Navajo direct / inverse (yi- / bi-) — the sourced ranking edge,
What’s claimed and how well supported
- Labels on this page
- SPECULATIVE ×1
- Strongest label used
- SPECULATIVE
- Main caveat (quoted)
- Does not: invent Navajo forms, glosses, or judgments; extrapolate to multiple-wh questions; treat the inverse as an English passive; supply a chemical stand-in; revise Bridge A; claim a joint field.
Counted automatically from the labels printed on this page. Counts are a reading aid, not a grade of the page.
Terms on this page (1), plain-language definitions
- Host (matrix)
- Majority element/phase whose lattice dominates the material.
Definitions quoted from the Chem for linguists primer glossary.
Date: 2026-10-04 PT (evening); Creamer deepen 2026-10-05 morning recovery
Status: Research note. Linguistics half is same-domain description from opened page images. No trees. No chemistry correlate found; none invented.
Audience: Grant (linguistics; less chemistry).
Does not: invent Navajo forms, glosses, or judgments; extrapolate to multiple-wh questions; treat the inverse as an English passive; supply a chemical stand-in; revise Bridge A; claim a joint field.
Why this note. Grant’s brief (2026-10-04): the project is argument ranking. The natural alignment puts the higher proto-agent argument in subject position. The edge is when a lower argument outranks it, so a marked voice (passive, or a Navajo order) is the more natural expression, and some permutations are impossible. Chemistry belongs only if it has an edge of that same shape. The earlier movement note (Derivational movement §2.5) refused a yi-/bi- example because the scans it had were illegible. This note opens clean page images of the alternation and checks it against that edge shape.
What is new (one sentence). Willie & Jelinek 2000 (page images, pp. 268–274) and Hale’s “Conflicting Truths” (page images, pp. 168–169) show that Navajo’s yi-/bi- alternation is a sourced instance of Grant’s edge: the roles stay fixed, a second scale (animacy, plus definiteness and topicality) decides which argument is topical, the marked bi- form is preferred when the patient outranks the agent on that scale, and several alignments are starred outright. Creamer deepen (2026-10-05): Creamer 1974 (opened ERIC Navajo Language Review 1.1 scan) supplies the eight-group status ranking that fills the multi-level animacy table Hale 1973’s note pointed at but that paper itself remains unopened.
1. Sources opened this run
| Source | How read | Used for |
|---|---|---|
| Willie, MaryAnn & Eloise Jelinek 2000, “Navajo as a Discourse Configurational Language,” in Fernald & Platero (eds.), The Athabaskan Languages (Oxford), pp. 252–278 | Scanned pages inside Carnie & Harley (eds.), Pronouns, Presuppositions and Hierarchies: Papers by Eloise Jelinek (manuscript collection, 20 Apr 2012, PDF on Harley’s Arizona site). No text layer; printed pp. 268, 270, 271, 272, 273, 274 read from page images. | Animacy hierarchy (61)–(62); definiteness (51)–(56); wh-questions (57)–(60); one bi- per clause; overrides (65)–(66) |
Hale, Ken, “Conflicting Truths” (MIT scan hale082.pdf, printed pp. 167–169; hostMajority element/phase whose lattice dominates the material. volume not identified from the scan) |
Page images | Direct/inverse pair (2); pro-drop pair (3); parsing pattern (4) |
| Fernald, Theodore, Navajo Language Academy grammar, Chapter 3 “Basic Clause Structure” (Word PDF, 2018 creation date) | Text layer. Navajo strings are font-garbled and not reprinted; English prose only. | Same meaning / different discourse use; inverse only with two third-person arguments |
| Creamer, Mary Helen 1974, “Ranking in Navajo Nouns,” Navajo Language Review / Diné Bizaad Náníl’įįh 1.1, pp. 29–38 (ERIC ED101549 scan; page images rendered 2026-10-05) | Page images (printed pp. 29–38). Native-speaker ranking for yi-/bi- with transitive action verbs; worked with Hale (spring 1971; her note 1–2). | Eight status groups; equal-rank optionality; unequal-rank topic = higher noun; lightning/infant exceptions; Group 7 movement-within-class rule |
Not opened: Hale 1973 (“A note on subject–object inversion in Navajo”) — still no reachable PDF of the Kahane festschrift article; Creamer 1974 (opened above) cites it in her note 2 and elaborates the ranking Hale’s program used. Willie 1991 dissertation (repository page found, PDF not fetched this run); Willie & Jelinek pp. 252–267 and 269 (not read); Perkins 1978; Uyechi 1996; Thompson 1996. An Arizona working-papers PDF that reprints (61)–(62) returned an HTML page instead of a PDF and was not used.
Orthography below is as printed on the page images. Glosses are the sources’ (W&J: 3ACC-3NOM-kicked; Hale: 3y.P.3.kick / 3b.P.3.kick). Bracketed translations and the ?* / * marks are theirs.
