [π: 3] Aordinary 3rd; Agree gives [π: 3] A
Spell-outʼa- + yi- → ʼoo- + joł Ayi- + kał Ayi- + goh A
Features B receives[no π, nat] B[π: 3] B[π: 3] B
dah, locativeMerge to the left of the verb Bthe same Bbikooh góyaa to the left B
LINKłid ⟶ 3i (PX-11′) Błid ⟶ [π: 3] (PX-11) Btó ⟶ [π: 3] (PX-11) B

Focus 1 (HN-13). Everything that differs between the first two columns is in the rows marked A. The B rows are the same apart from the feature bundle they receive. The word is where the choice is made, and the clause sees only its result.

8. Derivations

D1  Łid dah 'oogoh 'a cloud of smoke hangs' (H2) vs łid dah yikał (H3, minus the
    locatives)
1  Merge  [VP łid-theme  hang.FLOW]  *Øgoh            | [VP łid-theme  hang.TF] *Økał   B (head-final VP)
          The theme is VP-internal (VPISH). It is the only argument, a static
          located theme.
2  Merge  v (Ø classifier; no agent)                  | v (Ø; no agent)                 A (verb word begins)
3  Merge  subject head: 3i, listed with the base       | ordinary 3rd head               A (HN-4 / HN-5)
          [π: none, sort: _]                           | [π: _]
4  Agree  sort valued [nat]; nothing to copy for π     | [π] valued 3 from the theme     A (Agree: values, no new structure)
5  Move   none overt. The noun is not moved into the word. It stays at clause level           (HYPOTHESIS, as on earlier pages)
          and is paired by LINK in step 8.
6  Spell  'a- + yi- (PROG) → 'oo- ; + goh → 'oogoh     | yi- (PROG) + kał → yikał        A (Y&M p. 681 def 12; p. 760 def 15)
7  Merge  dah 'up, static' to the left of the verb    | the same                        B (right-headed clause)
8  LINK   łid ⟶ 3i slot (PX-11′)                       | łid ⟶ [π: 3] slot (PX-11)       B
   Steps 1, 4, 7 and 8 have the same role, item and clause frame. Only step 1's
   stem and step 3's listed head differ.

D2  tó yigoh (H10) vs 'Áhí dah 'oogoh (p. 681 def 17): one stem, two items
1  Merge  [VP tó  move.FLOW] *Øgoh                     | [VP 'áhí  move.FLOW] *Øgoh       B
2  Merge  v Ø                                          | v Ø                            A
3  Merge  ordinary 3rd head (base: *Øgoh + water)      | 3i head (base: dah + *Øgoh + gas) A (HN-8)
4  Agree  [π: 3]                                       | [nat], no π                    A
6  Spell  yi- + goh → yigoh                            | 'a- + yi- → 'oo- + goh          A
7  Merge  bikooh góyaa                                 | dah                            B
8  LINK   tó ⟶ [π: 3]                                  | 'áhí ⟶ 3i (PX-11′)             B
   The stem is the same. The item class, together with dah, picks the listed base.
   Whether dah is part of the base is OPEN (T84b).

D3  neestin (H11) vs dah 'oojoł (H6): rain stem and gas stem, the same 'hang'
1  Merge  [VP tózhool hang] *łtin (no dah)            | [VP k'os hang] *Øjooł           B
3  Merge  ordinary 3rd head (HYPOTHESIS, no label)     | 3i head                        A
6  Spell  neestin (P form; segmentation not given)    | 'oojoł                         A
8  LINK   tózhool ⟶ [π: 3]                             | k'os ⟶ 3i                      B
   The rain stem *łtin hosts a hanging-mist base with no 3i and no AR. The
   storm bases on the same stem have AR (RW-1), and the falling-rain base is
   Ø (RW-3).

Merge builds the VP, the v layer, the subject head and the clause around the verb (dah, the locatives). Agree only values features on a head that already exists: [π: 3] for an ordinary third-person head, a sort for the 3i, which has no person to copy. The listed choice of head in step 3 is neither Merge nor Agree. It is a lexical fact about the base.

9. The role schema (Focus 2)

HYPOTHESIS In Dowty's terms the hanging argument is the same in all five bases. It does not act on purpose, it is not sentient, and it is not moving or causing anything. It is a located theme, leaning Proto-Patient only weakly, if at all. If the subject marker were computed from Σ, the five bases would get one marker. They get two (HN-14). So the marker is not a reflex of the proto-role. This repeats RW-12 and GM-12 with a cleaner control: same role, same noun, same aspect in the tightest pair, and no AR.

OPEN This is evidence against one version of the strong thesis, "the role determines the marker". It does not touch the version Grant cares most about, that the verb selects a role slot rather than a lexical item. In fact HN-6 is consistent with role-slot selection. All five bases accept the same nouns, so none of them selects łid or kʼos as a lexical item. What each base does restrict is a class of thing (thin flat, flowing, non-compact). A role schema alone does not supply that class. So the most the data support is a role slot plus a class restriction listed with the base (cf. NF-11). Whether that counts as "theta roles, not lexical items" is a matter of definition, and it stays OPEN.

10. Fidelity notes

All forms were copied from the interface HTML saved on the box ([local working copy, not in the repository], [local working copy, not in the repository]). The interface was not checked against the print edition. (1) The homograph numbers 'a-((1))/((2)) cannot be matched to a printed prefix entry through the interface. For ʼoojoł and ʼoogoh the 3i reading rests on the p. 681 def 12 entry, which is on the same page and states that ʼa- + progressive yi- gives ʼoo-. (2) neestin shows only a P form and no prefix label. No 3i or AR is visible, but its Ø is HYPOTHESIS. (3) shoogish and sikaad are si- forms, so they differ from the progressive ʼoojoł/ʼoogoh in aspect as well as stem. HN-7 rests only on yikał, the one progressive Ø 'hang'. (4) The translations differ: yikał gets "a layer of smoke", ʼoogoh "a cloud of smoke". The stems classify the shape of the stuff. A defender could say the "item" differs at that level, and HN-11 is the reply. (5) The łtą́ search returned unrelated stems (peck, sit, slender object), which were not used. (6) In yítʼiʼ (HN-12) the verb is cited only inside another entry, and its own entry was not opened. (7) The words halgaidi, bílátahdi and shikʼijįʼ are glossed from the translations, not from their own entries.

11. Open questions

IDQuestionWhat would settle it
T84aIs the 3i/Ø split in the hanging set partly aspect? Do *Øjooł or *Øgoh have a si- neuter 'hang' with smoke, and does it keep the 3i?Paradigm pages ("View paradigm") for ʼoojoł, ʼoogoh, shoogish, sikaad; a si- form of the 3i bases
T84bIs dah part of the base that lists the 3i? Does *Øgoh take 3i with fog or smoke moving (no dah), or only when static? (cf. ʼiʼiigeeh p. 468 def 4, Ib. ʼiʼiijooł)All *Øgoh entries with a gas noun; the ʼiʼii- prefix entry
T84cDoes any *Økał, *Økaad or *Øgish base take a 3i with a gas noun? One would make HN-5 a fact about bases, not stems, as HN-8 already does for *Øgoh.A full read of the kał (135) and goh (136) sets saved on the box, and a kaad/gish search
T84dIs neestin's prefix string really free of 3i and AR (HN-3)? What do its other modes look like?The paradigm for p. 625 def 27; the nee-/nees- prefix entries
T84eT83e continued: is there any productive rule that predicts the 3i bases, for example a class of motion stems vs handling stems, and the sun as Y&M's noted exception (NF-1)?A list of all 3i bases with a [nat] subject (NF, GM, RW and HN sets), sorted by stem type

T83c answered (HN-9, HYPOTHESIS; data HN-1–HN-8 SOURCED). T83e partly answered (HN-11, HYPOTHESIS; carried on as T84e). Still open: T83a, T83b, T83d. The strong thesis stays OPEN (§9).

Where Young & Morgan put the indefinite ʼa-: in the listed prefix string of the base, not in the subject's person cell, at least for the ʼeʼe- verbs and the gas/wool pair on *Øjooł (T85e)

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

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). 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 derivation step annotated “(VPISH)” now reads “(VPISH hyp.)”: VPISH names the hypothesis that the subject starts in Spec,vP; it is not a node. 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.

T85e asked whether the choice between the 3i ʼa- and the plain third person (Ø) belongs to Y&M's verb entries. If it does, the light determiner of Grant's proposal is selected by the base, and we can ask where that selection is checked. In Y&M's own model paradigm for the ʼeʼe- prefix class (p. 364), a 3i subject has its own perfective row, ʼoʼoo-. Eleven Ø-classifier ʼeʼe- verbs have no first- or second-person subject: gas, the sun, unspecified vehicles, a household move, a vine, a splinter. In Y&M's example sentences, every one of them has its perfective in ʼiʼíí-, which is the plain third-person row, and none uses ʼoʼoo-. Y&M still say that the ʼa- of ʼeʼejooł "represent[s] the substance functioning as the subject" (p. 362 def 8). So the 3i of these verbs is the single ʼa- already inside the listed prefix ʼeʼe-, and the subject's person cell is the plain third person. SOURCED (forms) / HYPOTHESIS (analysis). Under analysis D of Light determiners, this listing can be checked at the step where the verb merges with its argument. Under analysis C it cannot.

Correction 2026-11-03: LD-3's inference is withdrawn and LD-9 is downgraded for the intransitive bases. Y&M's ʼa-((1)) table (p. 478) gives the same forms as its 3i row, and Y&M cite p. 364 only for the transitive ʼeʼe- bases. See two-tables (T88b).

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 Examples · 6 Answer to T85e (and part of T85d) · 7 Rulesets A and B · 8 Derivations · 9 Locality: D, C and A · 10 Fidelity notes · 11 Open questions

Notation is the same as on hang-subject. A = the word level (prefixal, left-headed). B = the clause level (right-headed). 3i = Y&M's indefinite ʼa-. Ø = the ordinary third person. d = Grant's light determiner (GRANT'S CONJECTURE, assessed here as HYPOTHESIS). [sort] is the class of thing a base takes. NC = non-compact or gaseous matter (Y&M's stem gloss for *Øjooł). No Syntax · Semantics · Interface conventions file is in the repo or the context folder.

1. Question

The essay Light determiners lists T85e: "Is the 3i/Ø choice part of Y&M's theme entries?" Earlier pages found that the marker goes with the verb base (RW-13 on rain-fire-subject, HN-9 on hang-subject). That was a conclusion from distribution: one base, one marker. It did not show that Y&M's lexicon records the marker as part of the base. There are two ways the lexicon could treat it, and they come apart. (i) The 3i is a person value, one row of a paradigm like 1sg or 3a, available to any base that has the row. (ii) The 3i is part of the base's listed prefix string. T85e decides which, at least for the bases opened. That matters for the light-determiner proposal because it says where d is chosen (§9).

2. Sources

SOURCED Young, Robert W. & William Morgan, Sr. 1987. The Navajo Language: A Grammar and Colloquial Dictionary, revised edition. Albuquerque: UNM Press. Read through the free University of Colorado searchable edition (verbs.colorado.edu/~mahu0110/ym/). Opened in this run:

  • the model paradigm for the ʼeʼe- class, p. 364 (paradigm view reached from ʼeʼejooł);
  • the model paradigm for na-((3)), p. 591 (reached from naʼanish, p. 549 def 24);
  • all 65 entries that match 'e'e, including the prefix entry ʼeʼe-, p. 362 def 2;
  • the prefix entries ʼadi- (p. 21 def 6) and naʼa- (p. 548 def 6);
  • the entries ʼadijooł (p. 16 def 15), naʼajooł (p. 548 def 11), naajooł (p. 576 def 6), yijooł (p. 760 def 11), ʼíjooł (p. 466 def 9), naʼanish (p. 549 def 24), bíníshchééh (p. 239 def 4) and ʼídlį́ (p. 465 def 21);
  • definition searches for "optional", "omitted", "with or without" and "3i subj".

Saved on the box in [local working copy, not in the repository].

Already on the site and not re-opened: RW-13, HN-9, HN-13, GM-6, PX-11 and PX-11′. The four analyses A–D come from the essay.

3. Rules first

LD-1  PERSON ROWS. Y&M's model paradigms list the 3i as a person row next to 1,  SOURCED (p. 364; p. 591)
      2, 3, 3a. In the 'e'e- class (p. 364) the 3 and 3i rows have the same
      form in every column but one. In the Ø/Ł-classifier perfective, row 3 is
      'i'íí- and row 3i is 'o'oo-. In the D/L-classifier perfective, both rows
      are 'o'oo-. In na-((3)) (p. 591) the rows are 3 naa-, 3o nei-, 3a nji-,
      3i na'a-, 3s naha-.
LD-2  THE 'e'e- LEMMA. Y&M p. 362 def 2: "'e- = 'a-, Pos. Ib, away out of sight  SOURCED (quote)
      + 'e- = 'a-, Pos. IV-V, 3i obj.-subj. pron., someone, something." So the
      prefix lemma itself contains one 3i, in a position labelled IV-V (object
      or subject).
LD-3  THE TEST CELL. The only cell of p. 364 that tells 3 from 3i is the          SOURCED (p. 364; 11 entries, Table 1;
      Ø/Ł perfective. Eleven Ø-classifier 'e'e- bases have no 1st- or 2nd-        forms from Y&M's example sentences,
      person subject (Table 1). In all 11 the perfective form in Y&M's example   not the interface's guessed paradigms)
      sentences begins 'i'íí- (row 3). None begins 'o'oo- (row 3i).
LD-4  Y&M CALL THAT 'a- THE SUBJECT. 'e'ejooł, p. 362 def 8: "'a-, Pos. IV-V is  SOURCED (quotes)
      the 3i pronoun ... Here it does represent the substance functioning as
      the subject of the verb." 'e'enééh, p. 363 def 6: "Lit. someone indefinite
      moves away with one." há'e'e'eeł, p. 405 def 16: "Lit.
      something floats away after it with him." 'e'e'aah, p. 362 def 9: "something
      roundish -- the sun -- moves away out of sight."
LD-5  SAME SLOT, OBJECT. In transitive 'e'e- bases, the same IV-V 'a- is the      SOURCED (p. 364 def 3; p. 364 row 1)
      object: 'e'eshjááh 'to plant, to put things (seeds) in the ground', 1sg
      perfective 'i'ííjaa' (row 1, Ø/Ł column, also 'i'íí-).
LD-6  SEPARATE HEADWORDS ON ONE STEM. On *Øjooł, Y&M give the Ø and 3i bases as  SOURCED (p. 548 defs 6, 11; p. 576 def 6;
      separate headwords with separate prefix lemmas and disjoint item lists:     p. 760 def 11; p. 466 def 9; p. 16 def 15;
      naajooł, na-((2)), 'tuft of wool, tumbleweed', vs na'ajooł, na'a-((1)),     p. 21 def 6)
      'cloud, smoke, gas, fog'. The lemma na'a- (p. 548 def 6) is "na-, Pos. Ib
      ... + 'a-, Pos.IV-V, 3i obj.-subj.pron., someone, something". Likewise
      yijooł (Ø, wool) vs 'íjooł ('a-((2)), rainstorm), and 'adijooł under the
      lemma 'adi- = "'a-, Pos. V, 3i subj.pron., something + di-". For the 3i
      bases, paradigm view says "A separate paradigm for this verb cannot be
      found" (nid 458, 9107, 10161). Y&M state the condition in the entry
      (p. 16 def 15): "when the subject is one of those entities listed above, it
      is represented in the verb by a 3i subject pronoun 'a-".
LD-7  THE 'SOMEONE' 3i AS A CELL. Elsewhere the 'someone' 3i is a row of an       SOURCED (p. 549 def 24; p. 239 def 4;
      ordinary person paradigm, cross-referenced to the base: na'anish "(3i        p. 465 def 21)
      subject 'a-, someone. See naashnish, to work ...)"; bíníshchééh lists
      "(bí'nílcheeh - 3i subject.)"; 'ídlį́ "(3i subject, someone. See nishłį́,
      to be.)".
LD-8  NO OPTIONAL 3i. Y&M mark optional classifiers ("Ø or d- classifier          SOURCED (searches, §2) / evidence from
      optional", p. 232 defs 1, 2, 4) and optional modes, but no entry found       silence only
      marks a 3i as optional or omissible.
LD-9  T85e ANSWER. Yes, for the bases opened. In the 'e'e- class, the 3i that    HYPOTHESIS (from LD-1–LD-6)
      Y&M read as subject is the IV-V 'a- of the listed prefix 'e'e-, and the
      subject's person cell is the plain row 3 (LD-3). On *Øjooł the 3i/Ø
      choice is a choice between two headwords (LD-6). In both, the 3i is part
      of what is listed, not a person value picked from a row.
LD-10 THE LINE IS NOT 'something' vs 'someone'. 'Someone' readings also use the  HYPOTHESIS (LD-4, LD-7)
      listed 'a- in the 'e'e- class (há'e'ebąąs 'a trip was made after water
      by someone', 'e'enééh). In other classes the same reading is a paradigm
      row (LD-7). Whether a 3i is listed or a cell depends on the prefix class
      and the base, not on its meaning.
LD-11 ENTRY FORMAT under D. A base is ⟨stem, prefix string, c-selection⟩. For    HYPOTHESIS (formal restatement)
      'e'ejooł: ⟨*Øjooł, 'a(Ib)+'a(IV-V), [__ dP[d:'a, sort:NC]]⟩. For naajooł:
      ⟨*Øjooł, na-((2)), [__ dP[d:Ø, sort:tuft]]⟩. The listed d is the IV-V 'a-
      of the prefix string. The marker is a function of base and sort together
      (HN-9), and both are visible on the label of one sister, dP (§9).
LD-12 T85d, PART. No base opened takes both 3i and Ø with one noun. The *Øjooł   SOURCED (LD-6, HN-6) / OPEN beyond
      pairs have disjoint item lists. HN-6's same noun with both markers was       these bases
      across different bases.
LD-13 FOCUS 1. Under D, d is the one item that is merged by B (inside dP, head-  HYPOTHESIS (D1–D4)
      final) and spelled out by A (in prefix position IV-V). The noun is merged
      and spelled out by B. The listing is stored in a word-level entry, but it
      is checked at a clause-level Merge (V + dP). This refines HN-13 ("the word
      is where the choice is made"): the word is where d is pronounced and where
      the listing is stored. Under D the check happens one step earlier, at B.
LD-14 COST FOR D. One surface slot (IV-V) hosts the listed 'a- whether it is the  HYPOTHESIS (LD-5)
      subject (D1) or the object (D4). If d reaches the word by head movement,
      the slot is not determined by where d starts. It is fixed by A's template.
      This is the essay's §5(b) problem (prefix order) seen in one class.

