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 1. The syntax audit found malformed old diagrams and overclaimed conclusions, not a demonstrated headedness switch. The core trees are corrected. Light d, NP-PRO, A/B cycles and B1/B2/B3 are hypotheses; their old working notes remain a historical record, not settled grammar. Current movement verdict and gaps.
2026-11-11 audit pass 3. The “roles, not words” framing was removed from the lede and the reading order (the thesis was dropped on 2026-11-02; those pages stay as a historical record). The “Newest” item now points to the 2026-11-11 sections. SPECULATIVE and NONE were added to the status legend. See the pass 3 report.
Note (2026-11-02): Grant has dropped the “roles, not words” thesis. Pages that test it are kept as a historical record. Current work: the clause-syntax audit (audit pass 3) and the movement diagnostics; light determiners remain a hypothesis. [2026-11-14: light d is withdrawn by Grant; see derivation.]
2026-11-14 overhaul (approved by Grant for the live site, 2026-11-14). The site is now organised around one clause structure (“the spine”). The older summaries further down this page (“Where v0 stands”, the dated “Added” notes) are kept as history; where they differ, the spine pages supersede them. No Navajo form, gloss, judgment or citation changed.
A research notebook on the syntax of the Navajo verb and clause. Every clause is built the same way: the verb merges with its object, the agent is added above them, tense above that, and negation and the clause-type head (question, statement) on top. The verb rises into tense and the subject rises to the front of the tense phrase, as in most languages studied so far. Each Navajo pattern is then the question of what else moves, and why, with its meaning written out so that scope matches structure. Every Navajo form is copied from a named published source with its example number, and every claim carries a status label.
What the site argues
- HYPOTHESIS One spine. V→v→T and subject raising to Spec,TP are assumed in every tensed clause (the spine). VPISH is used for what it is, a way of lining up syntax with meaning (each argument gets its role inside vP), not as an account of word order.
- HYPOTHESIS Verb-final order. The cheapest account has tense, negation and the clause-type head final, so nothing else has to move. Its rival puts tense first and moves the rest of the verb phrase above it. Present data do not decide between them (headedness).
- HYPOTHESIS The inverse is the object moving above the agent; it changes which phrase is highest, not who does what to whom (results).
- HYPOTHESIS Negation. Sentential negation (doo … da) has its head in the final da; constituent negation with hanii is a different thing (results).
- HYPOTHESIS Questions leave the question word where a noun would stand; the question marker is interpreted at the top of the clause and pronounced on a host (results).
- OPEN Still open: whether the polar particle is a head; whether the sourced prefix orders fit Grant’s Merge-site account of the word/clause headedness mismatch (unified structure §3); the landing site of the inverse movement; telicity (no minimal pairs; two narrow claims about enclitic forms are contradicted); and whether the noun phrase has a light n, parallel to light v in the clause (a structural question raised by Grant on 2026-11-14, not a premise). Light d, never needed in thirteen reviews, was withdrawn by Grant on 2026-11-14. See what resists.
Older published analyses (flat clauses with nouns adjoined, the word as a headedness changeover, topic and focus heads) were written before much of current syntax theory. Their data are used exactly as printed; their analyses appear on this site only as labelled contrast. The notebook began with the thesis that the verb takes roles, not words, as its arguments; that thesis was dropped on 2026-11-02.
Start here
- The spine: the premises, the steps every clause goes through, how meanings (LFs) are written, and the tests.
- Results on the spine: the first-principles reviews 1–13, restated on that one structure.
- One structure for Navajo (2026-11-15): the clause, the noun phrase and the verb word on one spine, with every phenomenon placed on it in one table.
- Headedness and the clause spine: what else moves to give verb-final order, and where each element sits.
- The topic pages, each opening with an “On the spine” summary.
