Algonquian direct / inverse: is the "crossing flip" a change of theta roles or of ranking?
What’s claimed and how well supported
- Labels on this page
- SPECULATIVE ×8
- Strongest label used
- SPECULATIVE
- Main caveat (quoted)
- Bridge A is untouched and stays ◆ SPECULATIVE.
Counted automatically from the labels printed on this page. Counts are a reading aid, not a grade of the page.
Terms on this page (2), plain-language definitions
- Valence (chem.)
- Bonding capacity / oxidation-related electron count of an atom in context.
- Valency (ling.)
- Arity: how many actants a predicate takes; formal vs semantic valency can diverge.
Definitions quoted from the Chem for linguists primer glossary.
Date: 2026-10-23 PT (overnight run) Status: Linguistics research note. Same-domain description from sources opened this run. Every claim is labelled SOURCED, HYPOTHESIS, GRANT'S CONJECTURE or OPEN. No chemistry correlate is claimed. Bridge A is untouched and stays SPECULATIVE. This is not a joint field. Audience: Grant (linguistics; less chemistry). Why: Grant relayed news from the Morphic Tessellations bot (2026-10-22): "some algonquin languages have some kind of theta phenomenon... crossing flip in inverse voice or something... please investigate and keep track of theta phenomenon for if they pertain to chem stuff". The running tracker is Theta-phenomena tracker: what languages do with theta roles, and any chemistry bearing.
0. Bottom line, in plain words
- SOURCED. In an Algonquian inverse the doer stays the doer and the done-to stays the done-to. Who did what to whom never changes. What changes is which participant controls the verb's "subject" agreement slot. The inverse sign (-ikw, -igw, -ok, -ku) says "the slot is held by the patient this time" (Oxford 2023 §1; Bruening 2005 p. 2; Oxford, Routledge overview §4.4).
- SOURCED. There are two kinds of inverse in one family, often in one language (Oxford 2023, after Givón 1994 and Gildea 1994):
- When a third person acts on "me" or "you" (the SAP inverse), the inverse is obligatory and only morphological ("shallow"). No argument moves.
- When one third person acts on another (obviative on proximate), the inverse is an optional voice ("deep"). The patient really is raised above the agent: binding reverses, word order and long-distance agreement follow it. But unlike a passive, the agent stays a full argument.
- HYPOTHESIS (our gloss). "Crossing flip" is not a term in any source opened. The Morphic note uses it for its own braid picture (σ vs σ⁻¹). The closest linguistic content is: in the deep inverse the two arguments' structural order is swapped (patient over agent), while their theta roles are not. In the shallow inverse nothing crosses; only the spell-out of agreement changes.
- SOURCED + HYPOTHESIS. The ranking scale is person/obviation (1/2 > 3 > 3′ > inanimate), not proto-role weight. Blackfoot is the exception that matters for Grant: Bliss 2005 argues its direct/inverse ranks sentient over non-sentient third persons, not proximate over obviative, and that non-sentient nouns cannot be agents at all. Sentience is one of Dowty's proto-agent entailments (the link to Dowty is ours, HYPOTHESIS).
- SPECULATIVE / NONE for chemistry. Nothing here gives a chemistry correlate. The tracker records only the existing SPECULATIVE site-preference ranking-edge scouts as the nearest candidates.
