Chapter 83The Antipassive / Detransitive i-

83.1 Form and function in the arity calculus

The prefix i- ‘antipassive / detransitivizing prefix’ occupies the object position of a transitive verb with a semantically generic 'thing, someone, people', reducing the verb's arity by one and rendering it intransitive. In Nakagawa's arity calculus the operation is a standard −1 step: a two-place transitive yields a one-place intransitive, and a three-place ditransitive yields a two-place transitive Nakagawa (2024: 190–192); Bugaeva & Kobayashi (2022: §4.2.3). The resulting intransitive takes suffixal S indexes — =as (1PL.EXCL.S) and =an (4.S) — where the base transitive had taken prefixal A indexes such as ci= and a=. This person-affix shift is the single operative diagnostic for the detransitivization Satō (2008: 96); Nakagawa (2024: 137, 140). The place of the operation in the wider valency system is set out in Chapter 73 (Valency, Voice, and Noun Incorporation: Overview).

The prefix sits immediately before the verb root, inside the person-index layer. Under Nakagawa's position-class template it occupies the same slot as the other valency-decreasing prefixes yay-, si-, and u-, and it combines freely with applicatives outside it and causative suffixes after the stem: ko-i-ruska (APPL-ANTIP-be.angry at), a re-transitivized applicative of the antipassive i-ruska 'be angry', is well-formed, and so is the further reflexivized si-ko-i-ruska Bugaeva & Kobayashi (2022: §4.2.3). The derivational chain ruska (vt) → i-ruska (vi) → ko-i-ruska (vt) → si-ko-i-ruska (vi) → i-ruska-re (vt, 'make sb angry') → ko-i-ruska-re (vd) shows that i- is not a terminal derivation but a mid-chain element and that the arity bookkeeping remains coherent at every step Bugaeva & Kobayashi (2022: §4.2.3). In the corpus, iruska 'be angry' appears 655 times (rank 264), making it among the most frequent antipassive derivatives ‹corpus-confirmed›.

83.2 Semantic domains and lexicalization

The antipassive is productive across a restricted range of semantic classes. Bugaeva, working from Nakagawa's corpus, identifies 48 verbal lexemes that form antipassive derivatives, concentrated in Perception/Cognition, Ingesting, Interaction, Communication, Grooming, and Traditional-Activity verbs Bugaeva & Kobayashi (2022: §4.2.3). Absent from the set are Creation, Transformation, and Contact-by-Impact verbs — a distribution consistent with the typological generalization that antipassives in non-ergative languages preferentially target low-transitivity predicates ‹contested›.

Many antipassive derivatives have lexicalized to a fixed, conventionalized object reading. i-ku no longer simply means 'drink something' but specifically 'drink alcohol' — a euphemistic narrowing noted by Nakagawa as 婉曲 (indirect expression) — while i-omante and i-carpa are restricted to ritual contexts Nakagawa (2024: 191–192); Bugaeva & Kobayashi (2022: §4.2.3). At the extreme of lexicalization the morphological boundary is no longer transparent: inkar 'gaze' (< i-nukar) and ipe 'food; eat a meal' (< i-e) have undergone phonological fusion and are synchronically nominal or verbal roots in their own right Nakagawa (2024: 192). Bugaeva and Kobayashi further report that i- is fossilized residually on the adverb i-mak 'behind' and the particle i-y-os(no) 'afterward, later', where the original Patient-incorporating reading is entirely opaque Bugaeva & Kobayashi (2022: §4.2.3).

The discourse motivation for the antipassive parallels what Bugaeva identifies typologically: i- is preferred when discourse focus falls on the Agent or the activity as a whole rather than on a specific, individuated Patient Bugaeva & Kobayashi (2022: §4.2.3). Bugaeva also notes that the prefix can antipassivize a Recipient or a direct Causee, as in i-newsar-ka 'amuse people' and i-oripak-ka 'make people feel humble' — an antipassivization of the causee rare across languages Bugaeva & Kobayashi (2022: §4.2.3).