Pages
Pages:
- Results on the spine
- One structure for Navajo
- Newest (2026-11-11): syntax audit pass 3; movement diagnostics (V-to-T was then labelled neither proven nor refuted; since 2026-11-13 it is the working default); English phrasal verbs and Navajo verbs; first-principles review 4: Navajo questions
- Hale’s future Qualifier d is the onset of doo: pp. 72 and 541 flip between thematic di and doo though 3rd shows one visible di (T98b)
- Distributive future and the missing di: p. 454 dahozhdiyoo- keeps zh before diyoo because that cell has only one di-piece, while singular hodizhdiyoo- has two (T97a, T97c)
- Where 4th-person zh sits relative to thematic di: in six hodii- future cells it follows di (ʼáhodizhdoo-, hodizhdiyoo-, náhodizhdoo-), so “subject d above every Qualifier” needs a finer spell-out rule (T96d in part)
- The space d ho-: the object prefix comes before the fourth-person subject in every table cell (hoji-, 150 of 150), and hashbįʼ 'build a hogan' cannot take a noun object, a ban that belongs to the verb root, not the prefix (T95b in part; T85a, T85b)
All other pages: Working notes.
Sections
Vocabulary: the template, prefixes and postpositions (advanced, verb dictionary).
Syntax: head direction, linking, the inverse (advanced: every derivation).
Background: Merge and Agree, VPISH, Dowty, event structure, Algonquian, sources.
Essays: eleven journal-style essays, one per topic (RSS, Atom, WebDAV).
Working notes: every page written run by run, unchanged, by topic.
- Glossary and morpheme index · Example sentences · Rules index · Open questions · Sources · About
Reading order
- The spine, then Results on the spine (2026-11-14 overhaul: read these first).
- The Navajo verb, start to finish (verb guide).
- Vocabulary, basic, then Syntax, basic.
- Essay 1: roles, not words (historical record: Grant dropped this thesis on 2026-11-02).
- Essays 2–11 in order (list), each linking to its working notes.
- Derivations and the rules index when you want the formal detail.
- Where the thesis stands and what is still open.
Status legend
| Badge | Meaning |
|---|---|
| SOURCED | Stated in a published source that was opened (this run, or a prior run on this project, re-checked where noted). The citation gives the page or example number. |
| HYPOTHESIS | A formalization, inference or comparison made on this site. No source states it in this form. |
| GRANT'S CONJECTURE | Grant's own claim, quoted or closely paraphrased. It is the thing being tested. |
| OPEN | Undecided. Needs Grant's texts, native-speaker judgments, or a source we could not open. |
| SPECULATIVE | A loose idea or analogy, not yet stated as a testable rule. |
| NONE | Nothing found: no source, form or test for the point exists yet. |
Where v0 stood (one line per phenomenon; older framing, kept as history)
This summary predates the 2026-11-14 spine overhaul and keeps its original wording. For the current framing see What the site argues. The home page’s earlier introduction read: A research notebook on the syntax of the Navajo verb and clause: how the verb word, its prefixes and the clause around it are built, tested against published data. (The notebook began with the question whether the verb takes roles, not words, as its arguments; that thesis was dropped on 2026-11-02.) Formal rules come first; every Navajo form is copied from a named published source with its example number, and every claim carries a status label.
Headedness. HYPOTHESIS The best-supported version of Grant's idea is not "the word is left-headed." It is narrower: functional heads inside the verb word come before their complements, while lexical heads (V, P, N) come after theirs, inside the word and outside it. That is Hale's assumption (2003, n. 4; 2001:689). Grant's version adds one claim on top: the boundary is the word, so functional heads outside the word go back to head-final. That claim can be tested on clause-level particles. They split: da’ comes first; ya’, -go and -ę́ę come last. This is the dataset that decides it.
Proto-role schema. HYPOTHESIS The schema can be stated as a set of slots with fixed entailments, each valued by exactly one verb-internal exponent. Under that statement, the inverse falls out as a mismatch between two orders: Dowty's proto-agent order, which is fixed by the event, and the animacy/status order, which is fixed by the referents. bi- appears exactly when the two orders pick different slots. The fourth person takes a slot out of the ranked set.
Added 2026-10-07: keeping vs bringing about
Page: Keeping a state vs bringing about a change. One Navajo posture stem in three event shapes, from page-checked forms in Goertz et al. 2006: change by the subject, change brought about by a causer, and a state held by a causer. The held participant is the object even though nothing happens to it. The page states the cut as proto-role entailments (onset vs extended causation, after Talmy). It sets the Navajo forms beside sourced English, Japanese/Korean and Chantyal contrasts, and derives all three shapes step by step. HYPOTHESIS The held object is not a problem for Dowty's selection principle or for the PAH. It does expose a gap in this site's ADD-CAUSER rule, which is now patched with a MAINTAIN operation.
