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

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

Is Navajo bi- a deep inverse? Oxford 2023's tests run on Navajo

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

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

Verbs on this page → dictionary: yiztał ‘kick’ · yiztsʼǫs ‘kiss’ · yishxash ‘bite’ · dabiiłtsą́ ‘see’ · yóóʼeelwod ‘run away’

Restated on the unified structure (2026-11-15). The inverse is restated as one extra Move on the clause spine: the patient moves (via an outer Spec,vP) above the agent, which has already raised to Spec,TP by the working default (H3, H6′, H21b). The event LF is the same as the direct; the trigger and landing site are OPEN. The head-direction mismatch argued below as rule sets A/B or a changeover is now stated as Grant’s Merge-site hypothesis and tested in §3.1–3.2. Flat-S and other pre-VPISH notation below is labelled contrast. Status on the spine: see unified structure, §4, inverse row. The data, judgments, translations and citations below are unchanged; the analysis below is kept as the record of what was argued then.

The Algonquian inverse page (T52) found that Algonquian has two inverses. A shallow one only changes the agreement marking. A deep one really moves the patient up into the subject slot (Oxford 2023). That page left T52d open: which kind is Navajo bi-? This page runs Oxford's four tests on the Navajo data in Willie 1991, one at a time. In plain words: on every test that can be run, bi- behaves like the deep kind. That holds even in the sentences where animacy forces bi-, because Oxford's Algonquian data show that a forced inverse can still be deep. The one thing that does not follow the patient is the plural prefix da-, which keeps counting the doer. So Navajo bi- is "deep" in where the participants sit, but its plural marking stays tied to the roles.

Update 2026-10-25 (~8:46pm PT): Saxon 2003 (two-osv) is a sourced "deep" account. The bi- object Moves to Spec,DiscP above the 3rd-person subject in Spec,NumP, and the prefix is read from the landing site, unlike A-bar Focus Movement, which keeps yi-. It is a structure, not a test, so D2 stays OPEN.

On this page: 1 Question · 2 Sources · 3 Oxford's tests · 4 Navajo data · 5 The tests run · 6 Forced but deep: rule RCC · 7 What does not flip: da- · 8 Derivations · 9 Grant's ideas · 10 Bottom line · 11 Open

Notation (the site's own; no Syntax · Semantics · Interface bot conventions file exists, so the provisional notation is kept). Σ = role-schema step; A = word level (rule set A: the prefixal, left-headed spine, spelled out as one word W); B = clause level (rule set B: right-headed clause syntax, noun linking, enclitics). W is the switch from A to B. ρA (agent slot), ρP (patient slot): role labels ours. r(ρ) = the animacy rank of the slot's referent as construed (eight ranks). INV-1…INV-7: inverse page. IA-1*: inverse as Agree and EPP. DIST-1: number & scope. Deep / shallow: Oxford 2023's terms (§3 below).

1. The question

Reminder (Grant mixes these up sometimes): Merge builds structure; moving something already built is internal Merge. Agree copies person and number features from an argument onto a head; nothing moves. A shallow inverse is Agree only. A deep inverse is Agree plus one internal Merge of the patient (Algonquian inverse §6).

2. Sources opened this run (2026-10-23, ~4:46am PT run)

  • Oxford, Will 2023. "A tale of two inverses." Syntax 26: 311–354. Author's preprint (lingbuzz/006567), text layer. Re-read: (4) p. 4; §4 pp. 8–17 (§4.1 contrast, §4.2 binding, §4.3 word order, §4.4 long-distance agreement); (39)–(41) pp. 25–26; §7.3, (44)–(45) and Table 4, p. 29; §8, Table 5, p. 31. Page numbers are the preprint's.
  • Willie, MaryAnn 1991. Navajo Pronouns and Obviation. PhD diss., University of Arizona. Box PDF. Navajo forms below are read from the page images (the text layer garbles Navajo): pp. 71, 77, 82–83, 86–88, 160–162. Text layer only: pp. 14, 76, 84–85, 135–139, 224, and the "no complement clauses" remark on PDF p. 230.