4. The data

Table 1. Ø-classifier ʼeʼe- bases (prefix lemma ʼeʼe- or háʼeʼe-) with no first- or second-person subject. The perfective column gives the form as it appears in Y&M's example sentence for the entry.

Base (p., def)Y&M gloss (item)Perfective in the examplep. 364 row it matches
ʼeʼejooł (362/8)drift away (fog, smoke, cloud, gas, stench)ʼiʼííjool3 (not 3i ʼoʼoo-)
ʼeʼeʼaah (362/10)set (the sun), become eveningʼiʼííʼą́ągo3
ʼeʼebąąs (362/6)bił --, ride away (unspecified wheeled vehicle)ʼiʼííbą́ą́z3
ʼeʼeʼeeł (362/14)float away (an unspecified vehicle)ʼiʼííʼéél3
ʼeʼeghááh (362/7)bił --, ride away (slow conveyance)ʼiʼííyáá3
ʼeʼemáás (363/5)bił --, roll away (inside a globular object)ʼiʼíímááz (printed without the initial ʼ)3
ʼeʼenééh (363/6)bił --, move away with household; "someone indefinite moves away with one"ʼiʼíínáá, ʼiʼííná3
ʼeʼetłʼóóh (365/8)grow away (a vine)ʼiʼíítłʼǫ́3
ʼeʼezhóósh (365/9)baa --, get a splinter in oneʼiʼíízhoozh3
háʼeʼebąąs (405/15)bił --, go away after (wheeled conveyance); "by someone"háʼiʼííbą́ą́z3
háʼeʼeʼeeł (405/16)bił --, go away after it by boatháʼiʼííʼééł3

Count: 11 of 11 match row 3, and 0 match row 3i. Not diagnostic, so left out: the D/L-classifier bases (ʼeʼelyeed → ʼoʼoolwod, 363/4; ʼeʼeʼnééh → ʼoʼooʼnaʼ, 362/15; háʼeʼeldlóósh, háʼeʼelyeed, háʼeʼetʼááh), because there rows 3 and 3i are both ʼoʼoo-. Also left out are the transitive bases with a first-person row (for example ʼeʼeshjááh, LD-5).

Table 2. The *Øjooł pairs (LD-6).

MarkerHeadword (p., def)Prefix lemmaItemsOwn paradigm page?
3inaʼajooł (548/11)naʼa-((1)) = na- + ʼa- 3i (548/6)cloud, smoke, gas, fogno ("cannot be found")
Ønaajooł (576/6)na-((2))tuft of wool, tumbleweed"View paradigm" offered
3iʼíjooł (466/9)'a-((2))rainstorm (níłtsą́)no
Øyijooł (760/11)Øwool—
3iʼadijooł (16/15)ʼadi- = ʼa- Pos. V 3i subj. + di- (21/6)gas, cloud, smoke, rain …no

5. Examples

The Leipzig lines segment only as far as Y&M's entries and the p. 364 table do. AWAY = Pos. Ib ʼa- 'away out of sight'. 'a = the IV-V 3i of the lemma. PFV.R3 = the perfective of p. 364 row 3. NC = non-compact matter. Glosses for words other than the verb come from Y&M's translations; they are ours.

(L1) p. 362 def 8Kʼos jóhonaaʼeí yichʼą́ą́h ʼiʼííjool
cloud sun 3.in.front.of AWAY-'a-PFV.R3-move.NC
'a cloud moved in front of the sun, a cloud moved in blotting out the sun.'
(L2) p. 405 def 15ʼAhbínídą́ą́ʼ shizhéʼé chizh bił háʼiʼííbą́ą́z
this.morning 1SG.POSS-father wood 3.with FOR-AWAY-'a-PFV.R3-move.wheeled
'my father went away after wood this morning (in a wagon).'
(L3) p. 364 def 3Kóhootʼéédą́ą́ʼ kohgo shidáʼákʼehgi ʼałtso ʼiʼííjaaʼ
this.time.last.year thus 1SG.POSS-field.at all AWAY-'a-PFV.1SG-handle.PL
'by this time last year I had finished planting on my farm.'
(L4) p. 548 def 11Sigháala bilid tʼahdii kóneʼé naʼajooł
cigar 3.POSS-smoke still in.here ABOUT-'a-move.NC(CI)
'some cigar smoke is still floating around in here.'
(L5) p. 576 def 6Háíshąʼ beʼaghaaʼgo ʼakǫ́ǫ́ honibąąhgóó naajooł
who.Q 3.POSS-wool-SUB [near.the.fire] ABOUT-3-move.NC
'whose wool is it that's drifting (tumbling) about near the fire?'

In L1, L2 and L3 the verb has the same ʼiʼíí- shape. Y&M read its ʼa- as the subject in L1 and L2 and as the object 'things' in L3. The segmentation of bilid as possessed łid is ours, from the translation. In L5, ʼakǫ́ǫ́ honibąąhgóó is glossed as one unit from the translation.

6. Answer to T85e (and part of T85d)

HYPOTHESIS (data SOURCED). T85e: yes. For the bases opened, the 3i is part of what Y&M list. It is not a person value chosen from a paradigm row. The clearest case is the ʼeʼe- class. Y&M's own model paradigm gives a 3i subject a distinct perfective, ʼoʼoo-. Yet every Ø-classifier base whose subject Y&M describe as "something", "an unspecified vehicle" or "someone indefinite" has the row-3 perfective ʼiʼíí- (LD-3, Table 1). Y&M nonetheless say the ʼa- stands for the subject (LD-4). The only way to fit both statements is that the subject 3i of these verbs is the one ʼa- in the lemma ʼeʼe-, and the person cell is the plain third person (LD-9). On *Øjooł, the 3i and Ø versions are separate headwords with separate prefix lemmas and separate item lists (LD-6). The "listed" half of HN-9 is therefore a fact about Y&M's lexicon, not only about distribution.

HYPOTHESIS One refinement (LD-10). Listing does not separate the nature 3i ('something') from the human 3i ('someone'). In the ʼeʼe- class, both are the listed ʼa-. In other classes (naʼanish, ʼídlį́), 'someone' is a row of an ordinary paradigm. So there are two ways to get a 3i, a listed d and a d from a person cell, and the prefix class decides which one a base uses.

OPEN T85d, in part. No base opened takes both markers with the same noun (LD-12), and Y&M never mark a 3i as optional (LD-8). That is evidence from silence in a dictionary. It is not a judgment that the other combination is ungrammatical.

7. Rulesets A and B

StepUnder D (kʼos … ʼiʼííjool, L1)Under C (same sentence)
Noun + dMerge [dP k'os d] Bd is not merged with the noun B (k'os at clause level)
Sort on dAgree d ← k'os, [sort: NC], inside dP Bnot yet available (no noun in the word) A
Base meets argumentV(*Øjooł, ʼeʼe-) + dP, c-selection checked on the sister's label BV + d on the spine; the sort is unvalued A
d into the wordhead movement (or post-syntactic attachment) to IV-V Aalready there A
Person cellrow 3 (no 1/2 person) Arow 3 A
Spell-out'a(Ib) + 'a(IV-V) + PFV → ʼiʼíí- + jool Athe same A
Listing checkdone at step 3 Bneeds a later step: upward Agree or a spell-out filter B→A

Focus 1 (LD-13). Under D, the line between A and B runs through the argument. d is built by B and pronounced by A. The noun is built and pronounced by B. Under C, d never belongs to B, and B's noun has to be matched to it later.

8. Derivations

D1  K'os jóhonaa'eí yich'ą́ą́h 'i'ííjool (L1), analysis D
    Assumes the moving cloud is an internal argument (VPISH; unaccusativity not tested, T87a).
1  Merge  [dP [NP k'os] d] ; d = 'a, [sort: _]        B (head-final: noun left of d)
2  Agree  d probes its complement: [sort: NC] from k'os B (inside dP; local)
3  Merge  [VP dP V] ; V = *Øjooł in base ⟨'e'e-, __ dP[d:'a, sort:NC]⟩
          c-selection is checked on the label of the sister: d:'a ✓ sort:NC ✓
                                                        B (head-final VP)
4  Merge  v (Ø classifier, no agent)                    A (verb word begins)
5  Move   d head-moves V → v (or attaches after syntax). It leaves k'os in dP.
          This is movement out of a complement, as in incorporation, so the
          essay's §5(a) problem (head movement out of a specifier) does not arise
          here IF step 3 is right (T87a).            B → A (the crossing)
6  Agree  subject person cell: no 1/2 → row 3 (Ø)      A
7  Spell  'a(Ib) + 'a(IV-V, = d) + PFV(Ø/Ł) → 'i'íí- ; + jool → 'i'ííjool
          (p. 364 row 3, not row 3i 'o'oo-)               A
8  Merge  [PP jóhonaa'eí yich'ą́ą́h] left of the verb      B
9  Move   the remnant [dP k'os t_d] reaches clause-initial position
          (how and where is OPEN, T87d)                  B
   The noun is spelled out in B. d is spelled out in A. The listing was checked at step 3.

D2  The same sentence, analysis C (d is a Voice-like head on the spine)
1  Merge  V = *Øjooł                                    A
2  Merge  d = 'a on the spine, [sort: _]; no noun in the derivation yet
                                                        A
3  Merge  v (Ø)                                         A
4  Spell  'i'ííjool (as D1 steps 6–7)                    A
5  Merge  k'os at clause level; PP left of the verb     B
6  ???    The base's [sort: NC] must now be checked against k'os. Two options:
          (i) d Agrees upward/outward with k'os (non-local; the probe is in A and
          the goal in B); or (ii) a filter at the interface pairs ⟨base, noun⟩ (= LINK,
          PX-11′, plus a sort check).                     B → A (late)
   C can state LD-11, but only as a look-back filter, not at Merge.

D3  L4 vs L5, analysis D: one stem, two listed bases
1  Merge  [dP [NP sigháala bilid] d:'a]           | [dP [NP be'aghaa'] d:Ø]      B
2  Agree  [sort: gas]                             | [sort: tuft]                 B
3  Merge  V in ⟨na'a-((1)), __ dP['a, gas]⟩ ✓        | V in ⟨na-((2)), __ dP[Ø, tuft]⟩ ✓ B
4  Merge  v Ø                                     | v Ø                          A
5  Move   d → word                                | d (null) → word              B → A
7  Spell  na + 'a + jooł → na'ajooł               | na + jooł → naajooł          A
   D predicts that [dP 'aghaa' 'a] fails at step 3 with na'a-((1)). Y&M list no
   such sentence. This is silence (LD-8), not a judgment.

D4  L3 'ałtso 'i'ííjaa', analysis D: the same IV-V 'a as the object
1  Merge  [dP pro d:'a] ('things', no overt noun; Y&M's gloss "things (seeds)")  B
2  Agree  [sort] from pro (what values it is OPEN, T87e)                           B
3  Merge  [VP dP V] ; V = *Øjááh in base 'e'e-                                     B
4  Merge  v; agent 1sg in Spec,vP                                                  A/B (VPISH hyp.)
5  Move   d → word                                                                 B → A
6  Agree  subject cell 1sg                                                         A
7  Spell  'a(Ib) + 'a(IV-V) + PFV.1SG → 'i'íí- ; + jaa' → 'i'ííjaa' (p. 364 row 1) A
   D1 and D4 put their d in one slot, IV-V, though one is a subject and one an
   object. The slot comes from A's template, not from where d started (LD-14).

Merge builds dP, VP, v and the clause. Agree only values features on heads that already exist. Move is head movement of d, or its post-syntactic equivalent, plus the clause-level movement of the remnant. The listed choice of d is a lexical fact about the base, and under D it is enforced at Merge (step 3).

9. Locality: D, C and A

HYPOTHESIS LD-9 says that what is listed is a pair: the base takes a d of a given kind together with a noun of a given sort (LD-11). Under D, d and the noun form one constituent before the base sees them. The pair is visible on one label at the moment the base merges, so the listing is ordinary local c-selection (D1 step 3). Under C, d is on the verbal spine and the noun is not yet in the derivation when d merges. The sort part of the listing has to be checked later, by non-local Agree or by an interface filter (D2 step 6). Under A (the prefix is the pronominal argument), the base can s-select a pronoun with a sort, which is local. The noun still has to match the pronoun afterwards through LINK. On locality, then, D and A both do better than C, and D does slightly better than A because its match happens inside one constituent (Agree in dP) rather than through LINK.

HYPOTHESIS Against D (LD-14), the ʼeʼe- class puts the listed ʼa- in a single slot whatever its role. If d really moves from its argument position, A's template has to override that position. This is the essay's §5(b) problem. It counts equally against C, which needs two spine heads spelled out in one slot.

OPEN The locality argument presupposes that the ʼeʼe- gas and vehicle argument is merged as the sister of V (D1 step 3). If it is an external argument, the local selector is v, and D's head movement comes out of a specifier, which is the essay's §5(a) problem. Unaccusativity diagnostics for these bases have not been opened (T87a).

10. Fidelity notes

All forms were copied from the interface HTML saved on the box ([local working copy, not in the repository], [local working copy, not in the repository]). The interface was not checked against the print edition. (1) The F/I/O/P/R lists in many entries are labelled "[Guess a paradigm]", which may mean the interface generated them. Table 1 therefore uses only perfectives that occur in Y&M's example sentences. (2) The p. 364 table was read from the interface's paradigm view. Its column heads ("PERFECTIVE Ø - Ł", "PERFECTIVE D - L") are taken as printed. (3) The ʼeʼe- paradigm also has a "PASS. A" row with ʼoʼoo- (Ø/Ł perfective). None of the Table 1 bases uses it. (4) The p. 591 table is the na-((3)) paradigm. naʼajooł belongs to na'a-((1)), and its own paradigm page could not be found, so naʼajooł's perfective naʼazhjool was not compared with a row (T87c). (5) No Navajo form here was supplied or checked by a speaker.

11. Open questions

IDQuestionWhat would settle it
T87aIs the single argument of the ʼeʼe- gas, sun and vehicle bases internal (unaccusative)? D's local check (D1 step 3) and its escape from §5(a) depend on it.A published Navajo unaccusativity diagnostic (e.g. in Hale & Platero, Willie 1991, or Speas 1990) applied to these bases
T87bWhich verbs actually use p. 364's 3i row ʼoʼoo- (Ø/Ł perfective)? A 3i subject on a transitive ʼeʼe- base?A search for Ø/Ł-classifier entries with a perfective in ʼoʼoo- and an example sentence
T87cIn the na- classes, does the listed 3i of naʼajooł (P naʼazhjool) match a 3i row or a row 3 (naazhjool of naajooł)? Do other prefix classes behave like ʼeʼe-?The na'a-((1)) prefix entries (pp. 544, 548, 554, 642) and any paradigm that covers them
T87dUnder D, how does the stranded noun reach clause-initial position (D1 step 9)? Remnant movement of [dP NP t_d], or base position?Sources on Navajo subject position and word order (Speas 1990; Willie 1991); T86b connectivity tests
T87eWhen d has no overt noun (L3 'things', the sun), what values its sort? A pro with a sort feature, or the base alone?Entries where the same 3i base occurs with and without an overt noun (cf. NF set on sun-subject)

T85e answered (LD-9, HYPOTHESIS; data LD-1–LD-8 SOURCED). T85d partly answered (LD-12; same-noun alternation not found in the bases opened). Still open: T85a, b, c, f; T86a–g. New: T87a–e.

Two tables for one string: the ʼeʼe- verbs of gas and unnamed vehicles fit Young & Morgan's 'away' table as 3i person forms just as well as they fit the ʼeʼe- table, so the listed-prefix argument of listed-d is withdrawn. What survives is a split: a named vehicle goes with a plain third person, an unnamed one goes with the 3i and no noun (T87b, T87c in part, T87e in part)

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

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.

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 page listed-d concluded that the 3i ʼa- of verbs like ʼeʼejooł 'drift away (fog, smoke, gas)' is listed in the verb's prefix string and is not the subject's person cell (LD-9). The argument compared the verbs' perfective forms with Y&M's ʼeʼe- paradigm on p. 364. This run opened the paradigm Y&M actually cite for the plain 'away' verbs, ʼa-((1)) on p. 478. Its 3i row is ʼeʼe- (imperfective), ʼiʼíí- (Ø/Ł perfective), ʼoʼoo- (D/L perfective). Those are exactly the forms of the gas and vehicle verbs. So the forms fit two analyses equally well. In one, the ʼa- is listed and the person cell is plain row 3 (p. 364). In the other, the ʼa- is the 3i person cell of an ordinary 'away' verb (p. 478). Y&M cite p. 364 only for the agentive ʼeʼe- verbs with first-person headwords. They describe ʼeʼelyeed as "3i pers. forms corresponding to ʼiishyeed", and they list fifteen other "3i person" forms as headwords of their own. LD-3's "only way to fit both statements" therefore does not hold, and LD-9 is downgraded. The one cell that could separate the two tables, the future, is internally inconsistent in the sources (TT-6). What the new data do show is a split in how the 3i relates to a noun. A named vehicle (a wagon, a barrel) takes the plain third person of the 'away' verb, in 3 of 3 sentences. An unnamed vehicle takes the 3i with no vehicle noun, in 15 of 15 sentences. Gas nouns stay overt with the 3i, in 4 of 4. SOURCED (forms, counts) / HYPOTHESIS (analysis).