Added 2026-10-07 (evening): which verbs take a plain ł-causer
Page: Which intransitive verbs take a plain ł-causer? Some Navajo intransitives become transitive by adding ł alone (‘the rope broke’ → ‘he broke the rope’). Others need the bigger causative with y- and bi- (‘the baby laughed’ → ‘I made the baby laugh’). The page checks Hale & Platero's 1996 lists against Dowty's entailments and his discussion of unaccusatives, with derivations C1–C4. HYPOTHESIS No single entailment predicts the split. A verb resists the plain ł-causer when its one participant is a living being doing something open-ended (crying, burping, sleeping), and possibly also when it acts on purpose. Whether that rule beats listing the verbs by hand is still OPEN.
Added 2026-10-08: the inverse (who outranks whom)
Page: The inverse: when a lower argument outranks the agent. With two third persons, ‘the horse kicked the man’ comes out as a bi- form, ‘the man got kicked by the horse’. The kicker is still the horse; what changes is which participant is the subject. The page reads Willie 1991 §3.3 in full and Sandoval & Jelinek 1989 (page-checked forms), and states the choice as rules over proto-role entailments plus an animacy ranking (NAT, EDGE, INV-0 to INV-7). It derives the horse/man pair, the equal-rank horse/mule pair and Willie's starred readings step by step. HYPOTHESIS The subject is whichever participant ranks higher, and bi- marks that this is not the agent. The only impossible cases are a lower-ranked subject. The choice is made inside the word, before any noun is placed: bare verbs already differ, and are already starred, with no nouns at all. Whether VPISH-style raising of the patient is real is OPEN.
Added 2026-10-09: number and quantifiers across the inverse
Page: Number and quantifiers across the inverse: da- follows the agent. The Navajo verb prefix da- (‘three or more’) always counts the doer. In ‘the boys saw the girl’ it counts the boys. It still counts the boys when the verb flips to the inverse, ‘my father was seen by the boys’, where the boys are no longer the subject (Willie 1991:82–84). So this prefix follows the proto-agent role. The definiteness requirement follows the subject position instead. The page also prints Willie's universal-quantifier pair (58)–(59), now page-checked. The pair does not yet decide the VPISH scope test (T5c). Derivations N1–N4; open questions T10a–T10f.
Added 2026-10-13: asking about a role: wh-questions, and the orders Navajo forbids
Page: Asking about a role: wh-questions and focus. In Navajo a question word like háísh ‘who’ behaves exactly like a noun. It has to be linked to a 3rd-person role in the verb, it takes the place a noun for that role would have, and which role it asks about is set by the same order and voice rules (Willie 1991 Ch. 8). So “who?” asks about a role slot, not a word, which fits Grant's thesis. Willie also lists the orders that are impossible. The question word for the lower argument can go ahead of the agent only when the two can't be confused: ‘What did John see?’ works because ‘what’ can't do the seeing, but with two people, ‘who John saw’ means only ‘Who saw John?’. That conflicts with a later paper (Willie & Jelinek 2000), and the site keeps both. A clause also can't have both ‘where’ and ‘who’, the first sourced fact for Grant's multiple-wh test.
Added 2026-10-11 (evening): which noun is which? Nouns attach to role slots, and the verb's meaning can overrule word order
Page: Which noun is which? Nominal adjunction. In Navajo the verb on its own is a full sentence, and nouns are optional extras added to make clear who is meant. Willie says a noun “by itself cannot be an argument -- it must be licensed by a pronoun” in the verb (1991:52). At the time this was read as the clearest page-checked support for Grant's thesis, since dropped, that the verb's arguments are roles, not words. Which noun goes with which role is decided first by position: a lone noun is the object, and with two nouns the outer one is the doer (pp. 59–60). But the verb's meaning can overrule position. ‘boy picked-round-things’ means ‘the boy picked them’, because a boy is not a round thing (p. 61). And the number of roles decides how a string of nouns is grouped: ‘the old man his car’ is one phrase with ‘is ruined’ and can be two with ‘ruined’ (pp. 58–61). By default the nouns line up as a mirror image of the prefixes, nested around the edge of the verb-plus-postpositions unit. That is a second sign that this edge, not the verb word, is where clause syntax starts. Willie's own ‘the boy is leading the horse for the girl’ breaks the mirror. Against a purely role-driven account: some of her overrides rest on knowledge about the world (“horses typically are led”), and one, ‘the boy is shooting at it’, has no visible reason at all (T18a). Derivations L1–L8; open questions T18a–T18h. Dictionary: ‘grind’, ‘pick (round objects)’, ‘count (plural objects)’, ‘shoot at’ and ‘cook (in something)’ added (97 entries), with new rows on ‘ruin’, ‘lead’, ‘hit with the fist’, ‘speak’, ‘sit’, ‘cry’, ‘kick’ and ‘see’.
