Chapter 90Oblique Incorporation and the Syntactic-vs-Lexicalist Debate

90.1 Oblique incorporation via the applicative

The basic architecture of noun incorporation in Hokkaido Ainu is the person-affix diagnostic: when a nominal stem merges with a verb, the person index moves to precede the entire noun-verb complex, making the incorporation a one-word structure Nakagawa (2024: 482–483); Satō (2022: §2). Object (patient) and patient-like subject incorporation — the overwhelming majority — follow this path directly. The route for oblique participants is different: an instrument, a location, a source, a purpose, or a comitative argument cannot incorporate until an applicative prefix has first promoted it to core-object status Shibatani (1990: 62, 66); Nakagawa (2024: 484); Satō (2022: §1). The further details of how applicatives stack and feed relativization belong to Chapter 82 (Applicative Stacking, Ordering, and the Feeding of Relativization); the present chapter develops the oblique-to-incorporated derivation itself and the theoretical debate over what that derivation implies.

Shibatani (1990: 67) traces the derivation through three steps, illustrated with the verb kuykuy ‘bite’. In the source phrase, the comitative participant pone tura ('bone' + comitative postposition tura 'with') is a peripheral argument: nea cep pone tura a=kuykuy 'I bit the fish with its bones'. The applicative ko- ‘applicative (goal / comitative)’ promotes pone to an additional core object, dropping the postposition and yielding the intermediate nea cep pone a=ko-kuykuy. The newly created object then incorporates: the person prefix a= moves outside the whole complex.

Two properties of the result bear on the theoretical debate in §Chapter 90 (Oblique Incorporation and the Syntactic-vs-Lexicalist Debate). First, the affix ordering is not arbitrary: the incorporated noun sits outside the applicative, which sits inside the verb root — the order applicative-inside / incorporated-noun-outside mirrors the derivational history, consistent with what Baker (1988) terms the Mirror Principle (Baker 1988); Shibatani (1990: 70–71). Second, the entire sequence remains a single verb word, because no person index intervenes between the incorporated noun and the root. Kaiser (1998) examines the applicative–incorporation interaction in Ainu systematically; the constraints she identifies align with the analysis in Shibatani (1990) and Nakagawa (2024: 484).

90.2 Postposition–applicative correspondences

Each postposition or case relation that can be expressed obliquely has a corresponding applicative encoding the same relation as a core object, and only through that applicative can the noun incorporate. Nakagawa (2024: 484) summarises the pairings; the table below assembles them alongside the relevant applicative prefix and an incorporated-form gloss. Nakagawa (2024: §19)

Sources: Nakagawa (2024: 484); Shibatani (1990: 62–66); Satō (2022: §1). The o- ‘applicative (allative, place)’ applicative is restricted to place nouns and is rare in prose Nakagawa (2024: 198–199); Shibatani (1990: 63) notes that the sequence o-tumi-osma, with the applicative, is unambiguously incorporated, whereas bare tumi osma is ambiguous between a phrase and a compound in third-person contexts (see §Chapter 90 (Oblique Incorporation and the Syntactic-vs-Lexicalist Debate)).

A corpus example illustrates the comitative feed in a running text. Nakagawa (2024: 198) cites a Chitose narration in which two coordinate clauses apply ko- to the comitative participants mun ‘grass, weeds’ and toy ‘earth, soil’, incorporating both into the verb tata ‘chop, mince’:

Once the oblique has been applicativized and then incorporated, the resulting verb word is structurally identical to an ordinary object-incorporated compound; the derivational history is preserved only in the affix order (applicative inside the noun) and in meaning. Kaiser (1998) treats the applicative–incorporation interaction as a single system on these grounds.

90.3 Adverb incorporation and stacking limits

Adverbs incorporate directly into the verb without any applicative prerequisite. The mechanism drops the adverbial suffix -no from the basic adverb form and places the adverb stem immediately before the verb root Shibatani (1990: 71–72); Nakagawa (2024: 484, 492). The minimal pair for pirka-no ‘well, finely’ illustrates the contrast:

A noun and an adverb may co-incorporate — Shibatani (1990: 72) cites kiraw-riki-kur-roski 'raise the horns high', with the noun kiraw 'horn' and the adverb riki 'high, up' both inside one verb word. The one-noun-at-a-time limit is the productive norm: Shibatani (1990: 72) states that two independently incorporable nouns cannot ordinarily stack. Nakagawa (2024: 488) records rare exceptions (cep-ya-o-kuta 'dump fish onto the shore', itak-mak-kus-te 'use hidden words'), which he places at the lexicalized fringe of the system rather than as evidence for productive double-NI. The quasi-incorporation strategy discussed in §Chapter 90 (Oblique Incorporation and the Syntactic-vs-Lexicalist Debate) is a separate, structurally different response to the ceiling on NI — it handles cases where NI is grammatically blocked, not cases where noun stacking would otherwise be possible.