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The data · 5 Examples · 6 Answers · 7 Rulesets A and B · 8 Derivations · 9 What this does to D · 10 Fidelity notes · 11 Open questions

Notation is the same as on listed-d. A = the word level (prefixal, left-headed). B = the clause level (right-headed). 3i = Y&M's indefinite ʼa-. Ø = the ordinary third person. AWAY = Pos. Ib ʼa- 'away out of sight'. d = Grant's light determiner (GRANT'S CONJECTURE, assessed here as HYPOTHESIS). "p. 478 row 3i" means the 3i row of the ʼa-((1)) model paradigm, and "p. 364 row 3" means row 3 of the ʼeʼe- model paradigm. No Syntax · Semantics · Interface conventions file is in the repo or the context folder.

1. Question

T87b asked which verbs actually use p. 364's 3i perfective ʼoʼoo-. T87c asked whether other prefix classes behave like ʼeʼe-. Answering T87b meant finding out what the 3i row of the p. 364 table is for. Answering T87c meant opening the neighbouring class, ʼa-((1)), whose plain 'away' verbs (ʼiishááh 'go away out of sight') Y&M cross-reference from several ʼeʼe- verbs. The comparison turned out to bear on listed-d's own answer, so this page checks that answer first. It then asks how the 3i and a noun fit together (Focus 2): when does the 3i come with a noun, and when without one?

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

2. Sources

Opened this run. Young, Robert W. & William Morgan Sr. 1987. The Navajo Language: A Grammar and Colloquial Dictionary, revised edn. Albuquerque: University of New Mexico Press. Read through the University of Colorado search interface (verbs.colorado.edu/~mahu0110/ym/). Entries and paradigm views were saved on the box under [local working copy, not in the repository].

  • Model paradigm ʼa-((1)), p. 478 (paradigm view). Model paradigm ʼeʼe-, p. 364 (paradigm view, re-read). Model paradigm ʼa-((2)), p. 123 (paradigm view).
  • All 65 ʼeʼe entries (search of the previous run, re-read for paradigm citations); entries pp. 362–365, 405.
  • ʼa-((1)) entries: ʼiishááh 480/13, ʼiibąąs 473/2, ʼiiʼeeł 475/1, ʼiimáás 476/18, ʼiinééh 477/4, ʼiizhóósh 487/13, ʼiijooł 474/15, ʼajooł 76/1.
  • Searches: ʼoʼoo (30 and 32 hits), "passive a" (154), "3i pers" (22 entries), the future strings ʼiʼdoo… for each base, and ʼiʼííjool and ʼiʼííbą́ą́z across the whole dictionary.

Nothing else was opened. Speas 1990, Willie 1991 and Hale & Platero are named in open items only.

3. Rules first

TT-1  THE 'a-((1)) TABLE. Y&M p. 478, person rows (singular/duoplural):         SOURCED (p. 478)
      row 3:  IMPF ʼii-  ITER ʼaná-   PFV(Ø/Ł) ʼíí-    PFV(D/L) ʼee-   FUT ʼadoo-  OPT ʼoo-
      row 3i: IMPF ʼeʼe- ITER ʼanáʼá- PFV(Ø/Ł) ʼiʼíí-  PFV(D/L) ʼoʼoo- FUT ʼadoo-  OPT ʼoʼó-
      PASS. A: ʼii-, ʼaná-, ʼee-, —, ʼadoo-, ʼoo-.
TT-2  THE ʼeʼe- TABLE (= LD-1). Y&M p. 364:                                     SOURCED (p. 364)
      row 3:  IMPF ʼeʼe- ITER ʼanáʼá- PFV(Ø/Ł) ʼiʼíí-  PFV(D/L) ʼoʼoo- FUT ʼiʼdoo- OPT ʼoʼó-
      row 3i: IMPF ʼeʼe- ITER ʼanáʼá- PFV(Ø/Ł) ʼoʼoo-  PFV(D/L) ʼoʼoo- FUT ʼiʼdoo- OPT ʼoʼó-
      PASS. A: ʼeʼe-, ʼanáʼá-, ʼoʼoo-, —, ʼiʼdoo-, ʼoʼó-.
TT-3  ONE STRING, TWO ROWS. p. 478 row 3i and p. 364 row 3 are identical in      SOURCED (TT-1, TT-2)
      IMPF, ITER, both PFV columns and OPT. They differ only in FUT
      (ʼadoo- vs ʼiʼdoo-).
TT-4  WHERE Y&M CITE p. 364. Of the 65 ʼeʼe entries, the "(364)" citation      SOURCED (65 entries, Table 1)
      appears in 16. 15 of them have a 1st-person headword (ʼeʼesh-/ʼeʼes-:
      ʼeʼestsééh, ʼeʼesgéés, ʼeʼeshtʼááh …); the 16th is the swarm verb
      ʼeʼesooł (T88e). None of the 3i-headed
      intransitive bases of listed-d Table 1 (ʼeʼejooł, ʼeʼebąąs, ʼeʼemáás,
      ʼeʼezhóósh …) carries a paradigm citation; for them the interface shows
      only "[Guess a paradigm]". LD-3 reached the p. 364 table through the
      interface's paradigm link for ʼeʼejooł, an entry for which Y&M print
      no paradigm citation.
TT-5  Y&M POINT THE 3i BASES AT 'a-((1)). ʼeʼelyeed (363/4): "3i pers. forms      SOURCED (quotes)
      corresponding to 'iishyeed, to run away out of sight". ʼeʼeghááh (362/7):
      "See ʼiishááh", and ʼiishááh (480/13) cites "(478)". ʼeʼemáás (363/5):
      "See ʼiismáás" [sic]. ʼeʼezhóósh (365/9): "Cf. baa ʼiizhóósh", and
      ʼiizhóósh (487/13): "See baa ʼeʼezhóósh". 16 entries carry the formula
      "3i person (forms) corresponding to X" (of 22 that mention "3i pers";
      e.g. chʼíʼílyeed 287/1, 'iigeh 473/17, naʼadá 544/21), so Y&M do make
      headwords out of 3i cells.
TT-6  THE FUTURE CELL DOES NOT SETTLE IT. Two example-sentence futures exist:     SOURCED (forms) / OPEN (the conflict)
      naa ʼiʼdoozhosh (ʼeʼezhóósh, 365/9) and nihíká ʼiʼdoolwoł (ʼeʼelyeed,
      cited at 311/10). Both are ʼiʼdoo-, which matches p. 364, not p. 478 row
      3i ʼadoo-. But Y&M call ʼeʼelyeed the 3i person forms of an 'a-((1))
      base (TT-5). Its attested future therefore conflicts with p. 478's 3i
      future as transcribed. p. 478's 3i ʼadoo- is also the only 3i cell
      without the doubled 'a, so it may be a transcription or print error.
      Not checked against print (T88a).
TT-7  T87b ANSWER. In the Ø/Ł perfective, p. 364's 3i row ʼoʼoo- occurs in       SOURCED (4 sentences) / HYPOTHESIS (reading)
      the example sentences found (searches for ʼoʼoo, 30 and 32 hits) only on
      the transitive base ʼeʼestsééh 'give an
      injection' (363/12): ʼoʼootsi in 29/12, 32/13, 42/9, 130/1. The agent
      is always unexpressed and the English is passive ("I was given an
      injection"). The 3i row and PASS. A are the same in that cell, so
      the form cannot say which. Either way it marks the external argument on
      top of the listed object ʼa- ("the needle", Y&M's "Lit."), and the
      overt-agent sentence uses row 3 (E9).
TT-8  NAMED vs UNNAMED VEHICLE. On the 'a-((1)) twins, the vehicle is an overt   SOURCED (counts) / HYPOTHESIS (generalisation)
      noun and the verb is row 3: tsinaabąąs … ʼííbą́ą́z (473/2), tóshjeeh …
      ʼíímááz (476/18, twice) = 3/3. On the ʼeʼe- vehicle and conveyance bases
      (ʼeʼebąąs, háʼeʼebąąs, ʼeʼemáás, ʼeʼeʼeeł, háʼeʼeʼeeł, ʼeʼeghááh,
      ʼeʼelyeed with bił), 15 of 15 example sentences have no vehicle noun.
      Y&M add "(in a wagon)", "(in a barrel)", "(in the car)" in the
      translation only. The vehicle nouns chidí and tsinaaʼeeł appear only
      inside goal phrases (chidí bighan 'garage', tsinaaʼeeł bá hazʼání
      'harbor').
TT-9  GAS KEEPS ITS NOUN. ʼeʼejooł (362/8): 4 of 4 sentences have the gas        SOURCED (counts; whole-dictionary search)
      noun overt (k'os, ʼáhí, chxííl, bitłʼid). These are the only 4 sentences
      with ʼiʼííjool in the dictionary. But łid 'smoke' also occurs with the
      row-3 form ʼííjool on the 'a-((1)) base ʼajooł 'get into one's eyes'
      (76/1; again in 657/13, 659/13, 796/14).
TT-10 LD-9 DOWNGRADED. For the 3i-headed intransitive ʼeʼe- bases, the forms      HYPOTHESIS (replaces LD-9 for these
      fit (a) listed IV-V ʼa- + p. 364 row 3 and (b) p. 478 row 3i of an       bases; LD-3's inference withdrawn)
      'a-((1)) base. Y&M's words favour (b) for ʼeʼelyeed and point that way
      for ʼeʼeghááh, ʼeʼemáás and ʼeʼezhóósh (TT-5); the one future favours
      (a) or shows a chart error (TT-6). Separate headwords no longer show
      listing, because Y&M also make headwords of 3i cells (TT-5). Either
      way, these verbs have one 3i and one argument: one ʼa-, one d.
TT-11 TWO WAYS A 3i d MEETS A NOUN. (i) Vehicle 3i: the d has no noun beside   HYPOTHESIS (from TT-8, TT-9; cf. PX-11′)
      it, and naming the vehicle goes with the Ø d of the twin base, so the
      3i and the vehicle noun are in complementary distribution, like a
      pronoun and a full DP. (ii) Gas 3i: the d co-occurs with the noun, so it
      takes a DP complement under D. This is PX-11′'s split (gas-subject),
      now with a Ø twin for (i).
TT-12 TWO 3i's IN ONE VERB. On p. 364 bases, the listed IV ʼa- is the object     HYPOTHESIS (TT-7)
      (the needle, the seeds) and the row-3i/PASS. A cell marks the subject,
      so ʼoʼoo- = AWAY + 'a(object) + 3i/PASS(subject). Two arguments, two
      d's, one surface ʼa- string. LD-5 stands; LD-9 is restricted to such
      transitive bases.

4. The data

Table 1. Where the 3i perfective appears. All forms are from Y&M example sentences unless marked "chart".

Base (p./def), classArgument expressed by the 3iPerfective formMatches
ʼeʼestsééh (363/12), ʼeʼe-, cites (364)agent, unexpressed (or passive)ʼoʼootsi (29/12, 32/13, 42/9, 130/1)p. 364 row 3i = PASS. A
ʼeʼestsééh, 1sg agentnone (1sg subject)ʼiʼíítsi (363/12)p. 364 row 1
ʼeʼeshtʼááh (365/6), cites (364)none; overt háíshį́į́ 'someone'ʼiʼííłtʼaʼp. 364 row 3
ʼeʼesooł (363/9), ł, cites (364)none; overt swarms (flies, children, sheep)ʼiʼíísolp. 364 row 3
ʼeʼelyeed (363/4), l, "3i pers. forms" of ʼiishyeedunnamed conveyance; helperʼoʼoolwodp. 478 row 3i (D/L) = p. 364 row 3 (D/L)
ʼeʼebąąs (362/6)unnamed wheeled vehicleʼiʼííbą́ą́zp. 478 row 3i = p. 364 row 3
ʼeʼemáás (363/5)unnamed barrel or tireʼiʼíímááz (printed without the initial ʼ)same
ʼeʼejooł (362/8)gas, with overt nounʼiʼííjoolsame
ʼiibąąs (473/2), 'a-((1))— (vehicle named: tsinaabąąs)ʼííbą́ą́zp. 478 row 3
ʼiimáás (476/18), 'a-((1))— (barrel named: tóshjeeh)ʼíímáázp. 478 row 3
ʼajooł (76/1), 'a-((1))— (smoke named: łid)ʼííjoolp. 478 row 3

Table 2. Ø twins and 3i bases on the same stem with the same AWAY prefix. Y&M cross-references are quoted.

Stem'a-((1)) base (row 3)ʼeʼe- base (3i)Y&M link
*Øbąąsʼiibąąs 473/2 'roll away (hoop, wheeled vehicle)'; bił 'ride in (as a passenger)'ʼeʼebąąs 362/6 'ride away (a passenger in an unspecified wheeled vehicle)'none printed
*Ømáásʼiimáás 476/18 'roll away (conveyed inside a barrel or tire)'ʼeʼemáás 363/5 '… inside an unspecified globular object'"See ʼiismáás"
*Øʼeełʼiiʼeeł 475/1 'float away'; bił "Lit. it floats … with him"ʼeʼeʼeeł 362/14 'float away (an unspecified vehicle)'none printed
*Øzhóóshʼiizhóósh 487/13 'a splinter enters' (tsin, dilk'is overt)ʼeʼezhóósh 365/9 'get a splinter' (no noun)both ways
*lyeedʼiishyeed 'run away out of sight'ʼeʼelyeed 363/4"3i pers. forms corresponding to"
*Øghááhʼiishááh 480/13 (cites 478)ʼeʼeghááh 362/7 'ride away (slow conveyance)'"See ʼiishááh"
*Øjoołʼiijooł 474/15 (wool, hay, wig); ʼajooł 76/1 (smoke into eyes)ʼeʼejooł 362/8 (fog, smoke, cloud, gas, stench)none printed

5. Examples

The Leipzig lines segment the verb only as far as Y&M's tables allow. PFV.3, PFV.3i = the perfective of row 3 or row 3i, with the table page. 'a = the IV-V 3i. Glosses of the other words follow Y&M's translations; the segmentation is ours.

(E1) p. 363 def 5Ndeiiʼnéego dibé bighan góneʼ shił yah ʼiʼíímááz
we.were.playing-SUB sheep 3.POSS-home inside 1SG-with in AWAY-'a-PFV(478.3i / 364.3)-roll.globular
'when we were playing I rolled into the sheep corral (in a barrel).' (Y&M print the verb without its initial ʼ.)
(E2) p. 476 def 18Ndeiiʼnéego dibéghan góneʼ tóshjeeh shił yah ʼíímááz
we.were.playing-SUB sheep.corral inside barrel 1SG-with in AWAY-PFV(478.3)-roll.globular
'as we (pl) were playing I rolled into the sheep corral in a barrel.'
(E3) p. 362 def 6ʼAhbínídą́ą́ʼ shicheii bił ʼiʼííbą́ą́z, kintahjigo
this.morning 1SG.POSS-grandfather 3-with AWAY-'a-PFV(478.3i / 364.3)-roll.hooplike town.toward
'my grandfather rode away toward town this morning (in a wagon).'
(E4) p. 473 def 2Shizhéʼé dáʼákʼeh góneʼ tsinaabąąs bił yah ʼííbą́ą́z
1SG.POSS-father field inside wagon 3-with in AWAY-PFV(478.3)-roll.hooplike
'my father rode into the field in the wagon.' (The interface prints bil for bił here.)
(E5) p. 363 def 4Dáʼákʼeh góneʼ shił yah ʼoʼoolwod dóó dádiʼnítą́
field inside 1SG-with in AWAY-'a-PFV.D/L(478.3i / 364.3)-run.one and 1SG.closed.the.gate
'I rode into the field and closed the gate.'
(E6) p. 362 def 8ʼÁhí yę́ę ńléí dził báátis góʼąą ʼiʼííjool
fog the.former yonder mountain 3.over through AWAY-'a-PFV(478.3i / 364.3)-move.NC
'the fog has drifted away over the mountain.'
(E7) p. 76 def 1 (excerpt)… łid shinákʼe ʼííjoolgo …
smoke 1SG.POSS-eye.into AWAY-PFV(478.3)-move.NC-SUB
'… when smoke got into [my eyes] …'
(E8) p. 130 def 1ʼAzeeʼ bił baa ʼoʼootsi ńtʼééʼ tʼah yówehgo ʼádzaa
medicine 3-with 3-into AWAY-'a-PFV(364.3i or PASS.A)-stick.slender.stiff but still more 3.became
'he was given an injection of medicine but he got worse.'
(E9) p. 363 def 12Tááʼ néílkáahgo ʼazeeʼííłʼíní ʼazeeʼ yił shaa ʼanáʼátsih
three days.pass-SUB doctor medicine 3-with 1SG-into AWAY.ITER-'a-(364.3)-stick.slender.stiff
'the doctor gives me an injection of medicine every three days.'
(E10) p. 365 def 9ʼEii tsin tʼáadoo nihítíhí, naa ʼiʼdoozhosh
those stick don't 2SG.break.up, 2SG-into AWAY-'a-FUT(364)-enter.at.angle
‘don't break up those sticks — you'll get a splinter.’

E1/E2 are close to a minimal pair from Y&M's two entries: same frame, same stem, same AWAY prefix. When the barrel is named (E2) the verb has the row-3 ʼíí-. When it is not (E1) the verb has ʼiʼíí-. E3/E4 show the same for the wheeled vehicle. SOURCED

6. Answers

HYPOTHESIS (data SOURCED). T87b. The Ø/Ł 3i row ʼoʼoo- of p. 364 is used, in Y&M's sentences, by the transitive agentive base ʼeʼestsééh, and only when the agent is not expressed (E8). The cell is shared with PASS. A, so it marks either an indefinite agent or a passive; the form cannot tell which. In both cases it sits on top of the listed object ʼa-. With an overt agent the same base uses row 3 (E9). So p. 364's 3i row belongs to a second argument. It is not the 3i of a one-argument verb (TT-7, TT-12).