Added 2026-10-11 (afternoon): who likes it? The experiencer sits on a postposition, and the subject can be “the area”
Page: Pleonastic ho-, psych verbs, and the experiencer inside a postposition. Navajo ‘I like it’ is shił nizhóní, literally ‘with me, it is nice’ (Willie 1991:47). The person who likes is the object of a little word meaning ‘with’, not the subject of the verb. In ‘I know it’, shił béé hózin, both the knower and the thing known are on such words, and the verb itself carries only ho-, a prefix Willie calls a “pleonastic” subject that “refers to an abstract space, area, or entity” (pp. 45, 48). This fits Dowty's account of psych verbs, where the one who feels and the thing that causes the feeling have equal claim to be subject (Dowty 1991:579). It also means the roles a verb needs can sit outside the verb word, in Willie's larger verb-plus-postpositions unit (p. 50). So this site's WORD-EDGE rule is revised: the verb's role schema closes at that unit, not at the word (WORD-EDGE′). What may be added to a sentence depends on the kind of value a slot has: a plain ‘it’ takes a noun (‘the water is deep’), the areal ho- takes only a place phrase (‘deep there in the water’), and the indefinite prefixes take nothing (pp. 35–39, 45–46). Against Grant's thesis: if ho- is a true expletive, a verb can have a slot that is not a role (T17a). Derivations H1–H7; open questions T17a–T17i. Dictionary: ‘be deep’, ‘like / be nice’, ‘know’, ‘be dense’, ‘shave’, ‘paint’, ‘ruin’ and ‘I am walking’ added (92 entries), with new rows on ‘walk (two)’ and ‘run away’.
Added 2026-10-11 (early morning): possessor raising: a noun goes into the verb, the slot count stays at two
Page: Whose neck? Possessor raising. In ‘the woman cut the chicken's head off’ the Navajo verb contains the word for ‘neck’, and the chicken is marked as the verb's object (Willie 1991 Ch. 7, pp. 182–198). Willie says the body part “is not an argument … but serves to derive a complex verb” (p. 187). The verb still has exactly two argument slots, the cutter and the one whose neck is cut. So a lexical item can enter the verb without taking an argument slot, which is a clean case for Grant's “theta roles, not lexical items, are the arguments”. The same string can also be heard two ways: with ‘neck’ inside the verb (‘the woman cut off the chicken's head’) or as a separate noun (‘the chicken cut off the woman's head’), and “the incorporation produces no phonological effects” (pp. 191–192). So the word edge where rule set A gives way to rule set B is structural, not something you can hear. Willie's rule for b-/y- on possessors also answers part of T11d. Derivations Q1–Q6; open questions T16a–T16h. Dictionary: ‘cut off the head’, ‘punch’, ‘eye-punch’, ‘hand-kiss’, ‘belly-kick’, ‘visit (someone's home)’, ‘be long’ and ‘live (plural)’ added (84 entries), with new rows on ‘speak’, ‘see’ and ‘kick’.
Added 2026-10-10 (evening): what the fourth person is for, and when it can carry a noun
Page: One fourth person, five uses. Normally a Navajo noun cannot be attached to a fourth-person verb form: ‘the boy, he cried’ is fine with the ordinary third person and starred with the fourth (Willie 1991:109–111). Willie names three exceptions: respect for one's mother, “I heard …” reports (which also need the word léiʼ ‘a, some’), and the opening of a traditional story (‘Once upon a time there was this coyote walking along …’). She treats respect, reports, stories, impersonal advice (‘one does not step over people’) and polite wishes as uses of one fourth person, whose job is to avoid naming or pointing at anyone (pp. 113, 121–122). That narrows T13a. The wish use breaks a rule this site drew earlier: in ‘My father wishes he could buy the horse’, the fourth-person buyer is the third-person father (p. 121). What links them is the role (the one who wishes is the one who would buy), not a noun and not a matching person. Derivations P1–P6; open questions T15a–T15g. Dictionary: ‘be called’, ‘pick up (plural)’, ‘walk (two)’, ‘walking’ (yiildlosh), ‘sleep’ (optative) and ‘think’ added (76 entries), with new rows on ‘kick’, ‘cry’, ‘buy’, ‘say’, ‘walk’, ‘sing’.