2. The alternation, as printed
Hale (2), both arguments third person:
- (2a) Łį́į́’ dzaanééz yiztał. horse mule 3y.P.3.kick — ‘The horse kicked the mule.’ (direct)
- (2b) Dzaanééz łį́į́’ biztał. mule horse 3b.P.3.kick — ‘The horse kicked the mule.’ (inverse)
Hale gives the same English translation for both. (2a) is the basic SOV order; (2b) fronts the logical object (OSV), and the object prefix changes from yi- to bi-.
Hale (3), one overt nominal:
- (3a) Dzaanééz yiztał. — ‘It kicked the mule.’ not: ‘The mule kicked it.’
- (3b) Łį́į́’ biztał. — ‘The horse kicked it.’ not: ‘It kicked the horse.’
Hale (4), his statement of the pattern: in (NP₁) NP₂ X AGR_OBJ V, if AGR_OBJ = yi-, NP₂ = object; if AGR_OBJ = bi-, NP₂ = subject. He calls it a parsing strategy in which NP–NP–V is taken “to involve a certain organization of the arguments and not another,” with “exceedingly strong conditions” on the material X.
Fernald (Ch. 3, prose): the direct/inverse pairs “have the same meaning but have different uses in discourse,” and “the inverse voice exists only for verbs that have two third person arguments; when a first, second, or fourth person argument is present, the verb occurs only in the direct voice.”
Willie & Jelinek (p. 270): “there can be no more than one occurrence of inverse bi- in a clause, since there can be no more than one voice alternation in a clause.” The alternation also applies to oblique (postpositional) objects when all arguments are third person (their (50)). They name the domain of the alternation the Maximal Verb Complex (Willie 1991).
3. What ranks the arguments
3.1 Animacy (W&J §11.4.4, p. 273)
“If the two third-person referents differ along an animacy scale, it is conventional to make the more animate or human referent the topic.”
- (61a) łį́į́’ tsé yiztał horse rock 3ACC-3NOM-kicked (Direct) — ‘The horse kicked the/a rock.’
- (61b) ?* tsé łį́į́’ biztał rock horse 3ACC-3NOM-kicked (Inverse) — [‘the rock was kicked by the/a horse’]
- (62a) ?* łį́į́’ ’ashkii yiztał horse boy 3ACC-3NOM-kicked (Direct) — [‘the horse kicked the/a boy’]
- (62b) ’ashkii łį́į́’ biztał boy horse 3ACC-3NOM-kicked (Inverse) — ‘The boy was kicked by the/a horse.’
Their reading: less animate referents are typically not topical and are more apt to be new information. “Speakers vary in respecting the hierarchy, and its use is declining now under language loss.” Making a less animate or volitional patient topical “is clearly ruled out” when that patient is indefinite or new. They read the inverse as a way of ensuring that agents low on the animacy scale are focused as new information while the high-ranking, definite patient is the topic.
3.2 Definiteness (W&J §11.4.3, p. 271)
In the direct form either argument can be definite or indefinite ((51)–(53)). In the inverse, the topical patient cannot be indefinite:
- (54) ’ashkii ’at’ééd bizts’ǫs — ‘The boy, the girl kissed him.’
- (55) ’ashkii ’at’ééd léi’ bizts’ǫs — ‘The boy, a girl kissed him.’
- (56) * ’ashkii léi’ ’at’ééd bizts’ǫs — [‘A boy, the girl kissed him.’]
3.3 Overrides (W&J p. 274)
The hierarchy “may be set aside” for natural forces with supernatural powers in Navajo culture (lightning, the rainbow; they cite Willie 1991 and Thompson 1996). A first-, second-, or fourth-person possessor can disqualify the sentence from the alternation altogether: (65) shilį́į́’ ’ashkii yiztał ‘My horse kicked the boy.’
3.4 Creamer 1974 — the eight-group status ranking (opened 2026-10-05)
Hale 1973 itself was not opened this run (no reachable PDF). Creamer 1974 is the opened primary that fills the multi-level ranking table the inverse note’s open question asked for. She writes as a native speaker, cross-checked with others, restricted to transitive action yi-/bi- verbs, and notes she worked with Hale in spring 1971; her note 2 cites Hale 1973 in the Kahane volume.
How the ranking decides the form (her general rules, English prose). If the two nouns differ in status, the higher-status noun is the preferred surface topic (initial NP), regardless of deep-structure actor/goal. If the higher noun is the actor → yi- (active). If the higher noun is the receiver → bi- (she calls this the passive form; the syntactic effect is “closely similar” to English passive, but see §5: this note still does not treat Navajo inverse as demotion). If the nouns are equal in status, either may be topic; the speaker’s choice of emphasis selects yi- or bi-.