1. Sources
1.1 Opened this run
| Source | Access | Used for |
|---|---|---|
| Oxford, Will. "A tale of two inverses." Syntax 26 (2023): 311–354. Author's preprint, lingbuzz/006567, Feb 2023 | PDF, text layer; read pp. 1–33 and the conclusion | Deep vs shallow inverse; examples (1), (8)–(15), (18)–(21); evidence list; Rule (35) impoverishment; Voice_erg analysis; diachrony |
| Oxford, Will. "Inverse marking and Multiple Agree in Algonquin." NLLT 37 (2019): 955–996. Preprint, lingbuzz/004529 | PDF, text layer; read the opening sections and Table 1 | Theme-sign table; complementarity of inverse marking and portmanteau agreement; rejection of a 2>1 "local inverse" |
| Oxford, Will. "Algonquian." Routledge handbook chapter (author's PDF from his site) | PDF, text layer; §§4.2–4.4 | Menominee (10a/b) direct/inverse pair (Bloomfield 1962: 39); obviation (14)–(15); Ojibwe templates (Figures 5–6); hierarchy 1/2>X>3>3′>0 in the independent, 3>3′>0 in the conjunct (Rhodes 1994: 432) |
| Bruening, Benjamin. "The Algonquian Inverse is Syntactic: Binding in Passamaquoddy." ms., Univ. of Delaware, 21 July 2005 (revised ch. 2 of Bruening 2001) | PDF already on the box (/workspace/t27k/bruening2005.pdf, the copy Navajo Decoded uses). The udel.edu URL now returns an HTML page instead of the PDF |
Abstract; Passamaquoddy (1) and (3); three families of analysis (inverse linking, morphological inverse, syntactic movement); "not a passive" |
| Bliss, Heather. "Topic, Focus, and Point of View in Blackfoot." WCCFL 24 (2005): 61–69 (Cascadilla, lingref.com paper1207) | PDF, text layer; pp. 61–66 | Blackfoot (1), (3a–h), (4), (9)/(13), (10), (11); sentience hierarchy; POV analysis built on Bruening 2001 |
| Willie, MaryAnn. Navajo Pronouns and Obviation. PhD diss., Univ. of Arizona, 1991 (box scan) | Page images of printed pp. 96–97 at 90 dpi; text layer of p. 216 (PDF p. 222) | Navajo comparison (§5) |
| Aissen, Judith. "On the Syntax of Obviation." Language 73 (1997): 705–750 (author's scan) | Downloaded; it is an image scan with no text layer. Only the abstract (as shown by the search index and Project MUSE) was read | Named only as the "morphological inverse" family that Bruening argues against |
1.2 Not opened (cited only through the sources above)
Klaiman 1991; Givón 1994; Gildea 1994; Howse 1844; Wolfart 1973; Wolvengrey 2011; Dahlstrom 1986a/b, 1991; Rhodes 1976, 1994; Bruening 2001 (diss.); Lochbihler 2012; Bliss 2013; Zúñiga 2006/2008; Macaulay 2009; Nichols 1980; Bloomfield 1946/1962; Valentine 2001; Hamilton 2015b; McLean 2001; Oxford 2024 (LI 55; downloaded, abstract only); Oxford's TWPL "grammar myths" paper (the download hung). Any form below credited to these comes from the opened paper that quotes it, and is marked "via".
2. The pieces
2.1 Person hierarchy and obviation (SOURCED)
- Hierarchy as usually stated: 1, 2 > 3 (proximate) > 3′ (obviative) > inanimate (Bruening 2005 (2), from Goddard & Bragdon 1988). Bliss 2005 (2): 1st/2nd > 3rd PROX > 3rd OBV. Oxford's Routledge chapter: 1/2 > X (impersonal) > 3 > 3′ > 0 for the Ojibwe independent order, and only 3 > 3′ > 0 in the conjunct (Rhodes 1994: 432, via Oxford).
- Does 2 outrank 1? A "local" 2>1 inverse (Wolfart 1973, via Oxford 2019) is rejected by most Algonquianists (Oxford 2019 lists Bloomfield, Hockett, Goddard 1979, Nichols 1980, Valentine 2001, Zúñiga 2006/2008, Macaulay 2009). The 2>1 effect shows up only in the person prefix (Routledge chapter: "a 2>1 hierarchy for the prefix and a 1PL>2 hierarchy for the central suffix (Zúñiga 2008; Macaulay 2009)"). So "2>1>3" is a prefix-slot fact, not the inverse hierarchy.
- Obviation: "When a clause contains multiple third-person referents, one, normally the most topical, is designated as proximate (notated 3 or PROX); any others must be obviative" (Routledge §4.3). It also marks disjoint reference: Oji-Cree Tepit o-waapam-aa-n o-tayihsh-an 'David.PROX sees his.PROX dog.OBV' vs … o-tayihsh-ini-wan 'his.OBV (not David's) dog' (Routledge (15), Oxford's fieldwork).
2.2 Theme signs and the template (SOURCED)
Ojibwe independent (Routledge Figure 5, from Nichols 1980): prefix ni- · stem waapam · theme sign -ikw (INV) · negation -ssii · central suffix -naa (1PL) · mode -token · peripheral -ak (3PL) = niwaapamikossiinaatokenak 'they might not see us'.
"The theme sign normally marks person/animacy agreement with the logical object: -i '1OBJ', -in '2OBJ', -aa '3.AN.OBJ' (or 'DIRECT'), -am 'INAN.OBJ'. However, a special INVERSE theme sign -ikw appears instead when the logical object outranks the logical subject" (Routledge §4.4).