Added 2026-10-10 (afternoon): “the one who …”: relative clauses and the fourth person
Page: The one who: relative clauses, nominalized clauses and the fourth person. A Navajo relative clause is a whole clause with a little word meaning ‘the one’ attached to its end (Willie 1991 Ch. 6). Which person ‘the one’ refers to is not decided by any noun. It is decided by the verb's grammar: with the ordinary prefix yi-, yiztał=yę́ę is ‘the one he kicked’; with the inverse prefix bi-, biztał=yę́ę is ‘the one who kicked him’. When a relative clause is followed by a second verb, both clauses must be about the same two people, as in the “when/if” chains. With ‘believe’, two third persons can be the same person or different people. In all four sentence types studied so far, one fourth person means “someone different” and two mean “the same person”. Derivations R1–R6; open questions T14a–T14f. Dictionary: ‘run away’, ‘believe’, ‘win’, ‘hear’, ‘be fixed’ added (70 entries), and relative-clause rows on ‘kick’, ‘see’, ‘kiss’, ‘laugh’, ‘rope’, ‘buy’, ‘fist-hit’.
Added 2026-10-10: same person or someone else, in “he said …” sentences
Page: Same person or someone else? The fourth person in reported speech and person focus. In a Navajo “he said he will come” sentence with only third persons, the two “he”s have to be different people (Willie 1991:143–144). That is the reverse of the “when/if” chains, where the second clause has to keep the same people. Put the fourth person on both verbs and they are the same person. Put a fourth person on one verb and a third on the other and they are different. So the fourth person does not mean “someone else” or “the same one” by itself. It is the plain third person whose default flips between the two constructions. The fourth person on ‘say’ is either a respect form for a specific person or a “they say” quotative. The two share a form and are kept apart. Also from Willie §5.2: putting a fourth person on the object of ‘see’ makes it mean “I saw her” and never “I saw the event”. Here the person value of a role slot decides what can fill it, with no noun involved. Derivations S1–S6; open questions T13a–T13f. Dictionary: ‘say’, ‘arrive’, ‘carry’ added (65 entries).
Added 2026-10-09 (evening): two clauses, one cast
Page: Two clauses, one cast: =go chains and the fourth person. Take a “when/if” clause marked with =go, followed by a second clause, both with transitive verbs and only third persons. Navajo then requires the second clause to be about the same two people (Willie 1991:135–139). The second verb can change who does what to whom, through yi-/bi-, but it cannot bring in anyone new. The fourth person is the way out, but only from the first clause and only for the doer. The person done-to has to stay the same unless the new one is indefinite. So the second verb maps roles onto a cast that is already fixed, which was then read as the site's “theta roles, not words” claim (since dropped) showing up across clauses. Navajo also turns out to be the reverse of the Greenlandic “fourth person”: it marks different person, not same. Derivations G1–G5; open questions T12a–T12f.
Added 2026-10-09 (afternoon): three things called “fourth person”
Page: Three things called “fourth person”: Navajo, Algonquian, Inuit. The names “fourth person” and “obviation” cover three different devices. In Algonquian (Passamaquoddy, Bruening 2005), obviation ranks two third persons by how high they sit in the clause. Navajo does that job with yi-/bi-, not with its fourth person. The Navajo fourth person (‘that other person’) is a separate person, outside the ranking. The West Greenlandic “fourth person” is the reverse: it must mean the same person as the main-clause subject (Lybech 2020). Bruening also shows that the Passamaquoddy inverse really moves the object above the subject, which is what VPISH predicts for Navajo bi-. That turns the site's open scope test into a concrete prediction: under VPISH, Willie's (58) should be ambiguous. Derivations X1–X4; open questions T11a–T11f.








