Qualities that raise status (her list, not a formal metric): (1) capacity for intent or purpose; (2) intelligence; (3) strength, vigor, aggressiveness, or special potency; (4) usefulness to man, or relatedness to man; (5) animation, or capacity for movement. She frames the sum as an inherent “capacity to act upon” relative to other beings. The rankings themselves, she says, were derived from customary usage sentences, not from the rationale first.
Rank groups, highest → lowest (as printed):
| Group | Members (Creamer’s English labels) |
|---|---|
| 1 | Persons (all humans equal for inversion, ignoring ethnicity/sex/rank/age). Exceptions: lightning with its characteristic striking verb = equal to persons; newborn infants lower than other humans (≈ large animals while helpless). |
| 2 | Larger animals and medium animals of special intelligence or relation to man (dog) and predators: horse, donkey, mule, bull, cow, elephant, lion, bear, wolf, wildcat. |
| 3 | Medium-sized animals: sheep, goats, turkeys, eagles, hawks, cats, chickens, deer, antelope, foxes, coyotes. |
| 4 | Small animals: squirrel, gopher, chipmunk, mice, rabbits, songbirds, snakes, frogs, toads, turtles. |
| 5 | Insects, spiders, worms, centipedes, scorpions. |
| 6 | Natural forces: windstorms, flood, sunshine/heat, forest or range fire. |
| 7 | Plants and inanimate objects. Special within-group rule: the noun the verb’s meaning implies as moving outranks the more stationary one. |
| 8 | Abstractions: old age, hunger, disease (or symptoms), “germs,” emotions, or other things seen only by effects. Her closing quip: in Navajo old age cannot kill you; you may only die of it. |
Unequal-rank pattern (her “not preferred” = *). When a higher-group noun acts on a lower-group noun, yi- is preferred and the inverse with the lower noun topicalized is marked not preferred (e.g. bear kills sheep: yi- good, bi- with sheep first *). When a lower-group noun acts on a higher-group noun, bi- with the higher noun first is preferred and the yi- form is * (e.g. dog bites boy: Ashkii łééchąąʼí bishxash preferred; Łééchąąʼí ashkii yishxash *). Equal-rank pairs allow both. Groups more than one step apart follow the same direction. Occasional personification of Group 7 (and some 2/3) in legendary writing is noted as an exception track, not expanded here.
What this adds to §3.1–3.3. W&J gave a coarse animacy contrast (human > animal > inanimate) plus definiteness and topicality. Creamer supplies the fine grain of that hierarchy as an eight-step status scale with named cutoffs, equal-rank optionality, and categorical “not preferred” cells when the scales disagree — the same edge shape Grant asked for, now with the multi-level table filled from an opened primary. Still not a passive demotion; still no invented forms; Hale 1973’s own prose and any table numbering in that paper remain unread.
Orthography caveat. Navajo strings above that carry diacritics were checked against the rendered page images of the ERIC scan (pp. 33–35). Strings whose diacritics were illegible on the scan are not reprinted; English glosses and the group labels are.
4. Wh-questions with the inverse (W&J pp. 272–273)
This is the only sourced wh material, and it is single-wh with at most one other nominal.
- (57a) háí-sh yizts’ǫs (Direct) — ‘Who did he kiss?’
- (57b) háí-sh bizts’ǫs (Inverse) — ‘Who kissed him?’
- (58b) háí-sh Jáan yizts’ǫs (Direct) — ambiguous: ‘Who was it, that kissed John?’ or ‘Who was it, that John kissed?’
- (59) háí-sh Jáan bizts’ǫs (Inverse) — ‘Which one (of you) was it, that was kissed by John?’
- (60a) Jáan háí-sh bizts’ǫs (Inverse) — ‘As for John, he was kissed by whom?’
- (60b) háí-sh Jáan bizts’ǫs (Inverse) — 1) ‘Which one (of you) was kissed by John?’ 2) *‘John was kissed by whom/which one?’
Their claims: Navajo has no obligatory wh-movement, but the wh-word may undergo leftward focus movement; -sh carries focus. A wh-word coindexed with bi- gets a presuppositional “which one” (D-linked) reading. “The inverse does not permit optional focus movement of the Wh-word.” “The bi-pronoun marks a topic/focus inversion that is incompatible with focus movement of the Wh-word.”
Grant’s multiple-wh test stays open. His proposed comparison (“who bet who what at what casino” against passive-like permutations) needs native intuitions and his stored texts. Nothing here ranks multiple-wh orders. Two sourced facts bear on how that test could be read, without predicting its outcome: (a) only one bi- per clause, so a clause has at most one voice alternation; (b) the inverse blocks focus movement of a wh-word that would otherwise front.