HYPOTHESIS Correction to listed-d (T85e). LD-3 argued that the gas and vehicle verbs must have a listed ʼa- because their perfective is p. 364's row 3, not its row 3i. That argument assumed p. 364 was their table. Y&M never assign it to them (TT-4), and the ʼa-((1)) table's 3i row produces the same forms (TT-3). Y&M explicitly call one of these verbs a set of 3i person forms (TT-5). LD-9 is therefore downgraded to TT-10: these verbs may equally be the 3i cell of an ordinary 'away' verb. Nothing in the forms decides between the two, apart from one future cell whose sources conflict (TT-6). The earlier T85e answer stays a HYPOTHESIS only for the transitive p. 364 bases (TT-12). For the intransitive ones it is reopened as T88b.

HYPOTHESIS T87c, in part. The neighbouring class ʼa-((1)) does not behave like ʼeʼe-. It has its own 3i row, and it pairs with the ʼeʼe- intransitives as Ø and 3i twins on the same stem (Table 2). The ʼa-((2)) table (p. 123) has no 3i row at all, because its ʼa- already is the 3i. Its PASS. A perfective ʼoo- differs from its row 3 ʼíí-. The na- classes (naʼajooł) were not reached; that part of T87c stays open.

HYPOTHESIS T87e, in part (Focus 2). For vehicles, naming and the 3i are complementary. The named vehicle takes the Ø d of the ʼa-((1)) twin (E2, E4). The unnamed one takes the 3i d and never has a vehicle noun (E1, E3, E5; 15/15). So the sort of a noun-less 3i d is not supplied by any noun. It comes from the base, whose gloss already says "unspecified wheeled vehicle". Gas is different: its 3i comes with the noun (E6; 4/4). Yet smoke itself can also take the Ø twin (E7). So not every gas noun forces 3i, and T85d's "same noun, both markers" now has a candidate, though across two headwords. OPEN whether łid ever occurs with ʼiʼííjool in a sentence. Y&M's gloss of 362/8 lists smoke, but none of its 4 sentences uses łid.

7. Rulesets A and B

StepE2 (barrel named)E1 (barrel unnamed)E8 (needle listed, agent 3i/passive)
Argument built[dP tóshjeeh d:Ø] Bd:3i, no complement Bd1:'a (listed object), no complement B
Sort on dAgree d ← tóshjeeh Bfrom the base's entry (no noun) Bfrom the base ("the needle") B
Base meets argumentV *Ømáás + dP BV *Ømáás + d BV *Øtsééh + d1 B
Second argumentnone (shił is an oblique) Bnone BVoice: d2:3i, or PASS (no d2) B
Into the wordd → V–v Ad → V–v Ad1, d2 → V–v–Voice A
Person cellp. 478 row 3 Ap. 478 row 3i (or p. 364 row 3 + listed 'a) Ap. 364 row 3i = PASS. A A
Spell-outʼíímááz Aʼiʼíímááz Aʼoʼootsi A
Noun in the clausetóshjeeh left of the verb Bnone B'azee' is bił's object, not the needle B

8. Derivations

All under analysis D (d takes a DP complement as v takes VP; only d joins the
verb word). VPISH: the single argument of the 'away' verbs is assumed to be
merged as V's sister (internal); that is untested (T87a).

D1  … tóshjeeh shił yah ʼíímááz (E2): vehicle named
1  Merge  [NP tóshjeeh] ; [dP NP d] , d = Ø, [sort: _]           B (head-final)
2  Agree  d probes NP: [sort: globular container] valued          B (inside dP)
3  Merge  [VP dP V] , V = *Ømáás, base 'a-((1)) ʼiimáás
          c-selection on dP's label: d:Ø ✓ sort ✓                  B
4  Merge  [shił] (oblique 'with me'), [yah] 'in', goal PP          B
5  Merge  v (Ø classifier)                                        A (word begins)
6  Move   d → V–v (complement head; the NP stays in dP)            B → A
7  Agree  person cell: d:Ø, 3rd → p. 478 row 3                     A
8  Spell  ʼa(Ib) + PFV(Ø/Ł, row 3) ʼíí- + mááz → ʼíímááz           A
9  Spell  [dP tóshjeeh t_d] pronounced at clause level, left of shił yah    B

D2  … shił yah ʼiʼíímááz (E1): vehicle unnamed
1  Merge  d = 3i, no complement (pro-like d)                      B
2  —      no Agree: nothing to value [sort]; the base's entry
          supplies it ("unspecified globular object", 363/5)       B
3  Merge  [VP d V] , V = *Ømáás                                   B
          (a) cell analysis: base ʼiimáás, d's value 3i chosen freely
          (b) listing analysis: base ʼeʼemáás requires d:3i
4  Merge  shił, yah, goal                                         B
5  Merge  v (Ø)                                                   A
6  Move   d → V–v                                                 A
7  Agree  person cell: (a) p. 478 row 3i  (b) p. 364 row 3         A
8  Spell  ʼa(Ib) + 'a(3i) + PFV → ʼiʼíí- + mááz → ʼiʼíímááz       A
          (a) and (b) spell out the same; only FUT could differ (TT-6)
9  —      no noun at clause level                                 B

D3  'Azee' bił baa ʼoʼootsi (E8): listed object, 3i/passive agent
1  Merge  d1 = 'a (listed object, "the needle"), no complement    B
2  Merge  [VP d1 V] , V = *Øtsééh, base ʼeʼestsééh (cites 364)
          listing checked: d1:'a ✓                                 B
3  Merge  [bił] 'with it' (object 'azee'), [baa] 'into him'        B
4  Merge  v (Ø classifier)                                        A
5  Merge  Voice: either d2 = 3i (indefinite agent) or PASS (no d2)  B
6  Move   d1 → V–v ; V–v → Voice                                  A
7  Agree  person cell for the external slot: 3i or PASS. A
          → p. 364 Ø/Ł PFV ʼoʼoo- (identical for both)             A
8  Spell  ʼa(Ib) + 'a(d1) + [3i/PASS] → ʼoʼoo- + tsi → ʼoʼootsi   A
9  Spell  'azee' bił, baa at clause level                          B
   Compare E9: Voice has an overt agent ʼazeeʼííłʼíní with d2 = Ø,
   so step 7 picks row 3 (ITER ʼanáʼá-). The listed d1 is the same.

D4  ʼÁhí yę́ę … ʼiʼííjool (E6): gas, noun overt
1  Merge  [dP [NP ʼáhí yę́ę] d] , d = 3i                            B
2  Agree  d ← NP: [sort: NC]                                      B
3  Merge  [VP dP V] , V = *Øjooł ; (a) base ʼiijooł with the 3i cell,
          which then has to be restricted to NC nouns, since 'iijooł's
          own nouns are wool, hay, wig; or (b) base ʼeʼejooł listing d:3i,NC   B
4  Merge  v (Ø)                                                   A
5  Move   d → V–v ; NP stays                                       B → A
6  Agree  person cell (a) p. 478 row 3i  (b) p. 364 row 3          A
7  Spell  ʼiʼííjool                                               A
8  Spell  [dP ʼáhí yę́ę t_d] at clause level                        B
   Under (a) the 3i cell of ʼiijooł must be sort-restricted (NC only), and
   E7 shows that smoke can also take the Ø cell of a different 'a-((1)) base
   (ʼajooł, into the eyes). Under (b) nothing new is needed.

9. What this does to D

HYPOTHESIS Analysis D survives, but its listed-d support is smaller. listed-d argued that the 3i/Ø choice is listed and is checked locally when the base merges with dP. For the transitive p. 364 bases that still holds: the listed object d is checked at first Merge (D3 step 2), and the 3i row is about the other argument. For the intransitive 'away' verbs, the cell analysis (a) needs no listing at all, except that gas 3i is sort-restricted (D4 step 3). Under D that restriction is still a local check, made on dP's label at first Merge, because the noun and d are one constituent there. Under C, the same restriction would have to reach a noun that is not yet in the derivation. So the locality advantage of D over C carries over to (a) unchanged, whenever the 3i has a noun.

HYPOTHESIS The vehicle 3i adds something for Focus 2. It is a d that never takes a noun, and its Ø twin is a d that does (D1 vs D2). In D's terms, that is the familiar split between a pronoun-like D with no complement and a D with an NP complement. On this view the 3i vehicle d is the pronoun-like one, while the 3i gas d is the full one (TT-11). Ruleset B builds both. The only difference at the A level is whether a noun is left behind at the clause level (D1 step 9 vs D2 step 9).

OPEN All of this assumes the single argument is internal (T87a), which is still untested.

10. Fidelity notes

All forms were copied from interface HTML saved on the box ([local working copy, not in the repository], [local working copy, not in the repository]). Nothing was checked against the print edition. (1) Paradigm view is keyed by the interface's own ids. It gave p. 478 under one id and an unrelated ha-((2)) table (p. 430) under another, and it said "cannot be found" for ʼiishááh, ʼiishyeed and ʼeʼestsééh. The p. 478 table is identified by its printed heading "'a-((1)) (ʼi-)" and the page number in the view; ʼiishááh cites "(478)". (2) p. 478's 3i future ʼadoo- may be a transcription error (TT-6). (3) The F/I/O/P/R lists marked "[Guess a paradigm]" were not used as evidence. All forms argued from come from example sentences or from charts. (4) "3 of 3" and "15 of 15" are counts of Y&M's example sentences, not judgments. Absence of a vehicle noun is evidence from silence. (5) The ʼeʼemáás cross-reference "ʼiismáás" is printed that way; the headword found is ʼiimáás (476/18). (6) No Navajo form here was supplied or checked by a speaker.

11. Open questions

IDQuestionWhat would settle it
T88aIs p. 478's 3i future ʼadoo- right, given that ʼeʼelyeed (which Y&M call the 3i forms of ʼiishyeed) has the future ʼiʼdoolwoł (311/10)?The printed p. 478 chart; further future sentences of the 22 "3i person" headwords
T88b(Reopens T85e for the intransitives.) Are the 3i-headed ʼeʼe- intransitives 3i cells of ʼa-((1)) bases (TT-10a) or listed bases (TT-10b)?Y&M's grammar text on 3i headwords; T88a; any 3i base whose meaning cannot come from its Ø twin
T88cDoes any unnamed-vehicle 3i verb ever co-occur with an overt vehicle noun as subject?A wider sentence search of Y&M for the 3i vehicle forms (all modes)
T88dDoes łid 'smoke' ever occur with ʼiʼííjool (3i), as the 362/8 gloss implies, alongside łid … ʼííjool (E7)?Other sources' example sentences; a fuller Y&M search on the ʼeʼejooł modes
T88eWhat is the IV ʼa- of ʼeʼesooł 'swarm away' (363/9, cites 364), whose overt swarming subjects take row 3? Is it the object of the causative ł ("to cause an air current")?Y&M's stem entry *ł-zooł and the biiʼ ʼassooł 'inflate it' entry

T87b answered (TT-7, HYPOTHESIS; forms SOURCED). T87c answered in part ('a-((1)) and 'a-((2)) compared; na- classes not reached). T87e answered in part (vehicles; gas still open). T85e: the listed-d answer is downgraded for intransitive bases (TT-10) and reopened as T88b. Still open: T87a, T87c (na-), T87d, T87e (gas); T85a, b, c, d, f; T86a–g. New: T88a–e.

Named or unnamed: Young & Morgan's own dictionary pairs the 3i ʼa- with an "unspecified" vehicle and the plain verb with a "named" one, and the vehicle noun never fills the 3i's slot. When the vehicle is named beside a 3i verb, it comes after the verb as the object of bee 'by means of it' (T88c answered, T88d in part)

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

Restated on the unified structure (2026-11-15). The indefinites (ʼa-, ʼadi-, areal ho-) are restated as personless pronouns in ordinary argument positions (H60): ʼadi- as the silent indefinite agent in Spec,vP that raises to Spec,TP by the default (H10b), ʼa- in VP. The 3i data on this page have not yet been individually restated. Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. Flat-S and other pre-VPISH notation below is labelled contrast. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §4, indefinite and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

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.

Two-tables found that a named vehicle goes with the plain third person and an unnamed one goes with the 3i and no noun (TT-8), and left open whether a 3i vehicle verb ever has an overt vehicle noun (T88c). This run searched Young & Morgan (1987) more widely. Y&M state the contrast in their own definitions: bébąąs (151/16) is 'as a passenger in a named wheeled conveyance' with prefix bí-, and bíʼíbąąs (214/11) is 'as a passenger in an unspecified vehicle' with prefix bíʼa-. The flying pair bétʼááh (157/1) and bíʼítʼááh (215/1) is the same. Four of their sentences switch inside one sentence: the unnamed leg has the 3i form and the leg that names the vehicle has a form without it. In 32 example sentences from entries defined with "unspecified" vehicles that contain a vehicle noun, the noun is never a bare argument of the 3i verb. In two of them the vehicle is named in a phrase with bee/yee 'by means of it', after the verb. SOURCED (forms, quotes, counts) / HYPOTHESIS (analysis).

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The 32 sentences · 5 Examples · 6 Answers · 7 Rulesets A and B · 8 Derivations · 9 What this does to D · 10 Fidelity notes · 11 Open questions

Notation is the same as on two-tables. A = the word level (prefixal, left-headed). B = the clause level (right-headed). 3i = Y&M's indefinite ʼa-. Ø = the ordinary third person. d = Grant's light determiner (GRANT'S CONJECTURE, assessed here as HYPOTHESIS), under analysis D (d takes a DP complement as v takes VP; only d joins the verb word). No Syntax · Semantics · Interface conventions file is in the repo or the context folder.

1. Question

Focus 2 asks how the prefixes and the nouns that fill their slots fit together. For vehicles the question is sharp. If the 3i ʼa- is a d, does it ever take the vehicle noun as its complement, the way a gas 3i takes ʼáhí 'fog' (TT-9)? T88c asked exactly that: does an unnamed-vehicle 3i verb ever co-occur with an overt vehicle noun? T88d asked whether łid 'smoke' ever occurs with a 3i perfective.

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

2. Sources

Opened this run. Young, Robert W. & William Morgan Sr. 1987. The Navajo Language: A Grammar and Colloquial Dictionary, revised edn. Albuquerque: University of New Mexico Press. Read through the University of Colorado search interface (verbs.colorado.edu/~mahu0110/ym/). Saved on the box under [local working copy, not in the repository].

  • Whole-dictionary searches of entry text: "unspecified" (191 entries), "named" (27), chidí (1033), łid (72), ʼoʼootʼaʼ (5), bébąąs, bíʼíbąąs, ʼílyeed (6).
  • Entries read in full: bébąąs 151/16, bétʼááh 157/1, bíʼíbąąs 214/11, bíʼítʼááh 215/1, ʼeʼeghááh 362/7, ʼeʼeʼnééh 362/15, ʼiighááh 474/2, ʼiiʼnééh 475/2, ʼílyeed 471/12, chʼíʼílyeed 287/1 and 287/2, yisdáʼílyeed 772/3, niʼnishtʼááh 642/4, ʼashtʼááh 130/6, béldlóósh 153/11, ʼiʼiigeeh 468/4, and the 32 sentences of section 4.

Nothing else was opened.

3. Rules first

NU-1  Y&M DEFINE THE PAIR BY NAMING. Same stem, same 'overtake' meaning:        SOURCED (quotes)
      bébąąs 151/16 "(as a passenger in a named wheeled conveyance)", prefix bí-
      bíʼíbąąs 214/11 "(as a passenger in an unspecified vehicle)", prefix bíʼa-
      bétʼááh 157/1 "(as a passenger in a named flying conveyance)", prefix bé-
      bíʼítʼááh 215/1 "(flying as a passenger in an unspecified vehicle)", prefix bíʼa-
      The "unspecified" member carries 'a- in Y&M's prefix field; the
      "named" member does not.
NU-2  Y&M CALL THESE 'a- FORMS 3i. chʼíʼílyeed 287/1: "3i person forms             SOURCED (quotes)
      corresponding to chʼíníshyeed"; its sense with bił is "(a passenger in
      an unspecified fast moving vehicle)". 'a- itself, 119/7: "3i obj. subj.
      pron., something, someone (unspecified or indefinite)".
NU-3  THE 3i IS THE VEHICLE. Y&M's literal glosses make the vehicle the            SOURCED (quotes) / HYPOTHESIS (reading)
      subject and the person the bił companion: 772/3 "(i.e. my old car ran
      to safety with me)"; 287/2 "Lit. to run down from a height with one".
      So the 3i d is the vehicle argument, not the rider.
NU-4  IN-SENTENCE SWITCHES. In one sentence, the leg with no vehicle noun has    SOURCED (4 sentences)
      the 3i form and the leg that names the vehicle has a form without 'a-:
      157/1a bił ʼoʼootʼaʼ … chidí naatʼaʼí bił bétʼaʼ
      157/1b bił dah ʼadiitʼaʼ … bichidí naatʼaʼí bikééʼ bił dah diitʼaʼ
      471/12 shił ʼílwod … chidí shił yílwod
      772/2  chidí naatʼaʼí shił yóóʼ ʼeetʼaʼ … shił yisdáʼootʼaʼ
      (28/6: ʼadii- = 'a- 3i subj. + di- + yi-.)
NU-5  NEAR-MINIMAL PAIRS ACROSS TWIN ENTRIES. Y&M's 3i and 'a-((1)) entries      SOURCED (sentences)
      on *Øghááh and *d-nééh reuse the same frames; only the twin names
      the vehicle (E1–E6).
NU-6  T88c COUNT. 32 sentences in entries defined with an "unspecified"         SOURCED (count; string search for 11 vehicle nouns)
      vehicle/conveyance/quadruped/boat contain a vehicle noun. In 0 is the
      noun a bare argument of the 3i verb. 24: the noun belongs to another
      clause (it broke down, ran out of fuel, I bought it …). 4: the noun is
      inside a place phrase (chidí bighan 'garage', tsinaaʼeeł bá hazʼání
      'harbor', bichidí bíighahgi 'beside his car'). 2: the noun is the object
      of bee/yee 'by means of it', after the 3i verb (772/3, 642/4). 1 is
      unclear (49/15), and in 1 the verb is not a 3i form (153/11).
NU-7  THE bee-PHRASE. A coreferent vehicle noun can appear with a 3i verb       SOURCED (2 sentences) / HYPOTHESIS (adjunct status)
      only as P's object, to the right of the verb, set off by a comma:
      shił yisdáʼoolwod, shichidí sání bee (772/3); niʼníłtʼaʼ, chidí naatʼaʼí
      ʼáłtsʼíísíyígíí yee (642/4). Read as an adjunct (afterthought), not as
      d's complement.
NU-8  COMPLEMENTARITY. Vehicle 3i d: no DP complement, ever, in the data.       HYPOTHESIS (from NU-4–NU-7; strengthens TT-11(i))
      Naming the vehicle as an argument goes with the Ø d of the twin base.
      Naming it as an adjunct (NU-7) leaves the 3i in place. This is the
      distribution of a pronoun with a doubling adjunct, not of a d with an
      NP complement.
NU-9  SMOKE WITH A 3i-SHAPED PERFECTIVE. ʼiʼiigeeh 468/4 'move away in a      SOURCED (form) / HYPOTHESIS (that 'i'íí- is 3i here)
      billowing cloud (dust, smoke, fog…)', prefix ʼiʼi-: łid shił yah ʼiʼíígoʼ,
      łeezh shił yah ʼiʼíígoʼ. The noun is overt with the doubled ʼiʼíí-
      perfective, as with gas on ʼeʼejooł (TT-9). Still no łid with ʼiʼííjool.