Nothing else was opened. Sandoval & Jelinek, Bruening 2005 and Hale 1997/2003 are cited only as already recorded on earlier pages of this site, and no Navajo form is taken from them here.

3. Oxford's tests SOURCED

Oxford 2023 (4), p. 4: a deep inverse is inverse voice (the patient is the subject); a shallow inverse is inverse alignment (the patient is only indexed like a subject). His §4 gives four ways to tell them apart:

TestDeep (3′→3 in Algonquian)Shallow (3→1/2 in Algonquian)
D1 Contrast (§4.1)A choice between direct and inverse, by topicalityNo choice: fixed by person
D2 Binding (§4.2)Reversed: the patient binds into the agentThe agent still binds into the patient
D3 Word order (§4.3)Follows the patient as subjectFollows the agent
D4 Long-distance agreement (§4.4)Targets the patientTargets the agent

Two refinements matter for Navajo.

4. Navajo data used SOURCED

All forms are from Willie 1991 as printed (page images). The glosses are Willie's own where she gives them. Where she gives only a translation, the word glosses (in square brackets) are ours, following her glossing of the same words elsewhere on the page.

(37a) Willie p. 71
łį́į́ʼdzaanéézyiztał
horsemule3.3.kicked
‘The horse kicked the mule.’
(37b) Willie p. 71
dzaanéézłį́į́ʼbiztał
mulehorseINV3.3.kicked
‘The mule was kicked by the horse.’

Willie p. 71: "The Inverse construction is comparable to a passive in that the subject 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."

(46) Willie p. 77
’ashkiiʼatʼéédléiʼbiztsʼǫs
[boygirlINDEFINV3.3.kissed]
‘The boy was kissed by a girl.’
(47) Willie p. 77
*’ashkiiléiʼʼatʼéédbiztsʼǫs
[boyINDEFgirlINV3.3.kissed]
[‘A boy was kissed by the girl.’]

Willie p. 77: "with the Inverse, the external argument (the patient, with Absolutive case) must be definite." Examples (48)–(51) on the same page show the same pattern with łaʼ ‘some’. In the direct voice either argument can be indefinite (pp. 76, 88).

Willie (70), p. 86: "If the agent is higher on the animacy scale than the patient, then the Direct voice is used: if the patient is higher on the scale, then the Inverse must be used. And if they are equal, the choice rests on discourse factors." Same page: "The voice alternation can only be used if all arguments are third person." And: "even when the sentence does not contain any nominals, the sentence must be interpreted with reference to the animacy scale."

(71) Willie p. 87
a.yishxashb.bishxash
3.3.bitINV3.3.bit

Willie (72), p. 87, the readings: (72a) yishxash: 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). (72b) bishxash: 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). P. 88: "Each paradigm permits what the other excludes."

(17) Willie p. 160
a.’ashkiiyiztsʼǫs1. ‘She kissed the boy.’ 2. *‘The boy kissed her.’
b.’ashkiibiztsʼǫs1. ‘She was kissed by the boy.’ 2. *‘She kissed the boy.’
(18) Willie p. 161
yiztsʼǫs-yę́ęyóóʼeelwod
3.3.kissed=REL3.ran.away
‘The one he kissed ran away.’
(19) Willie p. 161
biztsʼǫs=yę́ęyóóʼeelwod
INV3.3.kissed=REL3.ran.away
‘The one who kissed him ran away.’

Willie (16), p. 160: "The heads of relative clauses are absolutives … the 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". (19) is agent-headed "because the agent is the internal argument in the Inverse construction" (p. 161).