83.3 The label debate: antipassive or derivational prefix?

The functional description is uncontroversial; the grammatical label is not. The traditional analysis, associated with Tamura and reported through Bugaeva and Kobayashi, treats i- as a generalised object marker — a derivational prefix introducing an unspecified object 'someone, thing' — without applying the voice category antipassive Bugaeva & Kobayashi (2022: §4.2.3). Satō's textbook grammar follows this line, describing the identical prefix as derivational i- '人を' (people-as-object) and listing no antipassive in its voice inventory Satō (2008: 232–233). On this account i- is a lexical derivational element that selects a generic reading rather than a voice morpheme that restructures argument relations.

Bugaeva's antipassive reanalysis argues from three converging criteria that i- meets the typological definition of an antipassive: it systematically demotes the Patient/Theme/Recipient from a core object position; its output always carries an unspecified or generic object reading; and it is discourse-conditioned, appearing precisely where the speaker defocuses the Patient to foreground the Agent or the activity Bugaeva (2021: 232–237); Bugaeva & Kobayashi (2022: §4.2.3). Nakagawa endorses this reanalysis, noting that the term 逆受動態 (antipassive) has become standard in recent work following Bugaeva, and adopts it while acknowledging the derivational tradition Nakagawa (2024: 192). Ijäs's learner grammar also uses the antipassive label; her characterization — the prefix 'takes the place of the object and fills it with a generic meaning of thing/place' — renders Bugaeva's substance accurately Ijäs (2023) ‹contested›.

The two positions share substantial common ground: both agree that the prefix occupies the object position, that its semantic contribution is generic and unspecified, and that the verb becomes intransitive by the person-affix test. The disagreement is whether this slot-filling constitutes a grammaticalized voice alternation or a lexical derivation that happens to intransitivize. The width of the prefix's semantic scope — extending to Recipient and Causee demotion, not only Patient — and its discourse conditioning give the antipassive account stronger typological grounding. This chapter adopts Bugaeva's label while registering Tamura's and Satō's derivational reading as the earlier, still-maintained alternative Satō (2008: §29.2) ‹contested›.

83.4 Homophony: antipassive i- and the fourth-person object i=

Two surface strings i- must be kept distinct. The antipassive prefix i- is a derivational element in the slot immediately before the verb root; the fourth-person object marker i= is a person index belonging to the transitive object paradigm. Both are segmentally identical, and both can appear on the same verb surface, but their distributional profiles diverge Bugaeva (2021); Nam (2021: 62–63).

The antipassive prefix satisfies the object slot derivationally and renders the verb intransitive: the derived verb takes S suffixes and cannot simultaneously host a separate object marker. The fourth-person marker, by contrast, leaves the verb transitive and is cross-referenced by a prefixal A index alongside the i= object marker. A minimal diagnostic: i-nu=an (ANTIP-hear=4.S) is intransitive, proven by the suffixal =an; while a=i=nu (4.A=4.O=hear) is transitive and the i= there is the object marker, not the antipassive prefix Nakagawa (2024: 190–191). The arity computation — the same instrument that distinguishes applicative e- from affiliative e- and from the second-person marker e= — is the sole disambiguating tool in both cases Nakagawa (2024: 192–193).

83.5 Syntactic diagnostics and re-applicativization

The antipassive instantiates an S=A alternation: the Agent argument of the transitive base surfaces as the sole argument of the antipassive intransitive, cross-referenced by the same person indexes that mark the transitive A. The alternation is illustrated by Bugaeva and Kobayashi's minimal pair: nea kamuy a=ri 'I skin that bear' (transitive, specific Patient) versus i-ri=an '(I will) skin the catch' (antipassive, Patient removed from the frame) Bugaeva & Kobayashi (2022: §4.2.3). In the antipassive the prefixal a= is absent because there is no object slot to fill; the suffixal =an carries logophoric first-person-singular reference in the narrative context.

The 1PL.EXCL intransitive suffix =as cannot attach to a transitive verb, so its presence on i-nu=as and i-ku=as 'we drank (alcohol)' unambiguously proves the intransitive status of these antipassive derivatives Nam (2021: 62). The latter form is attested in a Saru oral narrative (Tamura 1993[1979]: 6, cited via Bugaeva 2021: 219 in Nam 2021: 62) and constitutes the core corpus evidence that the Person-Affix test extends to real discourse contexts, not only constructed elicitation.