4. The 32 sentences

Table 1. Every sentence, in an entry whose definition says "unspecified" vehicle, conveyance, quadruped or boat (160 such entries), that contains one of the vehicle-noun strings searched, chidí, tsinaabąąs, łį́į́ʼ, télii, tsinaaʼeeł, tóshjeeh, naatʼaʼí, naaʼnaʼí, naabąąs (two further strings searched matched nothing). Classification is ours, from Y&M's translations.

Where the vehicle noun isnEntries (p./def)
Bare argument of the 3i verb0—
Object of bee/yee, right of the 3i verb2772/3, 642/4
Inside a place phrase4287/16, 362/14, 362/15, 640/14
In another clause, with another verb2415/5, 28/17, 50/5, 130/6, 214/13, 277/13, 277/17, 367/13, 401/13, 408/15, 409/1, 471/12, 549/2, 559/16, 559/18, 559/19, 560/5, 638/23, 640/14, 682/3, 700/13, 739/1, 749/22, 772/2
Unclear149/15 (chidí naabąąs … ʼahééʼníłbą́ą́z 'drove it around in a circle'; T89d)
Verb not a 3i form1153/11 (télii bił ʼeeldloozh; Y&M's English adds "(on an unspecified quadruped)"; T89c)

Correction 2026-11-03 ~8:55pm PT: 49/15 is resolved on driver-3i (DV-4): chidí is the subject of naabąąs in the learning clause, so the row "Unclear" should read 0 and "In another clause" 25. The table is kept as written.

Table 2. Twin entries: same stem, Y&M's "unspecified" (3i) and named/plain members.

Stem3i member ("unspecified")Plain memberVehicle noun in the plain member's sentences
*Øbąąsbíʼíbąąs 214/11, bíʼa-bébąąs 151/16, bí-, "named"tsinaabąąs
*d-tʼááhbíʼítʼááh 215/1, bíʼa-bétʼááh 157/1, bé-, "named"chidí naatʼaʼí
*Øghááhʼeʼeghááh 362/7, ʼeʼe-, "slow moving conveyance"ʼiighááh 474/2, ʼa-((1)), "vehicle or conveyance"télii, chidí, łį́į́ʼ
*d-nééhʼeʼeʼnééh 362/15, "unspecified crawling conveyance"ʼiiʼnééh 475/2, ʼa-((1)), "as an army tank, caterpillar tractor"chidí naaʼnaʼí

5. Examples

The Leipzig lines segment the verb only as far as Y&M's prefix fields and tables allow. 'a = the 3i; AWAY = Pos. Ib ʼa-. Other words follow Y&M's translations; the segmentation is ours.

(E1) p. 362 def 7ʼEʼeʼáahgo shizhéʼé chidí bighan góneʼ bił ʼanáʼádááh
evening-SUB 1SG.POSS-father car 3.POSS-home inside 3-with AWAY.ITER-'a-go.one
'every evening my father drives into the garage.'
(E2) p. 474 def 2ʼEʼeʼaahgo shizhéʼé chidí bighan góneʼ chidí bił yah ʼanádááh
evening-SUB 1SG.POSS-father car 3.POSS-home inside car 3-with in AWAY.ITER(3)-go.one
'every evening my father drives his car into the garage.'
(E3) p. 362 def 7Tłʼéédą́ą́ʼ shizhéʼé bił yóóʼ ʼiʼííyáá lá dziłdi
last.night 1SG.POSS-father 3-with astray AWAY-'a-PFV-go.one EVID mountain-at
'last night my father rode off and got lost in the mountains.'
(E4) p. 474 def 2Tłʼéédą́ą́ʼ shizhéʼé łį́į́ʼ bił yóóʼ ʼííyáá lá dziłdi
last.night 1SG.POSS-father horse 3-with astray AWAY-PFV(3)-go.one EVID mountain-at
'last night my father rode off on horseback and got lost in the mountains.'
(E5) p. 362 def 15ʼIʼííʼą́ą́dą́ą́ʼ shizheʼé chidí bighan góneʼ bił yah ʼoʼooʼnaʼ
yesterday.evening 1SG.POSS-father car 3.POSS-home inside 3-with in AWAY-'a-PFV.D/L-crawl
'last evening my father drove into the garage (with a tractor).'
(E6) p. 475 def 2ʼIʼííʼą́ą́dą́ą́ʼ chidí naaʼnaʼí chidí bighan góneʼ shił yah ʼeeʼnaʼ
yesterday.evening tractor car 3.POSS-home inside 1SG-with in AWAY-PFV.D/L(3)-crawl
'I drove the tractor into the garage yesterday evening.'
(E7) p. 157 def 1Naatʼáanii Tó Naneesdizígóó bił ʼoʼootʼaʼ ńtʼééʼ biyeʼ chidí naatʼaʼí bił bétʼaʼ lá Tó Dínéeshzheeʼ bikáaʼdi
superintendent Tuba.City-toward 3-with AWAY-'a-PFV.D/L-fly but 3.POSS-son airplane 3-with 3-overtake.PFV-fly EVID Kayenta 3-over-at
'the Superintendent flew off to Tuba City but his son overtook him over Kayenta.'
(E8) p. 157 def 1Naatʼáanii Hoozdogóó bił dah ʼadiitʼaʼ ʼahbínídą́ą́ʼ, ʼáádóó bikʼíjįʼ biyeʼ tʼáá bí bichidí naatʼaʼí bikééʼ bił dah diitʼaʼ
superintendent Phoenix-toward 3-with up 'a-di-yi-fly this.morning, then after 3.POSS-son just 3 3.POSS-airplane 3-after 3-with up di-yi-fly
'the Superintendent flew off for Phoenix this morning and afterward his son took off after him in his own plane.'
(E9) p. 471 def 12Kǫʼ naʼałbąąsii ninádaaltłiʼígi shił ʼílwod dóó ʼóltaʼjįʼ chidí shił yílwod
train station-at 1SG-with 'a-PFV-run.one and school-to car 1SG-with PFV-run.one
'I arrived at the railroad station and went to the school by car.'
(E10) p. 772 def 3Dziłdi hodiiltłago shił yisdáʼoolwod, shichidí sání bee
mountain-at area.burned-SUB 1SG-with to.safety-'a-PFV-run.one, 1SG.POSS-car old 3-by.means.of
'I escaped the forest fire in my old car (i.e. my old car ran to safety with me).'
(E11) p. 642 def 4Shį́į́dą́ą́ʼ díí hastiin tónteel tséʼnaa niʼníłtʼaʼ, chidí naatʼaʼí ʼáłtsʼíísíyígíí yee
last.summer this man ocean across to.a.point-ʼa-PFV-Ł-fly, airplane small-REL 3-by.means.of
'last summer this man flew across the ocean in a small plane.'
(E12) p. 468 def 4Hastiin Nééz tłʼóoʼdi tsʼiilzéí yidiłidgo hooghan góneʼ łid shił yah ʼiʼíígoʼ
Mr. Long outside trash 3.was.burning.3-SUB hogan inside smoke 1SG-with in AWAY-PFV(ʼiʼíí-)-move.flowing
'when Mr. Long was outside burning trash, the smoke came billowing into the hogan on me.'

E1/E2, E3/E4 and E5/E6 are Y&M's own frames reused across twin entries. The only change is the vehicle noun, and with it the 3i. E7–E9 switch inside one sentence. E10 and E11 are the only cases in Table 1 where the vehicle is named in the same clause as a 3i verb, and both put it in a bee/yee phrase after the verb. SOURCED

6. Answers

HYPOTHESIS (data SOURCED). T88c. No: in Y&M's sentences a 3i vehicle verb never takes the vehicle noun as an argument (0 of 32, NU-6). Y&M's definitions draw the line themselves, "named" vs "unspecified" (NU-1). The vehicle can still be named in the same clause, but only in a postposition phrase with bee 'by means of it', placed after the verb (E10, E11). Y&M's literal gloss of E10, "my old car ran to safety with me", shows that this bee-phrase noun refers to the same vehicle as the 3i (NU-3). So the noun and the 3i can corefer, but the noun never stands in the 3i's slot.

HYPOTHESIS T88d, in part. łid 'smoke' does occur overt with a doubled ʼiʼíí- perfective, on the billowing-cloud verb ʼiʼiigeeh (E12), not on ʼiʼííjool. That fits the gas pattern of TT-9: the gas 3i keeps its noun. Whether ʼiʼíí- there is the 3i cell is not shown by Y&M's entry (prefix field ʼiʼi-, "[Guess a paradigm]"). OPEN for ʼiʼííjool itself.

7. Rulesets A and B

StepE4 (horse named)E3 (unnamed)E10 (unnamed + bee-phrase)E12 (smoke)
Argument built[dP łį́į́ʼ d:Ø] Bd:3i, no complement Bd:3i, no complement B[dP łid d:3i] B
Sort on dAgree d ← łį́į́ʼ Bfrom the base ("slow moving conveyance") Bfrom the base ("unspecified conveyance") BAgree d ← łid B
Base meets argumentV *Øghááh + dP BV *Øghááh + d BV *lyeed + d BV *Øgeeh + dP B
Into the wordd → V–v Ad → V–v Ad → V–v Ad → V–v A
Spell-outʼííyáá Aʼiʼííyáá Ayisdáʼoolwod Aʼiʼíígoʼ A
Noun in the clausełį́į́ʼ left of bił, in dP Bnone B[PP shichidí sání bee], right of the verb, adjunct Błid left of shił, in dP B

8. Derivations

All under analysis D. VPISH: the single argument of these motion verbs is
assumed to be merged as V's sister (internal); still untested (T87a). Pre-VPISH
flat-S notation is not used.

D1  … shizhéʼé łį́į́ʼ bił yóóʼ ʼííyáá (E4): vehicle named
1  Merge  [NP łį́į́'] ; [dP NP d] , d = Ø, [sort: _]                B (head-final)
2  Agree  d probes NP: [sort: quadruped] valued                    B (inside dP)
3  Merge  [VP dP V] , V = *Øghááh, base ʼiighááh ('a-((1)))           B
4  Merge  [bił] (companion: the rider), [yóó'] 'astray'            B
5  Merge  v (Ø classifier)                                        A (word begins)
6  Move   d → V–v ; NP stays in dP                                B → A
7  Agree  person cell: d:Ø → p. 478 row 3                         A
8  Spell  AWAY + PFV(row 3) ʼíí- + yáá → ʼííyáá                    A
9  Spell  [dP łį́į́' t_d] at clause level, left of bił               B

D2  … shizhéʼé bił yóóʼ ʼiʼííyáá (E3): vehicle unnamed
1  Merge  d = 3i, no complement (pronoun-like d)                   B
2  —      no Agree; [sort] comes from the base's entry
          ("slow moving conveyance", 362/7)                        B
3  Merge  [VP d V] , V = *Øghááh                                  B
4  Merge  bił, yóó'                                               B
5  Merge  v (Ø)                                                   A
6  Move   d → V–v                                                 A
7  Agree  person cell: 3i (p. 478 row 3i, or p. 364 row 3 + listed 'a; T88b)  A
8  Spell  AWAY + 'a(3i) + PFV → ʼiʼíí- + yáá → ʼiʼííyáá             A
9  —      no noun at clause level                                 B

D3  shił yisdáʼoolwod, shichidí sání bee (E10): unnamed d + coreferent adjunct
1  Merge  d = 3i, no complement                                   B
2  —      [sort] from the base ("unspecified conveyance", 772/3)   B
3  Merge  [VP d V] , V = *lyeed, prefix yisdá'a-                   B
4  Merge  shił (companion: me)                                    B
5  Merge  v (l classifier)                                        A
6  Move   d → V–v                                                 A
7  Agree  person cell 3i                                          A
8  Spell  yisdá- + 'a(3i) + PFV + l + wod → yisdáʼoolwod             A
9  Merge  [PP [DP shichidí sání] bee] as an adjunct to the clause,
          to the RIGHT of the verb (comma in Y&M)                    B
10 —      bee's 3rd-person object prefix be- is identified by the
          DP inside the PP; the 3i d and the DP corefer by
          interpretation only. No Agree between d and the DP:
          the DP is inside PP, and d has no complement.           B (interpretation)

D4  … hooghan góneʼ łid shił yah ʼiʼíígoʼ (E12): smoke, noun overt
1  Merge  [dP [NP łid] d] , d = 3i (if 'i'íí- is the 3i cell)       B
2  Agree  d ← NP: [sort: billowing mass]                          B
3  Merge  [VP dP V] , V = *Øgeeh, prefix 'i'i-                     B
4  Merge  shił, yah, goal                                         B
5  Merge  v (Ø)                                                   A
6  Move   d → V–v ; NP stays                                       B → A
7  Spell  ʼiʼíígoʼ                                                 A
8  Spell  [dP łid t_d] at clause level                             B

9. What this does to D

HYPOTHESIS The vehicle 3i is now a clean case of a d with no complement. Naming the vehicle as an argument replaces the 3i d with the Ø d of the twin (D1 vs D2), and Y&M build their headwords on that very contrast (NU-1). This is how a pronoun relates to a full DP: they compete for the same slot. Analysis D predicts the contrast without stipulation. D builds the noun inside dP, so a d that takes no complement leaves no place for a noun as an argument. If a noun appears at all, it has to come in as an adjunct, and E10/E11 show exactly that. Under analysis C, where the noun sits outside and values d by Agree, the bee-phrase noun of E10 would be the obvious goal. C then needs an extra statement saying why that noun may not be the 3i's goal but a bare noun may be the Ø d's goal. Under D the reason is structural: the bee-phrase noun is inside a PP, and the 3i d never had a complement.

HYPOTHESIS For Focus 1, the bee-phrase is the one place in these data where a noun phrase sits to the right of the verb. Ruleset B is head-final, so a right-hand phrase is unexpected unless it is an afterthought adjunct outside the core clause. Y&M's comma is consistent with that, but a comma is punctuation, not a prosodic or syntactic test (T89a).

HYPOTHESIS The gas 3i (TT-9, E12) is the other kind of d: 3i with a DP complement. So "3i" alone does not decide whether d takes a complement. The sort does. For vehicles, the 3i is the pronoun-like d. For gases and billowing masses, the 3i takes the noun as its complement. This matches TT-11, which now also rests on the 32 sentences of Table 1 and on E12.

10. Fidelity notes

All forms were copied from interface HTML saved on the box ([local working copy, not in the repository]). Nothing was checked against the print edition or by a speaker. (1) The count in NU-6 is a string search for eleven vehicle-noun strings in 160 entries. A sentence naming a vehicle with another word (e.g. chidítsoh was caught only because it contains chidí) could be missed. (2) "0 of 32" is a count of Y&M's example sentences, not a judgment; absence of the noun as argument is evidence from silence. (3) The classification of Table 1 rests on Y&M's English translations and our reading of clause boundaries. (4) Prefix fields like táʼdi-, niʼna-, háʼdi- in some "unspecified" entries contain a glottal stop after the outer prefix. We have not checked whether this is the 3i ʼa-, so they are not counted for NU-1. (5) In E8, the segmentation of ʼadiitʼaʼ follows Y&M 28/6 (ʼa- 3i subj. + di- + yi-); diitʼaʼ is segmented by analogy without the ʼa-. (6) E9's yílwod, E2's ʼanádááh and E6's ʼeeʼnaʼ are glossed as plain third person following two-tables TT-1 and the entries they appear in; yílwod's own entry was not opened. (7) 153/11's English "(on an unspecified quadruped)" goes with a clause that names télii 'burro' and has no 3i; we take it as a translation slip, not as data (T89c).

11. Open questions

IDQuestionWhat would settle it
T89aIs the post-verbal bee-phrase of E10/E11 an afterthought (right-dislocated) adjunct? Can a bee-phrase naming the vehicle also precede a 3i verb?Y&M sentence search for bee/yee with 3i vehicle forms; a grammar on right dislocation in Navajo
T89bWith the Ø twins, is the named vehicle always a bare noun, or can it also be a bee-phrase (638/23 has chidí bee with nisiikai, a different verb)?Sentences of the 'a-((1)) and bé-/bí- twins with bee
T89cIs 153/11's "(on an unspecified quadruped)" a translation slip, or does Y&M's "unspecified" ever go with a named noun?The printed page; other "unspecified quadruped" entries
T89dDoes ʼahééʼníłbą́ą́z (49/15), with chidí naabąąs as its object, contain the 3i 'a-? If yes, it is a counterexample to NU-6/NU-8 for transitive (driver) verbs.The paradigm of ʼahééʼásbąąs 49/15 and its Ø twin
T89eDo the 'a-((2)) vehicle verbs (ʼílyeed 471/12, ʼashtʼááh 130/6) have Ø twins like 'a-((1)) does, and is yílwod (E9) one?Y&M entry for yílwod/yishyeed; p. 123 paradigm

T88c answered (NU-6–NU-8, HYPOTHESIS; data SOURCED). T88d answered in part (NU-9). Still open: T88a, T88b, T88d (ʼiʼííjool), T88e; T87a, T87c (na-), T87d, T87e (gas); T85a, b, c, d, f; T86a–g. New: T89a–e.