(20) Willie pp. 161–162
a.ʼatʼééd’ashkiiyiztsʼǫs=yę́ęyóóʼeelwod
[girlboy3.3.kissed=REL3.ran.away]
‘The girl who kissed the/a boy ran away.’
b.ʼatʼééd’ashkiibiztsʼǫs=yę́ęyóóʼeelwod
[girlboyINV3.3.kissed=REL3.ran.away]
‘The girl who was kissed by the boy ran away.’

Willie (21), p. 162: "When both arguments of the relative have nominal adjuncts, the relativizing enclitic binds the external argument, which must be definite." She adds that Platero 1982 reports some speakers allowing two readings, but "this is not true of my speech."

(63c) Willie p. 82
ʼashiikéʼatʼééddayiiłtsą́
boysgirlPL.3.3.saw
‘The boys saw the girl.’
(63d) Willie p. 82
*’ashkiiʼatʼéékédayiiłtsą́
[boygirlsPL.3.3.saw]
(‘The boy saw the girls.’)
(64) Willie p. 83
a.dabiiłtsą́‘They/he (was) were seen by them/*him.’
PL.INV3.3.saw
b.shizhéʼéʼashiikédabiiłtsą́
1s.fatherboysPL.INV3.3.saw
‘My father was seen by the boys.’

Willie p. 82: "the agent falls under the scope of the distributive marker". In (64b) the agent (the boys) is plural and the patient (my father) is singular, and the sentence is good. P. 84 adds that with dabi- the external argument "may optionally be interpreted as plural", and a plural external argument can escape the definiteness requirement.

5. The four tests, run on Navajo

TestNavajo factsResult
D1 ContrastEqual rank: free choice by discourse ((70), (37), (72a.1/b.1, a.3/b.3)). Unequal rank: forced ((70), (72a.4/b.4)).SOURCED facts. HYPOTHESIS This matches Oxford's 3obv→3obv (optional, deep) plus 3obv→3px (forced, deep), not his 3→SAP (forced, shallow). See §6.
D2 BindingNo Navajo binding data in any source opened. T11b (fourth person compared) found the test may not be constructible. Update 2026-10-24 ~12:46pm PT: possessor coreference data found (Young & Morgan 1980 via Aissen 2000): ‘Mr. Longi's wife looks at himi’ needs bi-, ‘the boyi saw hisi horse’ needs yi- (possessor-binding).OPEN Silence, so it counts neither way. Update: HYPOTHESIS the four cells fit a deep bi- (Oxford (45)), but obviation and word order predict them too, and the antecedents are names, not quantifiers. Still OPEN.
D3 Word order and noun linkingThe noun next to the verb is the internal argument in both voices: the patient with yi-, the agent with bi- ((37), (17)). The outer noun is the external argument: the agent with yi-, the patient with bi-.SOURCED facts (Willie pp. 71, 160). HYPOTHESIS Order follows position, and the positions swap with voice: deep-like. Caveat below.
D4 Long-distance agreementNavajo has "no complement clauses … no long-distance Wh- movement" (Willie, PDF p. 230, text layer).SOURCED Not applicable.
D4′ Substitute: relativizationThe relative head is the internal argument when it has no noun ((18) patient, (19) agent) and the external argument when both have nouns ((20a) agent, (20b) patient).SOURCED facts. HYPOTHESIS Which participant can head the relative changes with voice, so it tracks positions, not roles: deep-like.
Extra: inverse pragmaticsThe inverse subject must be definite ((46)/(47), p. 77; p. 88), unless da- licenses a plural (p. 84).SOURCED This fits Oxford's deep inverse, where the patient is the more topical argument.
Shallow side?"The voice alternation can only be used if all arguments are third person" (p. 86).SOURCED No 3→1/2 inverse, so there is nothing for a shallow inverse to be.