90.4 The syntactic-vs-lexicalist debate: positions catalogued

Ainu noun incorporation has attracted theoretical attention because it sits uncomfortably within any single prior account. Shibatani (1988: 210–211) reviewed three frameworks at the 1988 LSJ annual meeting — Relational Grammar, Mithun's semantic account, and Baker's syntactic derivation — finding each insufficient on its own; the 1988 source is a one-page abstract, and the extended discussion appears in Shibatani (1990: 60–73). Subsequent contributions from Satō and Bugaeva have refined the picture further. The positions are catalogued below with evidence grades.

Baker's Case-theoretic prediction affords the clearest empirical test the debate contains. Baker (1988) derives from Case theory that unaccusative clauses cannot undergo applicative or P-incorporation, because the internal argument already fills the available Case position (Baker 1988). Ainu falsifies this directly: applicatives attach freely to unaccusative bases, yielding objects that can then incorporate. Shibatani (1990: 66) cites mokor a=e-wen 'I missed sleeping' and sake a=e-niste 'I am strong at wine', both with e- on an unaccusative base, as the falsifying data Shibatani (1990: 66) ‹contested›. This result disfavours the specific Case-theoretic derivation without settling whether an alternative syntactic mechanism or a lexicalist account better describes the system.

Shibatani's hybrid position requires reference to both the semantic role (low agentivity) and the grammatical position (derived or original object) of the incorporated element. Satō (2008: 226; 2022, §1) revises the semantic-role component: natural-force transitive subjects also incorporate (Type IV, 1.7% of Chitose tokens), so 'Patient' and 'unaccusative subject' are both too narrow. The correct generalisation is the incorporated noun's low degree of agentivity — a property shared by patients, natural forces, and possessed body-parts — and no other currently defined property predicts all four types jointly Satō (2022: §1). Shibatani's grammatical-relation requirement (object or patient-subject, including applicative-derived objects) survives the revision intact.

Community discussion in the aynu-corpora Discord has noted that Ainu has very few pure-noun endocentric or exocentric compounds — constructions built solely from two nouns without a verbal head — with nominal formation concentrated on verb stems (antitwilight, aynu-corpora Discord, June 2026; aynu-corpora Discord (2023–2026)) ‹contested›. Taken at face value, this would contrast Ainu with the kind of free lexical noun-noun compounding that Mithun (1984) treats as the prototype of Type I NI (Mithun 1984), lending support to treating Ainu NI as morphologically integrated into the verbal system. The observation is a community-proposed lead and has not been submitted to systematic quantification across the corpus.

A related terminological divide runs through the valency literature and is relevant to the syntactic-vs-lexicalist question. The prefix i- ‘antipassive / generic object’ is analysed as an antipassive on syntactic, semantic, and discourse grounds by Bugaeva (2021), and labelled accordingly by Nakagawa (2024: 190–192) and Ijäs (Lesson 9) Bugaeva (2021); Ijäs (2023). Satō (2008: 232–233) treats the same prefix as a derivational marker 'people, someone' without invoking the antipassive category ‹contested›. The antipassive analysis presupposes a syntactic operation demoting a grammatical object; the derivational analysis is compatible with a fully lexicalist account of the same data. The origin dispute and fuller analysis are in Chapter 83 (The Antipassive / Detransitive i-).

90.5 Satō's markedness calculus and the NI⊃QI hierarchy

Satō (2016: 88; 2022, §5) assigns a markedness penalty to each NI type by summing restriction factors and subtracting rescue factors. The penalty scores differ between the 2016 article and the 2022 Handbook chapter in their absolute values, but the rank order and the Chitose frequencies — drawn from the late Shirasawa Nabe's corpus, a single narrator — are identical across both sources. The table records both sets of scores; Satō (2022, §5.7) himself describes them as "tentatively defined" Satō (2022: §5); Satō (2016: 84–88).

Restriction factors: Subject NI [+10], Possessor Stranding [+15 in Satō 2022 / +10 in Satō 2016], Ambiguity [+4 in 2022 / +7 in 2016]. Rescue factors: Backgrounding [−5 in 2022 / −8 in 2016], Reflexive interpretation [−14 in 2022 / −9 in 2016]. The absolute values are tentative; the stable result is the rank order and the zero-frequency gap at the bottom Satō (2016: 88–89) ‹corpus-suggested›.