Driver verbs: on the transitive "drive/fly/sail it around in a circle" verbs, the 3i ʼa- fills the object slot and stands for the vehicle, and the car in 49/15 belongs to another clause (T89d answered, T89b in part)

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

Verbs on this page → dictionary: ní ‘say’ · sidá ‘sits; with yii’/bii’ ‘in’: sits in’

Restated on the unified structure (2026-11-15). The indefinites (ʼa-, ʼadi-, areal ho-) are restated as personless pronouns in ordinary argument positions (H60): ʼadi- as the silent indefinite agent in Spec,vP that raises to Spec,TP by the default (H10b), ʼa- in VP. The 3i data on this page have not yet been individually restated. Light d, used below, is withdrawn (Grant, 2026-11-14); light n is an OPEN structural question. Flat-S and other pre-VPISH notation below is labelled contrast. The “roles not words” theory, where it appears below, is dropped. Status on the spine: see unified structure, §4, indefinite and passive rows. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

2026-11-11 audit pass 3. A derivation step annotated “(VPISH)” now reads “(VPISH hyp.)”: VPISH names the hypothesis that the subject starts in Spec,vP; it is not a node. 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.

Named-unnamed found that a vehicle noun is never a bare argument of a 3i verb (NU-6, NU-8). That work was done on passenger (intransitive) verbs. It left one possible counterexample among the driver (transitive, ł-classifier) verbs: ʼahééʼníłbą́ą́z 'I drove it around in a circle' (Y&M 49/15), in a sentence that begins chidí naabąąs (T89d). This run read the whole "around in a circle" family in Young & Morgan (1987) together with two of its paradigm tables. Two findings. (i) 49/15 does contain the 3i. Its prefix is ʼahééʼa-, which Y&M gloss as "Pos. IV, 3i obj. pron., something", and the p. 50 table carries the ʼa- in every person row. (ii) chidí is still not its argument. Chidí naabąąs 'the car rolls around' is the object clause of bíhooshʼaah 'I learn it', a frame Y&M reuse word for word under a plain driver verb (45/8), another 3i driver verb (50/5) and dah diishłeeh (346/13). Across the family, the 3i driver verbs have 11 sentences, and none has a vehicle noun as an argument. In 5 of the 6 plain-twin sentences the vehicle is an overt object. The two tables show the 3i object ʼa- and the third-person object yi- in the same place, never together. SOURCED (forms, quotes, tables, counts) / HYPOTHESIS (clause analysis and everything under D).

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The circle family · 5 Examples · 6 Answers · 7 Rulesets A and B · 8 Derivations · 9 What this does to D · 10 Fidelity notes · 11 Open questions

Notation is the same as on named-unnamed. A = the word level (prefixal, left-headed). B = the clause level (right-headed). 3i = Y&M's indefinite ʼa-. Ø = the ordinary third person. 3o = Y&M's row for a third-person subject acting on a third-person object. CIRCLE = ʼahéé-/ʼahéná- 'around in a circle' (Y&M 45/7: ʼah-é Pos. O-Ib + ná- Pos. Ib). Ł = the ł classifier, which Y&M gloss "to cause to move …". d = Grant's light determiner (GRANT'S CONJECTURE, assessed here as HYPOTHESIS), under analysis D: d takes a DP complement as v takes VP, and only d joins the verb word. No Syntax · Semantics · Interface conventions file is in the repo or the context folder.

1. Question

Focus 2 asks how the prefixes fit with the nouns that fill their slots. NU-8 says the vehicle 3i d never takes a DP complement. Every sentence behind NU-8 has an intransitive verb, where the vehicle 3i is the subject and the rider is the object of bił. Driver verbs are causatives: the driver is the subject and the vehicle is what moves. That raises two questions. Does the vehicle 3i keep its role (the moving thing) and switch to object position? And does it ever take the car as its noun? T89d raised the one sentence that might show it does: Chidí naabąąs bíhooshʼaahgo … ʼahééʼníłbą́ą́z (49/15). T89b asked whether a plain (Ø) verb's named vehicle can also appear as a bee-phrase.

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

2. Sources

Opened this run. Young, Robert W. & William Morgan Sr. 1987. The Navajo Language: A Grammar and Colloquial Dictionary, revised edn. Albuquerque: University of New Mexico Press. Read through the University of Colorado search interface (verbs.colorado.edu/~mahu0110/ym/). Saved on the box under [local working copy, not in the repository].

  • Word searches: ahéé (113 entries), ahéná (30), naabąąs (240), hooshʼaah (34), shooʼíní, tʼáadoo. Definition search: "3i" (385).
  • Prefix entries: ʼa- 116/5, ʼahéé- 45/7, ʼahééʼa- 49/14, ʼahééʼii- 50/3, ʼahééʼni- 51/5, ʼahénáʼa- 44/17, adii- 28/6.
  • Verb entries read in full: 44/18, 44/19, 45/1, 45/8, 45/12, 46/1, 46/2, 46/3, 48/5, 48/10, 49/15, 49/16, 49/17, 50/1, 50/2, 50/5, 51/1, 51/2, 51/4, 51/9, 53/1, 53/18; also bíhooshʼaah 196/16, naabąąs 573/6, diishłeeh 346/13, and 588/6 and 183/9 for the chidí naa- 'drive a car' frame.
  • Paradigm tables (interface "paradigm view"): p. 50, headed ʼahééʼa- (the one linked from 49/15), and p. 46, headed ʼahéé-((1)), ʼahéná- (the one linked from 45/8).

Nothing else was opened.

3. Rules first

DV-1  Y&M LABEL THE DRIVER 'a- AS THE OBJECT. ʼahééʼa- 49/14: "'ahéé-,          SOURCED (quotes)
      around in a circle + 'a-, '(a)-, Pos. IV, 3i obj. pron., something".
      ʼa- 116/5: "Pos. IV, 3i direct obj. pron. … and Pos. V, 3i subj. pron."
      ʼahééʼii- 50/3: ʼahéé- + 'a- Pos. IV 3i obj. pron. + hi-/yi- Pos. VI.
      ʼahénáʼa- 44/17: "'a-, Pos. IV-V, 3i obj.-subj. pron."
DV-2  Y&M'S LITERAL GLOSSES MAKE THE 3i THE MOVING THING. 44/19 bił ʼahénáʼáshtʼááh:  SOURCED (quotes)
      "(Lit. to cause something to fly around in a circle with him)".
      51/9 bił ʼahééʼníshʼeeł: "(Lit. to cause something to move floating
      around in a circle with him.)" Definitions: 49/15 "(in an unspecified
      wheeled vehicle.)", 49/17 "(piloting a flying vehicle)", 51/1 "(in an
      unspecified floating vehicle)", 51/2 "(piloting an unspecified flying
      vehicle)".
DV-3  TWO TABLES, ONE OBJECT SLOT. p. 46 (ʼahéé-((1)), ʼahéná-) has a 3o row     SOURCED (tables) / HYPOTHESIS (slot reading)
      (perfective Ø-Ł ʼahééyiní-) beside row 3 (ʼahéná-). p. 50 (ʼahééʼa-)
      has ʼ in every row (1 ʼahééʼní-, 3 ʼahééʼní-, 3a ʼahéézhʼní-) and
      no 3o row. So after ʼahéé- we find either yi- (3o) or ʼ (3i
      object), never both. We read them as two values of one object
      slot.
DV-4  T89d. 49/15 ʼahééʼníłbą́ą́z = CIRCLE-3i-PFV-Ł-roll, 1sg subject (Y&M        SOURCED (forms) / HYPOTHESIS (clause boundary)
      P row 1 of p. 50; its Ø twin 45/8 has P ʼahééníłbą́ą́z). chidí is the
      subject of naabąąs 'it rolls around (a hooplike object or a wheeled
      vehicle)' (573/6), a plain-3rd intransitive. [chidí naabąąs] is the
      object of bíhooshʼaah 'to learn it' (196/16), subordinated by -go.
      So chidí is not an argument of ʼahééʼníłbą́ą́z. 49/15 is NOT a
      counterexample to NU-8; it moves from "unclear" to "another clause"
      in named-unnamed's Table 1 (24 → 25).
DV-5  THE LEARNING FRAME IS REUSED. Chidí naabąąs bíhooshʼaahgo 'when I was        SOURCED (4 entries)
      learning to drive (a car)' opens 45/8 (Ø driver), 49/15 (3i driver),
      50/5 (3i driver) and 346/13 (dah diishłeeh, Ø). The frame is the
      same before a Ø driver verb and before a 3i one.
DV-6  COUNTS (circle family, ł/Ø-classifier driver verbs, all sentences):       SOURCED (counts; our clause reading)
      3i drivers  (44/19, 49/15, 49/16, 49/17, 50/5, 51/1, 51/2, 51/9):
                  11 sentences; vehicle noun as argument 0; chidí in the
                  learning clause 2; vehicle named only in Y&M's English 1 (51/1).
      Ø drivers   (45/8, 45/12, 46/1, 46/2, 48/5, 53/18):
                  6 sentences; overt vehicle object 5; no overt object 1
                  (45/8, after the learning clause).
DV-7  MINIMAL PAIRS on one stem, ± 'a-:                                          SOURCED
      *ł'eeł  45/12 Ø  Kéyah táłkáaʼdi dah siʼánígíí binaago si tsinaaʼeel ʼahééníłʼéél
              51/9 3i  Kéyah táłkáaʼdi dah siʼánígíí binaago ʼahééʼníłʼéél
      *łbąąs  45/8 Ø   Chidí naabąąs bíhooshʼaahgo … ʼahééníłbą́ą́z
              49/15 3i Chidí naabąąs bíhooshʼaahgo … ʼahééʼníłbą́ą́z
      *łt'ááh 46/2 Ø   … bichidí naatʼaʼí dził binaago ʼahééiníłtʼaʼ
              44/19 3i … dził binaago ʼahééʼníłtʼaʼ
      In the first pair the only difference is the noun and the ʼ.
DV-8  THEME CONSTANCY. The vehicle 3i is the thing that moves in both verb     HYPOTHESIS (from DV-1, DV-2, DV-6; NU-3)
      types: the subject of a passenger intransitive with bił (50/2, 44/18;
      named-unnamed) and the object of a ł-causative driver verb (DV-2).
      Under VPISH + D it is one d, merged in one place, as V's internal
      argument. The ł voice adds the driver as an external argument. The
      object/subject difference comes from v, not from the d.
DV-9  NULL Ø vs 3i. A Ø driver verb can also lack an overt object (45/8,    SOURCED (45/8) / HYPOTHESIS (pro vs 3i) / OPEN (T90a)
      where it refers back to the car of the learning clause). So "no
      noun" does not by itself pick the 3i. Proposal: a Ø d with no
      complement is a definite pro; the 3i d is indefinite ("something",
      116/5). Y&M's English does not settle it: 45/8 "drove the car",
      49/15 "drove it".
DV-10 T89b IN PART. A named vehicle in a bee-phrase BEFORE a Ø verb:              SOURCED (1 sentence) / HYPOTHESIS (bearing on T89a)
      łį́į́ʼ bee ʼahéénísdzá (53/1, *d-ghááh 'walk, go'), "I made a circuit on
      horseback". So the bee-phrase is not tied to the 3i or to the
      post-verbal position. No bee-phrase occurs with a 3i verb in this
      family (T89a's second half still unanswered).

4. The circle family

Table 1. Every vehicle verb in Y&M's "around in a circle" family (prefixes ʼahéé-, ʼahéná-, ʼahééhi-, ʼahééʼa-, ʼahénáʼa-, ʼahééʼii-) found in the ahéé and ahéná searches, sorted by verb type. Classification is ours, from Y&M's prefix fields, classifiers and translations.

TypeEntries (p./def, prefix, classifier)Vehicle sentencesVehicle noun as argument
3i driver (transitive)44/19 ʼahénáʼa- ł; 49/15, 49/16, 49/17 ʼahééʼa- ł; 51/9 ʼahénáʼa- ł; 50/5, 51/1, 51/2 ʼahééʼii- ł110
Ø driver (transitive)45/8, 45/12, 46/2, 53/18 ʼahéé- ł; 46/1 ʼahéé- Ø (*Øleeh 'act with a cord'); 48/5 ʼahééhi- ł65 (object); 1 null object (45/8)
3i passenger (intransitive, bił)44/18 ʼahénáʼa-; 50/2 ʼahééʼa- d; 51/4 ʼahééʼii- d30 (51/4 has chidí bitooʼ 'gasoline' in another clause)
Ø passenger (intransitive, bił)45/1 ʼahéná- Ø; 46/3 ʼahéná- d; 48/10 ʼahééhi- d33 (subject: bitsinaaʼeeł, chidí naatʼaʼí ×2)

Altogether: none of the 14 3i sentences has a vehicle noun as an argument, and 8 of the 9 Ø sentences do. Y&M also cross-refer the 3i passenger verb to its Ø twin themselves: 50/2 ʼahééʼítʼááh "(Cf. ʼahénátʼááh.)" points to 46/3.

Table 2. The perfective cells of the two tables (Ø-Ł column; Y&M pp. 46 and 50, interface paradigm view).

Rowp. 46 ʼahéé-((1)), ʼahéná- (Ø object)p. 50 ʼahééʼa- (3i object)
1ʼahééní-ʼahééʼní-
2ʼahééyíní-ʼahééʼííní-
3ʼahéná-ʼahééʼní-
3oʼahééyiní-— (no row)
3aʼahéézhní-ʼahéézhʼní-
3iʼahééʼí-ʼahééʼí- (row offset; see fidelity note 4)

5. Examples

The Leipzig lines segment the verb only as far as Y&M's prefix fields and tables allow. 'a = the 3i object (Pos. IV, DV-1). PFV.1SG etc. name the table cell, not separate morphs. Other words follow Y&M's translations, and the segmentation is ours. tʼáadoo shooʼíní is glossed as a unit, 'without (him) seeing me', from Y&M's English only.

(F1) p. 49 def 15Chidí naabąąs bíhooshʼaahgo shizhéʼé tʼáadoo shooʼíní ńléí halgaidi ʼahééʼníłbą́ą́z
car 3.roll.about 3.O-learn.IPFV.1SG-SUB 1SG.POSS-father without.seeing.me yonder flat-at CIRCLE-'a-PFV.1SG-Ł-roll
'when I was learning to drive, I took the car out on the flat without my father's seeing me and drove it around in a circle.'
(F2) p. 45 def 8Chidí naabąąs bíhooshʼaahgo naakidi dibé bighan binaago ʼahééníłbą́ą́z
car 3.roll.about 3.O-learn.IPFV.1SG-SUB twice sheep 3.POSS-home 3-around CIRCLE-PFV.1SG-Ł-roll
'when I was learning to drive I drove the car twice around the sheep corral.'
(F3) p. 50 def 5Chidí naabąąs bíhooshʼaahgo halgaidi ʼahééʼiisbąąsgo náʼahóónáád
car 3.roll.about 3.O-learn.IPFV.1SG-SUB flat-at CIRCLE-'a-SER-IPFV.1SG.Ł-roll-SUB for.a.long.time
'when I was learning to drive a car I drove around in circles for a long time on the flat.'
(F4) p. 45 def 12Kéyah táłkáaʼdi dah siʼánígíí binaago si tsinaaʼeel ʼahééníłʼéél
[island] 3-around 1SG.POSS-boat CIRCLE-PFV.1SG-Ł-float
'I sailed (rowed, piloted) my boat around the island.'
(F5) p. 51 def 9Kéyah táłkáaʼdi dah siʼánígíí binaago ʼahééʼníłʼéél
[island] 3-around CIRCLE-'a-PFV.1SG-Ł-float
'I sailed (rowed, took a boat) around the island.'
(F6) p. 46 def 2Naalyéhé yá sidáhí bichidí naatʼaʼí dził binaago ʼahééiníłtʼaʼ
trader 3.POSS-airplane mountain 3-around CIRCLE-3O.PFV-Ł-fly
'the trader flew his plane around the mountain.'
(F7) p. 44 def 19ʼAdą́ą́dąąʼ naalyéhé yá sidáhí dzíł binaago shił ʼahééʼníłtʼaʼ
yesterday trader mountain 3-around 1SG-with CIRCLE-'a-PFV.3-Ł-fly
'yesterday the trader flew me around the mountain.' (Y&M: "Lit. to cause something to fly around in a circle with him")
(F8) p. 53 def 18Naalyéhé yá sidáhí dził Naatsisʼáán wolyéhígíí chidí naatʼaʼí yinaago ʼahééyiníłtʼaʼ
trader mountain Navajo.Mountain 3.is.called-REL airplane 3-around CIRCLE-3O.PFV-Ł-fly
'the trader flew the airplane around Navajo Mountain.'
(F9) p. 49 def 16Magellan wolyéhę́ę díí nahasdzáán nihił dah siʼánígíí tsʼídá ʼáłtsé yikááʼ ʼahééʼníłʼéél jiní
Magellan 3.was.called-REL this earth [that.we.live.on] very first 3-on CIRCLE-'a-PFV.3-Ł-float they.say
'Magellan was reportedly the first person to sail around the world.'
(F10) p. 45 def 1Magellan wolyéé ńtʼééʼ bitsinaaʼeeł nahasdzáán yinaago bił ʼahééníʼéél
Magellan 3.was.called PAST 3.POSS-ship earth 3-around 3-with CIRCLE-PFV.3-float
'Magellan sailed around the world in his ship.' (Y&M: "Lit. it floats around in a circle with him")
(F11) p. 53 def 1ʼAhbínídą́ą́ʼ ńléí dziłtsį́įdi łį́į́ʼ bee ʼahéénísdzá, béégashii hanishtáago
this.morning yonder mountain.foot-at horse 3-by.means.of CIRCLE-PFV.1SG-go.one, cow 1SG.look.for-SUB
'this morning I made a circuit on horseback to the foot of the mountain looking for a cow.'