HYPOTHESIS Caveat on D3 and D4′. On Willie's own analysis the nouns are adjuncts linked to the verb's pronouns. A purely morphological inverse could in principle tell the hearer "the noun next to the verb is the agent" without anything moving. So D3 and D4′ show that the positions swap, but they can't by themselves rule out a morphology-only story. Oxford's decisive test is binding (D2), and that one is open. The definiteness fact helps: a morphology-only inverse gives no reason for the patient to have to be definite, whereas a raised subject that is the topic does.

6. Forced but deep: rule RCC HYPOTHESIS

In INV-3 the site already states the ranking as "EXT = the slot whose referent ranks higher". Oxford's (45) suggests restating it as a constraint on structure, which makes the parallel to the Algonquian forced deep inverse exact:

RCC (Rank C-command Constraint; ours, after Oxford 2023 (45)). In a word whose verb has two third-person slots that the voice compares, a slot whose referent ranks higher may not be c-commanded by a co-argument slot whose referent ranks lower, at the point where W is sealed.

Checked against all eight readings of (72):

(72)RanksStructure at SealRCCWillie
a.1 he bit himA = PρA over ρPokok
a.2 he bit itA > PρA over ρPokok
a.3 it bit itA = PρA over ρPokok
a.4 it bit himP > Alower ρA over higher ρPviolated*
b.1 he was bitten by himA = PρP raised over ρAokok
b.2 he was bitten by itP > AρP raised over ρAokok
b.3 it was bitten by itA = PρP raised over ρAokok
b.4 it was bitten by himA > Plower ρP raised over higher ρAviolated*

HYPOTHESIS 8 of 8 match. RCC predicts the same facts as INV-3 for two-slot verbs, so this is a restatement, not a new prediction there. What it adds: (i) the obligatory bi- comes out as a deep inverse forced by a ban on bad c-command, exactly Oxford's account of 3obv→3px. Being obligatory is therefore no sign of a shallow, Agree-only inverse. (ii) The answer to T52's second half: Navajo's only obligatory side is this rank-forced deep inverse, not a person-hierarchy Agree inverse.

OPEN How hard is RCC? Oxford's PCC is a grammatical rule in Algonquian. For Navajo, the site has already recorded that Sandoval & Jelinek call a rank-violating sentence "pragmatically unacceptable" rather than ungrammatical, and that only about half of speakers in Young's survey obey the hierarchy (via Willie p. 16; inverse page, INV-3). So RCC may be a soft constraint, or it may vary between speakers (T5b, carried into T53b).

OPEN Three slots. RCC as stated would also apply to a theme that outranks a slot above it. The site's INV-5 says a theme never competes (Willie pp. 94–95). Oxford's ditransitive data (46)–(47), p. 29, show a parallel restriction in Algonquian (only the recipient takes part in the rank comparison). Whether RCC needs a clause excluding themes, or whether themes are low enough that it never matters, is untested (T53c).

7. What does not flip: da- counts the doer

SOURCED In Oxford's deep inverse the agreement head Infl agrees with the raised patient, because the agent is made inaccessible (pp. 25–26). In Navajo the distributive plural da- counts the agent in both voices: (63c)/(63d) in the direct, (64b) in the inverse, where the plural agent and singular patient are fine with dabiiłtsą́ (Willie pp. 82–83). The site has already recorded this as DIST-1 (number & scope).

HYPOTHESIS So Navajo bi- is not a copy of Oxford's deep inverse in every respect. Its positions swap (§5), but the plural agreement stays with the agent. This is what the site's IA-1* already says, following Bruening 2005's footnote 22 as recorded on the inverse as Agree and EPP page: the patient slot moves first, and the person/number head then agrees with the agent left below it. In Oxford's terms, the Navajo agent does not get the inherent Case that would hide it from agreement. Navajo is closer to Bruening's Passamaquoddy analysis than to Oxford's Voiceerg.

HYPOTHESIS Classification. Navajo bi- = a deep inverse in position (D1, D3, D4′, pragmatics), with agreement tied to the roles. It has no shallow counterpart at all.