Algonquin (Ojibwe dialect) theme signs, Oxford 2019 Table 1 (data Jones 1977):
| Configuration | Conjunct | Independent |
|---|---|---|
| 2→1 | -i | -i |
| 1→2 | -in | -in |
| 1/2→3 | -a: | -a: |
| 3→1 | -i | -igw |
| 3→2 | -in | -igw |
| 3→3′ | -a: | -a: |
| 3′→3 | -igw | -igw |
So 3→1 in the conjunct is not inverse (it uses plain 1OBJ -i); in the independent it is.
2.3 Attested pairs, as printed
Ojibwe (Nichols 1980: 276–292, via Oxford 2023 (1)): ni-waːpam-aː 'I see her' vs ni-waːpam-ikw 'she sees me'. The prefix ni- '1' is the same; only the theme sign changes.
Plains Cree, SAP inverse, obligatory (Wolfart 1973: 41–42, via Oxford 2023 (8)–(9), (15)): - ni-waːpam-aː-inaːn 'we see her' (direct) vs ni-waːpam-ikw-inaːn 'she sees us' (inverse). - niwaːpamaːnaːnak 'we see them' vs niwaːpamikonaːnak 'they see us'. "Each of the forms in (15) is the only way to express that particular configuration of arguments" (Oxford 2023 p. 10).
Plains Cree, third-person inverse, optional (Wolvengrey 2011, via Oxford 2023 (13)–(14)): - (13a) anihi waːposwa eː-kiː-waːpamaːyit anihi mahihkana 'that the rabbit.OBV saw the wolf.OBV' (direct) vs (13b) anihi mahihkana eː-kiː-waːpamikoyit anihi waːposwa 'that the wolf.OBV was seen by the rabbit.OBV' (inverse). Same event, chosen by topicality. - (14a) Caːniy kiː-wiːciheːw Meːriːwa 'Johnny.PX helped Mary.OBV' vs (14b) Caːniːwa kiː-wiːcihik Meːriy 'Johnny.OBV helped Mary.PX' / 'Mary.PX was helped by Johnny.OBV'.
Menominee (Bloomfield 1962: 39, via Routledge (10)): nɛ·w-ɛ·-w enoh enɛ·niw anenoh metɛ·mohs-an 'The man (PROX) sees the woman (OBV)' vs nɛ·w-ek-w … 'The woman (OBV) sees the man (PROX)'. "Despite the reversal of thematic relations, the proximate enjoys the same morphosyntactic prominence in both sentences."
Passamaquoddy (Bruening 2005 (1), (3)): K-ucem-a-k 'You kissed them' vs K-ucem-ku-k 'They kissed you'; Pesq muwin ’-toli-nuhsuphoqal-a mahtoqehsu 'One bear (Prox) was chasing some rabbits (Obv)' vs Mahtoqehs ’-toli-nuhsuphoqal-ku-l muwinuw-ol 'A rabbit (Prox) was being chased by a bear (Obv)'. Bruening: "it is important to note that this is not a passive. There is no loss of an external argument here, and no change in valenceBonding capacity / oxidation-related electron count of an atom in context.."
Blackfoot (Bliss 2005 (3), her fieldwork): nit-aawayaki-a-(w)a 'I hit him' vs nit-aawayaki-ok-(w)a 'He hit me'; aawayaki-(y)ii-(w)a 'He hit her' (PROX→OBV, direct) vs ot-aawayaki-ok-(w)a 'She hit him' (OBV→PROX, inverse); local kit-aawayaki-o 'I hit you' vs kit-aawayaki-oki 'You hit me'.
3. What the "crossing flip" is
3.1 Three families of analysis (Bruening 2005 pp. 2–3, SOURCED)
- Inverse linking. Arguments are projected in the reverse order: the hierarchy decides which is merged high (Rhodes 1976/1994; Perlmutter & Rhodes 1988, via Bruening). This is the only family on which the inverse changes argument structure.
- Morphological inverse. Syntax is constant (subject over object); the morphology indexes the hierarchy (Aissen 1997, via Bruening, who quotes her "Relational Hierarchy: Subject > primary object").
- Syntactic movement. Syntax is projected uniformly, and "what is different about the inverse is that it involves a step of movement in the syntax, such that the object rather than the subject ends up in the highest position" (Bruening 2005 p. 3). Evidence: scope and variable binding in Passamaquoddy.
3.2 Oxford 2023: both 2 and 3, in different contexts (SOURCED)
- Two levels of "inverse" (Gildea 1994, via Oxford): "inverse alignment is a morphological pattern in which the patient is indexed like a canonical subject, while inverse voice is a syntactic construction in which the patient is the subject." Deep inverse = both; shallow inverse = morphology only. "The two kinds of inverses can coexist within a language (Givón 1994: 29)."