F9 and F10 describe the same event in two ways. With the 3i driver verb, Magellan is the subject and the ship is the ʼ (F9). With the Ø passenger verb, the named ship bitsinaaʼeeł is the subject and Magellan is the object of bił (F10). SOURCED (sentences) / HYPOTHESIS (the role reading).

6. Answers

T89d: answered. SOURCED (forms, labels, tables) / HYPOTHESIS (clause boundary). ʼahééʼníłbą́ą́z does contain the 3i: its prefix field is ʼahééʼa-, and DV-1 and DV-3 locate the ʼ in the object slot. But chidí is not its argument. chidí is the subject of naabąąs inside the learning clause (DV-4). The proof is that the same clause, word for word, also precedes the Ø driver verb of 45/8 and the 3i driver verb of 50/5 (DV-5). NU-8 survives, and its range grows from passenger verbs to driver verbs (DV-6, DV-8). 49/15 also becomes a fifth case of a switch inside one sentence. The named car is the subject of a Ø verb (naabąąs), and the driving that leaves the vehicle unnamed takes the 3i.

T89b: answered in part. SOURCED (one sentence). A named vehicle can be a bee-phrase with a Ø verb, and it can come before the verb (F11). F11's verb is 'walk, go', not a vehicle verb. No Ø vehicle verb with a bee-phrase was found.

New generalisation HYPOTHESIS: the vehicle 3i is always the theme. It is the subject of an intransitive and the object of a causative (DV-8). Y&M's position labels do not split the two cases neatly. They give "Pos. IV, 3i obj." for ʼahééʼa- on the driver verbs (49/15–17) and also on the intransitives 50/1 and 50/2. Yet adii- 28/6 is "Pos. V 3i subj.". Whether the intransitive vehicle 3i sits in the object position (unaccusative) is T90b.

7. Rulesets A and B

StepF4 (Ø, boat named)F5 (3i, boat unnamed)F8 (Ø, 3o)F7 (3i, 3rd subject)F2 (Ø, null object)
Object built[dP sitsinaaʼeeł d:Ø] Bd:3i, no complement B[dP chidí naatʼaʼí d:Ø] Bd:3i, no complement Bd:Ø, no complement (pro) B
Base meets objectV √ʼeeł + dP BV √ʼeeł + d BV √tʼááh + dP BV √tʼááh + d BV √bąąs + d B
Driver addedv:Ł + 1SG Bv:Ł + 1SG Bv:Ł + trader (3) Bv:Ł + trader (3) Bv:Ł + 1SG B
Into the wordd → V–v Ad → V–v Ad → V–v Ad → V–v Ad → V–v A
Cellp. 46 row 1 Ap. 50 row 1 Ap. 46 row 3o Ap. 50 row 3 Ap. 46 row 1 A
Spell-outʼahééníłʼéél Aʼahééʼníłʼéél Aʼahééyiníłtʼaʼ Aʼahééʼníłtʼaʼ Aʼahééníłbą́ą́z A
Noun in the clauseboat, left of the verb, in dP Bnone Bplane, left of yinaago, in dP Bnone Bnone; car in the learning clause B

Focus 1 reading HYPOTHESIS: the choice between Ø and 3i, and whether d has a complement, is made at B, in the clause syntax. Where the object surfaces, as yi- or ʼ after ʼahéé- and left of the subject cell, is fixed at A, inside the word. The 3o yi- appears only when the subject is also third person (row 3o), so it is a person-combination cell of ruleset A. It is not a separate d.

8. Derivations

All under analysis D. VPISH: the driver is merged in Spec,vP (external); the
vehicle d is V's sister (internal). The ł classifier is read as the causative
voice/v head (Y&M's "to cause to move …" for *ł-stems). Pre-VPISH flat-S
notation is not used.

D1  Kéyah táłkáaʼdi dah siʼánígíí binaago ʼahééʼníłʼéél (F5): 3i driver
1  Merge  d = 3i, no complement                                   B
2  —      [sort] from the base: "a boat" (51/9 definition)         B
3  Merge  [VP d V] , V = √'eeł 'move floating', prefix ʼahénáʼa-        B
4  Merge  [PP [island] binaago] 'around the island', adjunct       B (head-final P)
5  Merge  v = Ł (causative); [vP DP_1SG [v' VP v]]                  B (VPISH hyp.)
6  Move   V → v ; d (object) → V–v                                B → A (word begins)
7  Agree  subject cell ← 1SG; object slot ← 3i                     A
8  Spell  CIRCLE ʼahéé- + ʼ (3i, Pos. IV) + PFV row 1 ní- + ł + ʼéél
          → ʼahééʼníłʼéél (p. 50 row 1, Ø-Ł perfective)                   A
9  —      no noun at clause level; 1SG pro                         B

D2  … binaago si tsinaaʼeel ʼahééníłʼéél (F4): Ø driver, boat named
1  Merge  [NP tsinaa'eeł] with possessor 1SG ; [dP NP d] , d = Ø  B
2  Agree  d ← NP: [sort: floating vehicle]                         B (inside dP)
3  Merge  [VP dP V] , V = √'eeł, prefix ʼahéé-                     B
4  Merge  PP adjunct 'around the island'                           B
5  Merge  v = Ł ; DP_1SG in Spec,vP                                B
6  Move   V → v ; d → V–v ; NP stays in dP                         B → A
7  Agree  subject cell ← 1SG; object slot ← Ø (3), silent with a 1SG subject   A
8  Spell  ʼahéé- + PFV row 1 ní- + ł + ʼéél → ʼahééníłʼéél (p. 46 row 1)       A
9  Spell  [dP si tsinaa'eel t_d] left of the verb                  B

D3  … chidí naatʼaʼí yinaago ʼahééyiníłtʼaʼ (F8) vs … shił ʼahééʼníłtʼaʼ (F7)
1  Merge  F8: [dP chidí naat'a'í d:Ø] ; F7: d:3i, no complement   B
2  Merge  [VP d(P) V] , V = √t'ááh 'move flying'                   B
3  Merge  F7 only: [PP shi- -ił] 'with me' (the rider)             B
4  Merge  v = Ł ; subject DP 'the trader' (3) in Spec,vP          B
5  Move   V → v ; object d → V–v                                  A
6  Agree  F8: subject 3 + object Ø(3) → row 3o ; F7: subject 3 + object 3i → row 3 of p. 50   A
7  Spell  F8 ʼahéé- + yi- (3o) + ní- + ł + t'a' → ʼahééyiníłtʼaʼ               A
          F7 ʼahéé- + ʼ (3i)  + ní- + ł + t'a' → ʼahééʼníłtʼaʼ                A
8  —      the 3i takes the place of the 3o yi-: one object slot (DV-3)  A

D4  Chidí naabąąs bíhooshʼaahgo … ʼahééʼníłbą́ą́z (F1): two clauses
    Embedded clause
1  Merge  [dP chidí d:Ø] ; [VP dP V] , V = √bąąs 'roll', Ø classifier   B
2  Move   d → V–v ; Agree row 3 → naabąąs (573/6)                  A
3  Spell  [chidí naabąąs] , chidí left of its verb                  B
    Learning verb
4  Merge  the clause as complement of the object d of bíhooshʼaah
          (bí- 3 object 'it', 196/16); v = Ł ; 1SG subject         B
5  Spell  bíhooshʼaah ; Merge C = -go 'when'                        A, then B
    Matrix
6  Merge  d = 3i, no complement ; V = √bąąs ; v = Ł ; 1SG subject   B
7  Merge  adjuncts shizhéʼé tʼáadoo shooʼíní, ńléí halgaidi, and the -go clause   B
8  Move   d → V–v ; Agree p. 50 row 1                             A
9  Spell  ʼahééʼníłbą́ą́z                                             A
10 —      chidí's d is inside the -go clause; the matrix 3i d has no
          complement. Their link (it was the car I drove) is
          interpretation only, as with the bee-adjunct of NU-7.    B (interpretation)

D5  Chidí naabąąs bíhooshʼaahgo naakidi dibé bighan binaago ʼahééníłbą́ą́z (F2)
1–5 as D4                                                          B/A
6  Merge  d = Ø, no complement (pro, anaphoric to the car)         B   HYPOTHESIS (DV-9)
7  Move, Agree p. 46 row 1 ; Spell ʼahééníłbą́ą́z                      A
    F1 and F2 differ only in the matrix d: 3i (indefinite) vs Ø pro (definite?). T90a.

9. What this does to D

HYPOTHESIS Analysis D handles the driver verbs with no new machinery. The vehicle d is merged in the same internal position whether the verb is a passenger intransitive or a driver causative. The ł voice only adds a driver above it, so it does not matter which grammatical function the vehicle 3i ends up with (DV-8). This is the VPISH expectation. The theme is the same argument in both, and only the presence of an external argument changes whether it surfaces as subject or object. The vehicle 3i still never takes a complement (DV-6), now across 14 sentences in both verb types. Tables p. 46 and p. 50 add a new point: in the word, the 3i object competes with the 3o yi- for one slot (DV-3). This is what D predicts if Ø and 3i are two values of the same d. The difference is that Ø can take a noun complement and the 3i cannot. Analysis C (an outside noun values d by Agree) still has to say why the car of F1's learning clause cannot be the 3i's goal. D gets this for free because that noun is inside another clause. The finding that does not follow from D alone is DV-9. A Ø d can also go without a noun (F2), so the 3i is not simply "the d without a noun". If DV-9 is right, the 3i is an indefinite pronoun-d and the silent Ø is a definite one. This parallels a full DP against pro.

Correction to named-unnamed Table 1: 49/15 belongs in "another clause" (now 25), not "unclear" (now 0). A note has been added there; the original table is kept as written.

10. Fidelity notes

All forms were copied from interface HTML saved on the box ([local working copy, not in the repository]). Nothing was checked against the print edition or by a speaker. (1) 45/12's object is printed si tsinaaʼeel (with a space and final l), and its P form is ahééníłʼéél without the initial glottal stop. Elsewhere Y&M write sitsinaaʼeeł (45/11) and the sentence has ʼahééníłʼéél. We copy the sentence as given and suspect conversion errors. (2) 46/2's sentence has ʼahééiníłtʼaʼ, while the p. 46 3o cell and 53/18 have ʼahééyiní-. We take this as a dropped y and gloss it as 3o. (3) The p. 46 3o row has five cells for seven columns, so its alignment is uncertain. Only the perfective ʼahééyiní- is used here, and it is confirmed by 53/18. (4) The p. 50 3i row also has five cells, so which column ʼahééʼí- belongs to is unclear (T90c). (5) 45/1's prefix field reads ʼahénáʼa-, but its parenthesis reads (ʼahéná-). We count it as Ø because its stem is *Øʼeeł, whose own ʼ may be what the prefix field caught (T90d). (6) 44/18's classifier field says ł, but its stem is *ldlóósh and P is ʼahééʼíldloozh. We count it as an intransitive. 49/17's stem note prints *ltʼááh while its classifier field says ł; we follow the field and the gloss "to cause to move". (7) Table 1 is limited to the ahéé/ahéná word searches, so a vehicle verb in this family spelled another way could be missed. (8) In 44/19, paradigm (45) was not opened. Its 3rd-person form is matched to p. 50 row 3, which has the same shape. (9) The learning clause is analysed from 196/16 (bíhooshʼaah 'to learn it', 3rd yíhooł-) and 573/6. That chidí naabąąs is its object is our reading of Y&M's English "learning to drive".

11. Open questions

IDQuestionWhat would settle it
T90aDoes a noun-less Ø driver verb (45/8) differ in reading from its 3i twin (49/15), definite pro vs indefinite 'something' (DV-9)? Does the Ø null object always need an antecedent in the discourse?More Ø vehicle-verb sentences with no overt object; grammar statements on 3rd-person pro vs 3i
T90bY&M label ʼahééʼa- "Pos. IV, 3i obj." even on intransitives (50/1 'something moves … the clock hand', 50/2 passenger). Is the intransitive vehicle 3i in object position, i.e. unaccusative (cf. T87a), against adii- 28/6 "Pos. V 3i subj."?Y&M grammar on Pos. IV vs V for 3i; other intransitive prefixes labelled IV
T90cWhat is the p. 50 table's 3i row (ʼahééʼá-, ʼahééʼí-, …): a 3i subject together with the 3i object, or a passive-like cell?Print p. 50; Y&M's paradigm introduction
T90d45/1 ʼahénáʼeeł: is the prefix ʼahénáʼa- (field) or ʼahéná- (parenthesis)?Print p. 45
T90eDoes a 3i driver verb ever take a bee-phrase naming the vehicle, before or after it (with T89a)?Y&M sentence search for bee/yee across all ł-causative 3i vehicle entries

T89d answered (DV-4, HYPOTHESIS; data SOURCED). T89b answered in part (DV-10). Still open: T89a, T89c, T89e; T88a, T88b, T88d, T88e; T87a, T87c, T87d, T87e; T85a, b, c, d, f; T86a–g. New: T90a–e.

Position labels: Young & Morgan's "Pos. IV, 3i obj." on ʼahééʼa- belongs to the prefix entry, not to the verb, so it does not show that the passenger 3i is an object (T90b answered)

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

Verbs on this page → dictionary: ní ‘say’ · sidá ‘sits; with yii’/bii’ ‘in’: sits in’

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. Flat-S and other pre-VPISH notation below is labelled contrast. 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 derivation step annotated “(VPISH)” now reads “(VPISH hyp.)”: VPISH names the hypothesis that the subject starts in Spec,vP; it is not a node. 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.

Driver-3i noticed something odd. Young & Morgan (1987) gloss the prefix ʼahééʼa- as "Pos. IV, 3i obj. pron." (49/14), and they cite it for the passenger intransitives 49/19 and 50/2, where the 3i is the thing that moves with the rider. If that label is about the verb, the passenger 3i sits in the object slot, which would be a sign it is unaccusative (T90b, cf. T87a). This run read the position labels of all 298 prefix entries in the dictionary's "3i" search and joined them to 121 verb entries that cite those prefixes. The label does not track the verb. The same morphemes, ʼahéé-/ʼahéná- 'around in a circle' plus ʼa-, carry "Pos. IV" in one prefix entry (49/14) and "Pos. IV-V" in another (44/17). Each of the two entries is cited by both causatives, where the 3i is the object, and intransitives, where it is the subject. Across 46 passenger-vehicle 3i intransitives with a single cited prefix, 42 cite a prefix labelled IV-V or IV/V, 2 cite V only, and just the 2 ʼahééʼa- verbs cite IV only. Wherever Y&M comment on an intransitive 3i at the level of the verb, they call it the subject. SOURCED (quotes, labels, counts) / HYPOTHESIS (what the labels mean, and everything under D) / OPEN (whether the passenger 3i is an internal argument).

On this page: 1 Question · 2 Sources · 3 Rules first · 4 The circle family · 5 Passenger verbs · 6 Examples · 7 Answers · 8 Derivations · 9 What this does to D · 10 Fidelity notes · 11 Open questions

Notation follows driver-3i. A is the word level (prefixal, left-headed) and B is the clause level (right-headed). 3i is Y&M's indefinite ʼa- and Ø is the ordinary third person. CIRCLE is ʼahéé-/ʼahéná-. Pos. IV and Pos. V are Y&M's prefix positions, which they gloss "obj. pron." and "subj. pron." respectively for the 3i. "IV-V" and "IV/V" are their labels for a 3i assigned to both positions. d is Grant's light determiner (GRANT'S CONJECTURE, treated here as HYPOTHESIS). Under analysis D, d takes a DP complement as v takes VP, and only d joins the verb word. "Label" always means the label printed in a prefix entry. No Syntax · Semantics · Interface conventions file is in the repo or the context folder.

1. Question

Focus 2 asks how the prefixes fit with the nouns, or the silent arguments, that fill their slots. DV-8 on driver-3i proposed that the vehicle 3i is always the theme. It is the object of the ł causatives and the subject of the passenger intransitives, and under VPISH + D it is one internal d in both. That proposal needs the passenger 3i to be an internal argument, which is unaccusativity, T87a. T90b asked whether Y&M's "Pos. IV, 3i obj." on ʼahééʼa- already shows this, since that prefix heads the intransitives 49/19, 50/1 and 50/2. If Pos. IV means the 3i sits in the object slot, the passenger 3i is internal. The counter-evidence is adii- 28/6, which is labelled "Pos. V, 3i subj."

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

2. Sources

Opened this run. Young, Robert W. & William Morgan Sr. 1987. The Navajo Language: A Grammar and Colloquial Dictionary, revised edn. Albuquerque: University of New Mexico Press. Read through the University of Colorado search interface (verbs.colorado.edu/~mahu0110/ym/). Saved on the box under [local working copy, not in the repository] and [local working copy, not in the repository].

  • Definition search "3i" (385 entries, from the T89d run). Further definition searches this run: "3i subj" (35), "3i person" (20), "3i pers" (22), "3i obj" (296), "something" + "it" (95), "corresponding to" (218), "something moves" (11), and the word search ʼahééʼa (12).
  • All 298 prefix entries in these results. Their "Pos." labels were extracted by script and spot-checked by hand.
  • Prefix entries quoted below: ʼahééʼa- 49/14, ʼahénáʼa- 44/17, ʼa- 116/5, adii- 28/6, chʼíʼa- 286/15, ʼidiyii- 465/10, ʼałnááʼa- 98/8, háʼa- 401/18, biʼdi- 212/3.
  • Verb entries read in full: 16/15, 28/16, 33/5, 33/13, 44/18, 44/19, 45/1, 45/2, 49/15–19, 50/1, 50/2, 51/9, 286/19, 362/8, 401/9, 406/14, 465/11, 584/8, plus the 56 passenger entries listed in §5 (definition line and cited prefix).