OPEN P. 84 says the external argument of dabi- "may optionally be interpreted as plural". Whether da- can ever count the raised patient alone (plural patient, singular agent) is not shown by the examples opened (T53d). Update 2026-10-23 ~12:46pm PT: answered, no: see can da- count the patient? (Willie (64a) ‘by *him’, (65b), (67b)). There the optional patient plural is treated as a second, closest-goal probe that reaches the patient only after it has moved (PL-2H; HYPOTHESIS).

8. Derivations

Steps are labelled by rule set: Σ role schema, A word (W), B clause. Each step names its operation: Merge (builds structure), Agree (values person/number), Move (internal Merge), Link (a noun linked to a slot), Seal (spell-out of a unit). VPISH: both participant slots are first-merged inside the verb's phrase, ρA in Spec,v and ρP as complement of V. The spine is the site's Qual > M/S > v > V, with Voice (yi-/bi-) above QualP. No flat-S tree is used. Verb words are not segmented beyond Willie's gloss line.

D-1. (37a) łį́į́ʼ dzaanééz yiztał (direct, equal rank)

  1. Σ Σ(√KICK) = {ρA (the kicker), ρP (the one kicked)}. Theta roles are fixed here and at external Merge, and nothing later changes them.
  2. A Merge V with ρP; Merge v, ρA in Spec,v (VPISH). v has no EPP (IA-1*, direct).
  3. A Merge M/S; Agree M/S ⟶ ρA [3] (the closest goal). Merge Qual; Merge Voice.
  4. A RCC check: r(ρA) = r(ρP) ("of equal animacy", Willie p. 72). ρA c-commands ρP, and that is allowed. Voice spells out as yi- because ρA is the highest slot. Seal D1 → W = yiztał, as printed.
  5. B Merge the nouns and Link them (INV-6): dzaanééz, next to W, links to the internal slot ρP; łį́į́ʼ links to the external ρA. Seal D2: łį́į́ʼ ⌢ dzaanééz ⌢ W ✓ ‘The horse kicked the mule.’

D-2. (37b) dzaanééz łį́į́ʼ biztał (inverse, equal rank: optional deep inverse)

  1. Σ As in D-1: the same two roles.
  2. A Merge V with ρP; Merge v[EPP], ρA in Spec,v. Discourse chose the EPP version (equal rank).
  3. A Move: v's EPP raises ρP to outer Spec,v, over ρA (the "crossing" of the Algonquian page). This is internal Merge, not Agree.
  4. A Merge M/S; Agree M/S ⟶ ρA [3], the agent left below (IA-1*; this is where Navajo differs from Oxford's Voiceerg, §7). Merge Qual, Voice.
  5. A RCC check: equal ranks, so ρP c-commanding ρA is allowed. Voice spells out as bi-. Seal D1 → W = biztał.
  6. B Link (INV-6): łį́į́ʼ, next to W, links to the internal slot, now ρA; dzaanééz links to the external slot, now ρP. Seal D2 ✓ ‘The mule was kicked by the horse.’ The horse is still the kicker.

D-3. (72a.4) and (72b.4): the two stars, schematic (no nouns)

  1. Σ Σ(√BITE) = {ρA, ρP}. Referents as construed: (72a.4) ρA = "it", ρP = "him", so r(ρP) > r(ρA). (72b.4) the reverse.
  2. A (72a.4) Merge with v without EPP: ρA c-commands ρP. Agree M/S ⟶ ρA. RCC: the higher ρP is c-commanded by the lower ρA → violated. Seal would give yishxash, but not on this reading: * ✓.
  3. A (72b.4) Merge with v[EPP]; Move ρP over ρA. RCC: the higher ρA is c-commanded by the lower, raised ρP → violated. Seal would give bishxash, but not on this reading: * ✓.
  4. B No B step: there are no nouns. Both stars are made inside rule set A, before W is sealed, which fits Willie's remark that the scale applies "even when the sentence does not contain any nominals" (p. 86).