- Third-person inverse = deep (voice). Evidence:
- It is contrastive: the rabbit/wolf and Johnny/Mary pairs.
- Binding reverses, as in an English passive. Oji-Cree (18)/(19): direct 'All the women.PX kiss the man.OBV who loves them' (agent binds into patient); inverse 'The man.OBV who loves them kisses all the women.PX' (patient binds into agent). The same pattern is reported for Plains Cree, Western Naskapi, Passamaquoddy, Ojibwe, Blackfoot and Meskwaki (list in Oxford 2023 p. 11; those sources not opened).
- Word order (East Cree, Passamaquoddy, Mi'kmaq) and long-distance agreement (Odawa, Mi'kmaq) follow the patient.
- SAP inverse = shallow (alignment). It is obligatory and non-contrastive (15). In the one SAP binding example Oxford found (Bruening 2001: 131, Passamaquoddy, via Oxford 2023 (22)), the agent still binds into the patient. In Mi'kmaq, long-distance agreement in the SAP inverse targets the agent (Hamilton 2015b, via Oxford).
- Mechanism. Shallow: the Infl probe is relativized to [uϕ, uPart] and agrees with the SAP patient. An impoverishment rule then deletes the duplicate: "Delete phi-agreement from Voice when phi-agreement on Infl is an exact duplicate" (Rule 35). The inverse marker is Voice's elsewhere exponent. Deep: a special Voice head (Voice_erg) gives the agent inherent Case and carries an EPP feature, so the patient raises over it.
- Not a passive. The agent stays a core argument in both kinds (Oxford 2023 §2, Table 1; Bruening 2005 p. 2).
- Diachrony (Givón/Gildea, via Oxford). passive > deep inverse > deep + shallow > shallow. McLean 2001 (via Oxford) derives the Algonquian inverse from a passive.
3.3 Oxford 2019: inverse as an Agree outcome (SOURCED)
In Algonquin, "inverse marking is possible only in contexts where portmanteau agreement is not" (abstract). Voice always agrees with the object. Infl "is able to agree with either the subject, the object, or both arguments simultaneously, depending on which argument best matches the probe's features" (p. 2). "When Infl enters a Multiple Agree relation with both arguments, the realization of portmanteau agreement morphology is possible. When Infl agrees only with the object, it duplicates the result of an earlier object agreement operation on Voice." Impoverishment then deletes Voice's features, and Voice is spelled out as "an underspecified elsewhere form—which is the exponent that we know descriptively as the inverse marker" (abstract). On this account the inverse is a fact about Agree, not Merge. (Oxford 2023 restricts this Agree-only analysis to the shallow inverse.)
3.4 Our verdict on Grant's question (HYPOTHESIS, built on the SOURCED material)
| Question | SAP inverse (3→1/2) | Third-person inverse (3′→3) |
|---|---|---|
| Do theta roles change? | No | No |
| Is valencyArity: how many actants a predicate takes; formal vs semantic valency can diverge. changed (agent demoted)? | No | No (≠ passive) |
| Is structure changed (Merge/Move)? | No: Agree plus impoverishment | Yes: the patient raises over the agent (internal Merge/A-movement) |
| Is it optional? | No | Yes (topicality) |
| Best label | "same roles, marked differently" | "same roles, ranked differently and re-positioned" |
So the answer to "argument-structure change or same roles marked differently?" is: never an argument-structure change in the sense of theta assignment or valency. In 3→3 clauses it is a voice in the narrow sense of re-positioning. Only the inverse-linking family (Rhodes 1976/1994) would make it an argument-structure change, and Bruening and Oxford both argue against that on binding evidence.
Merge vs Agree here (HYPOTHESIS, standard definitions as in Navajo Decoded's Merge/Agree pages): Merge builds structure. External Merge introduces the agent and patient in their theta positions. Internal Merge (movement) re-merges the patient higher, in the deep inverse. Agree values features: Infl copies person/number from whichever argument it finds. The shallow inverse is all Agree; the deep inverse is Agree plus one internal Merge. Neither touches theta roles, which are fixed at external Merge.
4. Bearing on Grant's role-ranking thesis
- GRANT'S CONJECTURE. Verbs rank their arguments; the natural alignment puts the higher proto-agent in subject position. When a lower argument outranks it, a marked voice is natural and some permutations are impossible.