Nothing else was opened. Y&M's grammar chapter on prefix positions was not opened, because the interface has only the dictionary.

3. Rules first

LB-1  TWO SLOTS, ONE STRING. When one 3i ʼa- stands alone between Pos. I-III        HYPOTHESIS (formal)
      material and the Pos. VI-IX material, nothing in the string shows
      whether it fills IV or V. The two slots are adjacent, and no exponent
      can come between them. Only a form where BOTH slots are filled can show
      the order, and then IV comes first: biʼdi- 212/3 "Pos. IV, obj.          SOURCED (212/3)
      pron., him + '(a)di-, Pos. V, 3i subj. pron. = bi'di-, forms Passive B".
      So in a single-3i verb, which slot the 3i fills cannot be read off
      the form.
LB-2  LABELS ARE PRINTED ON PREFIX ENTRIES. Y&M give "Pos." labels in the       SOURCED
      prefix entries. A verb entry cites one prefix in parentheses, e.g.
      "('ahéé'a-.)" in 49/19. Of the verb entries read in full for this
      page, only 33/13 prints a Pos. label inside the verb entry (LB-8).
LB-3  LABEL COUNTS over the 298 3i prefix entries: IV obj. 131; IV-V           SOURCED (counts; script + hand check)
      obj.-subj. 124; V subj. 21; IV-V written without "subj." 9 (several
      of these, e.g. 98/8 and 401/18, read "IV/V (3i object/subject
      pronoun)"); IV obj.-subj. 2; V obj. 1 (yináʼa- 765/9); no clear
      label about 13.
LB-4  THE GENERIC ENTRY HAS BOTH. ʼa- 116/5: "Pos. IV, 3i direct obj. pron.  SOURCED
      … and Pos. V, 3i subj. pron." So Y&M separate the two positions in
      principle.
LB-5  SAME MORPHEMES, TWO LABELS. 49/14 ʼahééʼa-: "'ahéé-, around in a         SOURCED
      circle + 'a-, '(a)-, Pos. IV, 3i obj. pron., something (used
      thematically in some constructions)." 44/17 ʼahénáʼa-: "'ahéná-,
      'ahéé-, around in a circle + 'a-, '(a)-, Pos. IV-V, 3i obj.-subj.
      pron., something."
LB-6  THE LABELS CROSS THE VERB TYPES. In the circle family (§4), the IV        SOURCED (citations, Lit. glosses)
      entry 49/14 is cited by the causatives 49/15, 49/16 and 49/17 (3i =   / HYPOTHESIS (function reading)
      object, "to cause to move") and by the intransitives 49/19 and 50/2
      (3i = the thing that moves with the rider) and 50/1. The IV-V entry
      44/17 is cited by the causatives 44/19 and 51/9 ("Lit. to cause
      something to fly/move floating …") and by the intransitive 44/18
      (rider on an unspecified quadruped).
LB-7  PASSENGER COUNT (§5). There are 56 passenger-vehicle 3i intransitives      SOURCED (counts) / HYPOTHESIS (set)
      ("bił ---", unspecified vehicle). 46 cite a prefix with a single
      entry: IV-V or IV/V 42, V only 2 (286/19 chʼíʼa-, 465/11
      ʼidiyii-), IV only 2 (49/19, 50/2, both ʼahééʼa-). 10 cite a prefix that has
      homonymous entries with different labels (ʼadii-, ʼani-, ʼa-,
      ńʼdii-). These are left unresolved.
LB-8  VERB-LEVEL STATEMENTS SAY SUBJECT. 16/15: "it is represented in the       SOURCED
      verb by a 3i subject pronoun 'a-". 362/8: the 3i "does represent the
      substance functioning as the subject of the verb". 406/14 "(3i
      subject.)". 584/8 "(the boat is the subject)". The literal glosses
      (49/19 "something floats around in a circle with him"; 286/19,
      465/11, 28/16, 33/5 "something … with him/me") put the 3i in subject
      position of the English. The only mixed case is 33/13 (thunder):
      "'a-, Pos. V, 3i obj. pron., something ?", with Y&M's own query.
LB-9  T90b ANSWER. "Pos. IV" on ʼahééʼa- is a property of that prefix entry. It    HYPOTHESIS (data SOURCED)
      is probably written with the causative and thematic uses in mind
      (49/14's own note: "used thematically"). It is carried over to every
      verb that cites the entry, so it is NOT evidence that the passenger
      3i of 49/19 and 50/2 fills the object slot. By LB-1 the form could
      not show this in any case.
LB-10 WHAT STAYS OPEN. Y&M's "subject" (LB-8) is a grammatical function.      OPEN
      Under VPISH an unaccusative theme and an unergative agent are both
      surface subjects, so LB-8 does not decide T87a either. Neither the
      labels nor the strings decide whether the passenger 3i is internal.
      DV-8 stays a HYPOTHESIS.
LB-11 T90d IN PART. 45/1 and 45/2 have the same mismatch: the Prefix field      SOURCED (fields, forms) / HYPOTHESIS
      says ʼahénáʼa-, the parenthesis says ('ahéná-). 45/2's forms have no
      3i (P ʼahénáldloozh, I ʼahénáldlóósh, sentence with overt mąʼii
      'coyote'), and its stem *ldlóósh has no initial glottal stop that
      the field could have picked up. So the field is unreliable for this
      pair, and the Ø reading of 45/1 (driver-3i note 5) gains support.
      Print p. 45 is still needed.

4. The circle family: label against function

SOURCED (citations, Y&M glosses) / HYPOTHESIS (the "3i function" column is our reading of Y&M's English and of the classifier).

Cited prefix (label)3i = object of a ł causative3i = subject of an intransitiveThematic / other
ʼahééʼa- 49/14 (Pos. IV)49/15, 49/16, 49/17 (3)49/19 boat, 50/2 plane, 50/1 'pass (an hour)' (3)49/18, 51/8, 51/10, 51/11, 51/12, 51/13 (6, not classified)
ʼahénáʼa- 44/17 (Pos. IV-V)44/19, 51/9 (2)44/18 quadruped (1)45/1, 45/2: field only, forms Ø (LB-11)

Both rows have entries in both of the first two columns. If the label recorded the verb's argument position, the IV row would have only objects and the IV-V row would hold the intransitives. Neither is the case.

5. Passenger verbs: which label their prefix carries

SOURCED (each entry's definition line and cited prefix). The set is the entries found by the searches in §2 whose definition begins "bił ---" (or shił/nihił) for travel in or on an unspecified vehicle or mount, plus 406/14 "(3i subject.)" and the stem entry 771/16. It is not guaranteed complete.

Label of cited prefixnEntries
IV-V or IV/V4244/18; 99/1; 106/1; 226/2; 227/3; 253/12; 287/13; 318/16; 376/24, 376/25, 377/1, 377/2 (elevator, escalator); 401/13, 401/15, 401/20, 401/22, 403/10; 402/4; 403/25; 405/16; 406/9, 406/13, 406/14, 408/18, 409/2; 407/4; 468/20; 548/14, 548/16, 553/9, 553/10, 554/17; 558/3; 584/8, 584/15; 639/16; 677/7; 744/27; 749/22; 771/16, 771/17, 771/19
V only2286/19 (chʼíʼa- 286/15 "Pos.V, 3i subj.pron."), 465/11 (ʼidiyii- 465/10 "Pos. V, 3i subject pron.")
IV only249/19, 50/2 (ʼahééʼa- 49/14)
unresolved (homonymous prefix entries)1028/16, 33/5 (ʼadii-((1)): entries labelled IV ×5, V ×2; the inceptive 28/6 is V); 115/17, 122/4 (ʼani-: IV-V and IV); 468/1, 468/2, 472/13, 681/24, 683/7 (ʼa-: 116/5 IV+V, others IV-V); 676/13 (ńʼdii-: IV and IV-V)

So the IV-only label is the outlier among passenger verbs (2 of 46). It is also the label of the prefix that heads the most causatives in its family.

6. Examples

Forms and translations are Y&M's. The segmentation and glosses are ours (HYPOTHESIS). PFV is the perfective cell, CL a classifier, Ł the causative ł classifier.

E1. Y&M 49/19, verb ʼahééʼíʼeeł ʼbił --, to ride in a circle (in a boat). (Lit. something floats around in a circle with him.)' Prefix ʼahééʼa- (Pos. IV).
… yikááʼ bił ʼahééʼníʼéél jiní
… 3-on.surface 3-with CIRCLE-3I-PFV-float.PFV it.is.said
'Magellan was reportedly the first man to sail around the world (circumnavigate the globe).'

E2. Y&M 50/2, ʼahééʼítʼááh ʼbił --, to fly around in a circle (as a passenger in an unspecified flying vehicle)'. Prefix ʼahééʼa- (Pos. IV).
ʼAhbínídą́ą́ʼ ńléí dził binaago shił ʼahééʼítʼaʼ
this.morning yonder mountain 3-around 1SG-with CIRCLE-3I-PFV-fly.PFV
'I flew around that mountain this morning.'

E3. Y&M 44/19, bił ʼahénáʼáshtʼááh 'to ride him around in a circle (in a plane): (Lit. to cause something to fly around in a circle with him)'. Prefix ʼahénáʼa- (Pos. IV-V).
ʼAdą́ą́dąąʼ naalyéhé yá sidáhí dzíł binaago shił ʼahééʼníłtʼaʼ
yesterday trader mountain 3-around 1SG-with CIRCLE-3I-PFV-Ł-fly.PFV
'yesterday the trader flew me around the mountain.'

E4. Y&M 286/19, chʼíʼíʼeeł ʼbił ---, to sail out … (Lit. something floats horizontally out with him.)'. Prefix chʼíʼa- (Pos. V).
Tłʼéeʼgo shił chʼíʼníʼéél …
at.night 1SG-with out-3I-PFV-float.PFV
'I sailed out by night …'

E5. Y&M 45/2, ʼahénáldlóósh 'to go around in a circle (on all four), to trot around in a circle'. Prefix field ʼahénáʼa-, parenthesis (ʼahéná-.).
Mąʼii tsé yinaago ʼahénáldloozh
coyote rock 3-around CIRCLE-CL-go.on.all.fours.PFV
'the coyote trotted around the rock.'

E1 and E3 are glossed alike up to the classifier (CIRCLE-3I-PFV-…), yet one cites an entry labelled IV and the other an entry labelled IV-V. E4 has a V-labelled prefix and the same passenger frame as E1. Only Ł separates E3 from E2.

7. Answers

T90b, answered (HYPOTHESIS; data SOURCED). No. The "Pos. IV, 3i obj." of 49/14 does not show that the intransitive vehicle 3i of 49/19 and 50/2 is in object position. Y&M print position labels per prefix entry (LB-2). The same morphemes get IV in one entry and IV-V in another (LB-5), and both entries head causatives and intransitives alike (LB-6, §4). 42 of 46 passenger verbs cite an IV-V or IV/V prefix and 2 cite V only (LB-7). Wherever Y&M speak at the level of the verb, the intransitive 3i is the subject (LB-8). The form could not decide the question anyway (LB-1).

T87a, not advanced (OPEN). Ruling the labels out leaves unaccusativity untested. "Subject" in LB-8 names a grammatical function, not a position at first Merge.

T90d, in part (HYPOTHESIS). LB-11: the interface's Prefix field is unreliable for 45/1–45/2, so 45/1 is probably Ø.

8. Derivations

All under analysis D, with VPISH (subjects start inside vP). Pre-VPISH
flat-S notation is not used. "B" = clause syntax (right-headed); "A" =
inside the word (left-headed prefix string). The arrow B → A marks the
changeover, where heads move into the word.

D1  E1 … yikááʼ bił ʼahééʼníʼéél (49/19), passenger 3i, Ø classifier
    Option U (unaccusative; DV-8)          | Option E (unergative)
1  Merge d = 3i, no complement          B  | same                          B
2  Merge [VP d V], V = √'eeł 'move        | Merge [VP V] (no internal d)  B
         floating'                       B  |
3  Merge PP [bi- -ił] 'with him' (rider), adjunct to VP                      B
4  Merge PP [yi- -káá'] 'on its surface' (the globe), adjunct                B
5  Merge v (Ø CL; no external argument) B  | Merge v; d = 3i in Spec,vP    B
6  Move  d → T-domain subject position (EPP); V → v → … (head chain)    B → A
7  Agree subject agreement ← 3i (the only argument d)                       A
8  Spell CIRCLE ʼahéé- + ʼ (3i) + PFV ní- + Ø + ʼéél → ʼahééʼníʼéél    A
         The 3i exponent is in the same place under U and E: one ʼ after
         CIRCLE, before PFV. Whether that place is "IV" or "V" is not
         visible (LB-1).
9  Spell yikááʼ bił left of the verb word; jiní after it                 B
    U and E differ only at steps 2 and 5, which are B steps. A spells them
    out the same way. That is why neither the form nor the label (LB-9)
    can settle T87a.

D2  E3 … shił ʼahééʼníłtʼaʼ (44/19), driver: 3i is the object
1  Merge d = 3i, no complement                                              B
2  Merge [VP d V], V = √t'ááh 'move flying'                                 B
3  Merge PP [shi- -ił] 'with me' (rider)                                    B
4  Merge v = Ł (causative); DP 'the trader' (3) in Spec,vP                  B
5  Move  V → v; object d → V–v; subject DP → subject position          B → A
6  Agree subject cell ← 3; object slot ← 3i (p. 50 row 3)                    A
7  Spell ʼahéé- + ʼ (3i) + ní- + ł + tʼaʼ → ʼahééʼníłtʼaʼ                    A
8  Spell [naalyéhé yá sidáhí] … shił left of the verb word                    B
    The 3i lands in the same place as in D1 (after CIRCLE, before
    ní-). Only ł (A) and the external DP (B) are different. Here the 3i
    really is the object, but that is shown by Ł and the agent, not by
    the 3i's position or by the label (44/17 says IV-V).

D3  Where the label could matter: a form with both slots filled
1  Merge [VP DP_him V]; d_obj = Ø/3 pronoun 'him'                           B
2  Merge v (passive); implicit agent d = 3i in the subject role         B (VPISH hyp.)
3  Move  both d's into the word                                         B → A
4  Spell IV bi- + V ʼ(a)di- = biʼdi- (212/3, Passive B)                       A
    Only here is "IV before V" visible. No passenger verb in §5 has a
    second argument d, so none can show its 3i's slot.

9. What this does to D

HYPOTHESIS Nothing in D changes, but one apparent piece of support is gone. DV-8 (one internal vehicle d, object under Ł and subject without it) could have drawn on the IV label of ʼahééʼa-. It cannot now, so it rests only on the VPISH expectation that motion themes are internal. For Focus 1, the result is clearer. The word-level ruleset A spells the 3i into one place whatever its clause-level (B) function and first-Merge position (D1 vs D2). The A/B changeover therefore erases the distinction T87a is about. A ruleset A that only orders exponents needs no "IV vs V" choice for a single 3i: one 3i cell after Pos. I and before the mode cell is enough. Two cells are needed only when two argument d's enter the word together (D3). On this view Y&M's IV/V split is a fact about multi-d words, not about every 3i. For Focus 2 nothing new is claimed. This run did not recount sentences, so NU-8 (the vehicle 3i d takes no DP complement) is neither strengthened nor weakened.

10. Fidelity notes

Everything here was copied from interface HTML saved on the box. Nothing was checked against the print edition or with a speaker. (1) The labels were extracted by script by searching back from each "3i" for the nearest "Pos." Odd spacing and punctuation ("Pos.IV", "Pos,Id", "IV/V") were normalised by hand for the entries quoted. The about 13 unclear prefix entries are not counted in any label class. (2) When a verb cites a prefix with several homonymous entries ("((1))", "((2))"), the interface's numbering does not always line up with an entry. Those 10 verbs are reported as unresolved, not assigned. (3) The 45/2 sentence is printed with spaces (ʼ ahéná ldloozh), and the P form is taken from the paradigm line. (4) 771/16 has a broken headword in the interface. It is counted from its definition line, "Lit. something in the nature of a wheeled vehicle … with me". (5) The "3i function" readings in §4 depend on Y&M's English and on the classifier. Neither is a syntactic test. (6) 50/1 ('pass (an hour)', "something moved around to complete one circle --- the clock hand") is not a passenger verb. It is listed in §4 but left out of §5.

11. Open questions

IDQuestionWhat would settle it
T91aDo 49/14 ("Pos. IV") and 44/17 ("Pos. IV-V") differ in print, or is one of them an interface error?Print pp. 44 and 49
T91bBy what rule do Y&M assign IV, IV-V or V to a 3i prefix entry: by the first verb listed, by the thematic use, or by something else?Y&M grammar section on the prefix positions (not in the interface)
T91cIs there a source-backed unaccusativity diagnostic for Navajo that can be run on the passenger 3i verbs (49/19, 50/2, 286/19)? Without one, T87a and DV-8 stay where they are.Navajo literature on unaccusativity, e.g. any sources in the context folder that discuss it, applied to these entries
T91dWhich homonymous entry do the 10 unresolved passenger verbs cite (ʼadii-((1)) for 28/16 and 33/5; ʼani-, ʼa-, ńʼdii-)?Print, where the numbering is visible
T91eWhat does 33/13's "ʼa-, Pos. V, 3i obj. pron., something ?" mean for an intransitive thunder verb, and is yináʼa- 765/9 (the one V-obj. prefix) the same kind of case?Read 765/9 and its verbs; print p. 33

T90b answered (LB-9, HYPOTHESIS; data SOURCED). T90d answered in part (LB-11). Still open: T90a, T90c, T90d (print), T90e; T89a, T89c, T89e; T88a, T88b, T88d, T88e; T87a, T87c, T87d, T87e; T85a, b, c, d, f; T86a–g. New: T91a–e.