D-4. (19) biztsʼǫs=yę́ę yóóʼeelwod: relativization tracks the raised structure

  1. Σ Σ(√KISS) = {ρA, ρP}.
  2. A As D-2: v[EPP] and Move ρP over ρA; Agree M/S ⟶ ρA; RCC satisfied (no rank conflict is given). Seal D1 → W = biztsʼǫs.
  3. B Merge =yę́ę (REL). Willie (16): it binds the internal argument (absolutive) when there are no nouns. After the Move in step 2, the internal slot is ρA. So the head is the kisser: ‘The one who kissed him ran away’ ✓. With yi- (no Move) the internal slot is ρP, giving (18) ‘The one he kissed’ ✓.
  4. B Merge the relative as the subject of yóóʼeelwod and Seal D2 ✓. HYPOTHESIS The enclitic reads the structure built in rule set A, so the deep-inverse Move is visible to a rule-set-B operation. That is a point where the word boundary does not hide A's work from B.

9. Grant's ideas

10. Bottom line

HYPOTHESIS T52d answered as far as the data go: by every one of Oxford's tests that can be run on Navajo (contrast, word order/linking, relativization as a substitute for long-distance agreement, plus definiteness), bi- is a deep inverse. The patient really takes the subject position.

HYPOTHESIS Forced but deep. The obligatory bi- (patient outranks agent) is the Navajo counterpart of Oxford's forced deep inverse, not of the shallow SAP inverse. New rule RCC states it as a ban on a higher-ranked slot being c-commanded by a lower one, and fits all eight readings in (72).

SOURCED No shallow side. Navajo has no inverse for third person acting on first or second person (Willie p. 86).

SOURCED + HYPOTHESIS Not identical to Algonquian. da- counts the agent in both voices (Willie (63)–(64)), whereas Oxford's deep inverse makes agreement follow the patient. Navajo looks more like Bruening's Passamaquoddy analysis (IA-1*).

OPEN Binding, the decisive test, has no Navajo data in any source opened, and RCC's grammatical status (hard vs soft vs varying between speakers) is open. Update 2026-10-24: possessor coreference data now found; they fit "deep" but do not decide it (possessor-binding §5–6).

OPEN Chemistry. No source opened mentions chemistry, and this page claims no chemistry analog.

11. Open questions this page adds

#QuestionWhat would settle it
T53aBinding/scope under bi-: can the raised patient bind into or take scope over the agent (Oxford's D2)?Speaker judgments on quantifier/possessor sentences; T10a, T11b. OPEN Update 2026-10-24 ~12:46pm PT: possessor data found and they fit "deep", not decisive; see possessor-binding (T57a–d).
T53bIs RCC a hard grammatical constraint (as Oxford's PCC is), a soft one, or one that varies by speaker?Young's survey itself; Sandoval & Jelinek re-read; Willie (73). OPEN
T53cDoes RCC need a clause exempting themes (INV-5), parallel to Oxford's ditransitive restriction?A three-slot verb with a theme that outranks the agent or goal, from a source. OPEN
T53dCan da- ever count the raised patient alone (plural patient, singular agent) under bi-?Closed 2026-10-23 ~12:46pm PT: no (can da- count the patient? §4). SOURCED
T53eClause chains with =go (Willie pp. 135–139, (9)): does the "switch in thematic roles" track position (deep) or role? The subscripts in the text layer are damaged and conflict with T31b.A clean scan of Willie pp. 136–137. OPEN
T53fPlatero 1982's speakers who allow two readings of a two-noun relative: do they treat bi- as less deep?Platero 1982 (not opened). OPEN

Builds on the Algonquian inverse (T52), the inverse (T5), the inverse as Agree and EPP (IA-1*), number & scope (DIST-1), relative clauses and fourth person compared (T11).