5. Against Grant’s edge shape
| Edge property (Grant, 2026-10-04) | Navajo yi-/bi-, as sourced | Verdict |
|---|---|---|
| Natural alignment: higher proto-agent argument is the subject | Direct yi- with SOV is “the basic transitive order” (Hale); the agent is the subject | Match |
| Roles do not change under the marked alignment | Hale and Fernald: same meaning; Hale translates (2a) and (2b) identically | Match. Fits Grant’s own correction that discourse does not change proto-role counts. |
| A lower argument outranks the higher one | In (62) the patient (boy) outranks the agent (horse) on animacy, not on proto-agent entailments | Match, with the scale named: animacy / topicality / definiteness |
| The marked alignment is then more natural | (62b) inverse is the good form; (62a) direct is ?* | Match (gradient: speaker variation; declining under language loss) |
| Some permutations are impossible | (56) *; (60b) reading 2 *; no more than one bi- per clause; inverse only with two third-person arguments | Match (categorical bans) |
| The marked form is a passive | Not a passive. Both arguments stay as the verb’s pronominal arguments; W&J call it a voice contrast; their English translations use passives only as glosses | Differs from English passive: no demotion |
What the ranked units are. On W&J’s pronominal-argument analysis, the verb’s person prefixes are the arguments and overt nominals are adjuncts ordered by topic and focus. bi- records which argument pronoun is linked to the topic. So what the ranking realigns is the link between verb-internal, role-bearing pronouns and the topic/focus positions. The lexical nouns do not change roles. This is the closest sourced statement found so far of Grant’s idea that in Navajo the units in play are theta roles rather than lexical items. W&J do not use that wording, and this note does not say theta roles move.
Tie to the lexicalization line (one sentence, no merger). The alternation sits in the object-prefix slot of the same verb word whose listed base and theme the lexicalization page describes; the sources state it for transitives with two third-person arguments generally, not theme by theme, so nothing here says the inverse is listed.
6. Chemistry side — open, not invented
No chemistry source opened this run, and no chemical correlate is claimed. The earlier scout (passivization-edge-chemistry-scout-v0.md (not published here)) tested inverse electron demand and it failed, chiefly because there is no second voice for the same two molecules.
What the Navajo case adds is a sharper search specification, taken from the sourced shape above. A chemical edge of the same shape would need all of these:
- One unchanged event with two participants whose roles are fixed.
- A second ranking scale that is independent of those roles (as animacy is independent of agent/patient).
- When the two scales disagree, a marked alignment is preferred and the default alignment degrades.
- Some alignments banned outright, not just disfavored.
- Both participants staying inside the unit under the marked alignment (no demotion).
Grant’s Zipf frame (phrase ↔ molecule as the shared level; HIS, SPECULATIVE, not a joint field) stays the way to look. Nothing in this note is a Zipf result.
7. Open questions
- Read W&J pp. 252–267 and Willie 1991 for the full statement of the Maximal Verb Complex and the obviative fourth person.
- Hale 1973 (“A note on subject–object inversion in Navajo”) still has no reachable PDF; Creamer 1974 (opened 2026-10-05, §3.4) now supplies the eight-group status ranking that fills that table from a primary in Hale’s program. Opening Hale 1973 itself remains desirable for his own wording and any table numbering, but is not blocking.
- Grant’s multiple-wh / casino test: needs his texts and native intuitions. Do not pre-fill.
- Chemistry: search against the five-point specification in §6, within the phrase ↔ molecule level. A source that does not already state the correspondence is not a reason to stop.
Pointer (2026-10-05 early): open question 4 is taken up in Spinel inversion ranking edge (spinel cation inversion passes points 2, 3 and 5 of the §6 spec, partial on 1 and 4, crystal level not molecule; candidate, not correlate).
Pointer (2026-10-05 midday): open question 1 is taken up in Fourth person vs ranking (Willie 1991 §4.2.7 and W&J 2000 pp. 252–267, 270, 277–278 opened: the fourth person is outside the animacy hierarchy and lets the direct form survive a low-agent / high-patient clash instead of the inverse — an escape from the scale, not a second inverse; spec point 6 added there).
Pointer (2026-10-05 late evening): open question 4 at the molecule level is taken up in TBP apicophilicity ranking edge (five-coordinate phosphorus substitution: fixed entering/leaving jobs, an independent apicophilicity scale, a marked pseudorotation step visible in the product, connectivity bans; the marked step restores the role slot rather than letting the second scale win; candidate, not correlate).
Corpus note: navajo-inverse-ranking-edge-v0.md. Forms and judgments are as printed on the Willie and Jelinek 2000, Hale, and Creamer 1974 page images. The yi-/bi- alternation is a same-domain instance of the ranking edge. Creamer supplies the eight-group status ranking. No chemical correlate is claimed or invented. Not Bridge A. Not a joint field.