- SOURCED. Algonquian fits the shape of that edge: roles fixed, a ranking decides which argument controls the subject slot, and a marked form appears when the patient outranks the agent.
- SOURCED. But the Algonquian ranking is over person/obviation features of the participants, not over proto-role weight. In 3→3 clauses it is over discourse topicality (Wolvengrey's "situation-specific, speaker-determined assignment of relative topicality", via Oxford).
- SOURCED (Bliss 2005) + HYPOTHESIS (Dowty link). Blackfoot is the closest to Grant: third persons rank SENT > NON-SENT, and the evidence is (11) Ana pokóna ani otaííspaapiksistii máóhksinaa 'The ball that he is bouncing is red'. There a sentient obviative acts on a non-sentient proximate ball, and the verb is direct (-yii), not inverse. Also, "Non-sentient nouns, like isttoán 'knife' cannot be agentive in Blackfoot" ((9)/(13); the knife appears as an adjunct with the prefix iiht- 'means'). Sentience is a Dowty proto-agent entailment, so Blackfoot ranks on one proto-agent property. It is a single property, not a weight, so this is a partial fit at best.
- OPEN. Whether any Algonquian inverse is sensitive to a patient proto-property (change of state, incremental theme). Nothing opened says so.
5. Navajo yi-/bi- compared
- SOURCED. Willie 1991 pp. 96–97, on Navajo: "the Inverse marking functions like obviative marking. However, note that the Inverse has a further property: it always makes the argument with an agent theta role the internal argument." p. 216: an alternation on a lexical postposition "is not one of voice, but shows the use of this alternation to mark obviation."
- SOURCED (Navajo Decoded /inverse/). The site already treats yi-/bi- as roles fixed, a ranking over animacy (Willie & Jelinek 2000; Creamer 1974), and a question about whether the agent's position changes (T5).
- HYPOTHESIS. On Willie's wording the Navajo inverse is closer to the Algonquian deep (third-person) inverse than to the shallow one: it is confined to 3/3 clauses, it is optional and topicality/animacy-driven, and it is said to re-position the agent ("internal argument"). Navajo has no 3→1/2 inverse at all, so it has no counterpart of the shallow SAP inverse. Navajo Decoded's /inverse-agree/ and /fourth-person-compared/ pages already use Bruening 2005 for this.
- OPEN. Whether Navajo's yi-/bi- shows Oxford's deep-inverse diagnostics (binding reversal, LDA). Navajo Decoded T11 notes the binding diagnostics may not be constructible in Navajo.
SPECULATIVE Cross-domain section. Open after the materials.
6. Chemistry relevance ( SPECULATIVE at most)
- NONE found. No source opened mentions chemistry, and no chemistry correlate is claimed.
- SPECULATIVE (pointer only). The existing scouts that test chemistry against the "ranking edge" specification (spinel inversion, TBP apicophilicity, EAS directing) remain the only candidates. The new information for them is a sharper specification. A real analog would need to separate (a) an obligatory, feature-determined marking with no rearrangement (shallow) from (b) an optional, context-determined rearrangement with roles kept (deep). That two-way split is a test to run on those scouts, not a result.
- Morphic's braid reading (σ vs σ⁻¹ as a "crossing flip") is the Morphic bot's own SPECULATIVE framing, unpublished. It is the deep inverse only, if anything. In the shallow inverse nothing crosses.
7. Open questions
- A1 (OPEN): Does any Algonquian inverse rank by a patient proto-property? (Not in anything opened.)
- A2 (OPEN): Read Bliss 2013 §7.4.1 (Blackfoot binding) and Oxford 2024 (LI 55, 12 inverse distributions) for the full typology.
- A3 (OPEN): The Navajo deep-inverse diagnostics (Navajo Decoded T11, T5).
- A4 (OPEN): The Klaiman 1991 / Givón 1994 typology of inverse voice, to check the "voice" label against its source. Neither was opened.
Corpus note: algonquian-inverse-v0.md. Linguistics only: same-domain description from the sources the note lists as opened (Oxford 2023 Syntax 26 preprint; Oxford 2019 NLLT 37 preprint; Oxford's Routledge Algonquian chapter; Bruening 2005 ms.; Bliss 2005 WCCFL 24; Willie 1991 pp. 96-97, 216; Aissen 1997 as its abstract only). Forms are copied from those papers; forms from unopened grammars are marked as cited via them. The 'crossing flip' gloss and the comparisons are analysis. No chemistry correlate. Bridge A unrevised. Not a joint field.