The systematic zero at the foot of the table arises because the inflectional possessor that body-part nouns require makes the incorporated noun definite and thereby removes the reflexive-rescue option. Satō (2022, §4) illustrates the ungrammatical gap with *a=i-par-oyki 'I am fed by others' (lit. 'fed at the mouth'): i= would supply the possessor for par ‘mouth’, but i= on an object creates a definite, non-reflexive reading that blocks the rescue Satō (2022: §4).

This gap is filled by what Satō terms quasi-incorporation (QI) — [O V] units that are tight, idiomatic, and non-compositional in meaning but remain transitive: the person index for the main object sits outside the QI complex as a whole, whereas in true NI the person index precedes the incorporated noun and the result is intransitive Nakagawa (2024: 494–498); Satō (2022: §6). Nakagawa (2024: 494–498) calls the same items 連動詞 (serial verbs / third-class verbs). The Chapter 88 (Object (Patient) Noun Incorporation) chapter lists the QI inventory in full; the example below shows the transitive structure that distinguishes it from NI:

The NI⊃QI accessibility hierarchy that Satō (2016: 88–92; 2022, §6) derives runs: Object NI > Natural-force intransitive subject NI > Possessor-requiring intransitive subject NI > Natural-force transitive subject NI > Possessor-requiring Object NI — with NI covering the left-hand, accessible portion and QI taking over once NI is structurally or semantically blocked. The hierarchy extends Mithun's four-type scale into the QI domain and provides a unified account of why certain noun-verb collocations lexicalize as NI while others lexicalize as QI Satō (2016: 88–92); Satō (2022: §6–7).

A register asymmetry runs across the whole system. Nakagawa (2024: 480) notes that incorporation is commoner in verse than in prose or conversation; Ainu rarely incorporates two nouns into one verb. The verse preference is consistent with observations in Mithun (1984) that NI often develops in rhetorically dense contexts (Mithun 1984), though the Ainu data do not support a diachronic claim about direction of change ‹speculative›.

90.6 Third-person undecidability

The person-affix test that defines incorporation fails for third-person contexts. Hokkaido Ainu marks third-person A and O with a zero morph: a sequence like cep koyki has no surface index to place inside or outside, so the test cannot determine whether the sequence is a phrasal cep koyki or an incorporated cep-koyki Nakagawa (2024: 482); Shibatani (1990: 62). For any incorporable pair with two third-person arguments, lexicographic representation and interlinear glossing must choose between the two analyses without a decisive test.

Two secondary diagnostics are sometimes invoked but neither is conclusive in the general case. Accent and intonation might separate a compound from a phrase, but Nakagawa (2024: 482) states that no firm basis for an accentual test has been established. A derivational prefix on the nominal-verbal sequence (e.g. an applicative over the whole pair) proves one-word status when attested, since derivational morphology cannot straddle two separate words; but the prefixed form and the bare form are distinct constructions and the diagnostic applies only where the prefixed form independently exists. The open-class-verb cases — apto as 'rain fell', upas as, rera as — are formally undecidable in the third person, and Nakagawa (2024: 493–494) leaves their word-boundary status open.

Where the fourth person replaces the third, the test is again decisive. The fourth-person subject suffix =an is licensed only on intransitives; its presence on a noun-verb sequence is unambiguous evidence of incorporation and intransitivization. This diagnostic has been observed in community practice with the pair kinakar=an (incorporated, intransitive) beside kina a=kar (phrasal, transitive with the fourth-person A prefix a=) for the colocation of kina ‘plant material, vegetables’ and kar ‘make, gather’ (thegodofneet, aynu-corpora Discord, December 2024; aynu-corpora Discord (2023–2026)). The observation directly corroborates Nakagawa (2024: 482–483) and Satō (2022, §2) ‹corpus-confirmed›.

The undecidability of third-person cases does not affect the theoretical positions catalogued in §Chapter 90 (Oblique Incorporation and the Syntactic-vs-Lexicalist Debate) — the incorporated forms with first- and fourth-person agreement supply sufficient data for the relevant contrasts — but it does place a ceiling on frequency counts derived from surface strings. The Chitose percentages in §Chapter 90 (Oblique Incorporation and the Syntactic-vs-Lexicalist Debate) are from one narrator; comparable counts for Saru or Horobetsu are not available because the undecidability of third-person tokens in other narrators' texts makes systematic NI frequency comparison across dialects intractable at present Satō (2016: 84) ‹speculative›. The typological comparison — Mithun's four-type hierarchy applied to Ainu — is introduced in Chapter 73 (Valency, Voice, and Noun Incorporation: Overview); subject incorporation and possessor-stranding data are developed in Chapter 89 (Subject Incorporation and Possessor-Stranding); the transitivity diagnostics that ground the person-affix test in the wider grammar belong to Chapter 57 (Transitivity, Valence Classes, and Ambitransitivity).