The antipassive can be re-applicativized: the applicative prefix e- reintroduces a content or instrument object to the now-intransitive stem, raising arity by one and yielding a transitive again. The result is that e- and i- cancel each other's arity effects while changing the type of object expressed:

This pattern, where an applicative overlays an antipassive to shift the object type without restoring the suppressed Patient, is productive and has a direct structural parallel in the re-applicativization of other arity-decreasing prefixes. The derivational chain through ruska documented in Chapter 73 (Valency, Voice, and Noun Incorporation: Overview) traces the same logic: the arity decrements and increments compose arithmetically, and each step is attested in the corpus.

83.6 Origin: the three-way reconstruction debate

Three accounts of the historical source of antipassive i- are in active competition, and none is settled in the sources consulted directly for this grammar. The dispute turns on the direction of the relationship between the antipassive prefix and the fourth-person object marker i=, and on what the ultimate etymon of both might be.

The traditional view, associated with Kindaichi, Tamura, and Shibatani and reported through Nam, derives the antipassive from the bound fourth-person pronoun: the clitic was grammaticalized into a derivational voice prefix as its object-demoting function became conventionalized Nam (2021: 64) ‹contested›.

Bugaeva reverses the direction. In her 2021 analysis the antipassive i- arose from object-incorporation of a generic noun 'thing / place / time', in the same way that the body-part prefixes he- 'head' and ho- 'bottom' arose from incorporated body-part nouns. Once the incorporated generic noun grammaticalized into a Patient-demoting voice morpheme, the antipassive form — superficially resembling the pronoun — extended to unspecified speech-act participants, and this extension gave rise to the person index i=. The antipassive thus preceded the pronoun Bugaeva & Kobayashi (2022: §4.2.3); Bugaeva (2021: 235–239) ‹contested›.

Nam's 2021 reconstruction agrees with Bugaeva that the antipassive is older than the pronoun but proposes a different etymon. Drawing on the parallel between the antipassive prefix, the clause-final nominalizer -i (< *ihi), and the possessed suffix -hV/-VhV, he reconstructs the source as Proto-Ainu *hi / *ihi 'thing, place, time'. The initial consonant was lost through the accumulated phonological effect of a long incorporation history: incorporated noun-stems that are word-initial rarely follow a vowel, which would have preserved the h, and word-initial *h was subsequently lost in that position. Cross-linguistic parallels include Winnebago wa- and Bantawa kha, both generic-noun-origin antipassive markers Nam (2021: 65–74) ‹contested›.

A single fossilized form bears directly on all three positions. The transitive verb ikka ‘steal, take by stealth’ has been analyzed as i-kV-ka (ANTIP-steal-CAUS), a frozen antipassive re-transitivized by the direct causative suffix -ka, synchronically opaque but historically preserving the derivational layer. This analysis is attributed to Bugaeva (2010) by nukopoli in the aynu-corpora Discord community (nukopoli, aynu-corpora Discord 2026-05-19) aynu-corpora Discord (2023–2026); Bugaeva (2010). Since Bugaeva (2010) is not consulted directly for this grammar, and since the competing segmentation ik-ka vs. i-kV-ka cannot be decided from the surface form alone, the fossil analysis is graded ‹contested›. Vovin's competing reconstruction of Proto-Ainu *ihka for the same root implies no antipassive layer (reported via the same Discord discussion). If the Bugaeva/community segmentation is correct, ikka parallels the lexicalized antipassives inkar and ipe at a deeper historical stratum where the causative has since merged with the root ‹speculative›.

The width of the antipassive's semantic scope — extending to non-human objects and to ritual 'thing/place' readings (i-omante, i-mak 'behind') — fits the generic-noun hypothesis better than the pronoun-first account, since a grammaticalized human pronoun would be expected to bias the antipassive toward human Patients rather than generating the observed range Nam (2021: §4.1) ‹speculative›. The residual-distribution evidence and Nam's phonological argument converge on this conclusion.

83.7 Relation to the middle ci- and the A/P split in arity-decreasing voices

The middle prefix ci-, treated in full in Chapter 87 (Middle, Anticausative, and the (Non-)Passive Question), likewise reduces arity by one but from the affected side: it presents an event as experienced by the Patient rather than suppressing the Patient from the frame. The paired minimal set in Nakagawa is: a=cari 'I scatter (something)' (transitive, Patient as object, agent indefinite) versus ci-cari 'it is scattered' (middle, affected noun as sole subject, agent absent) Nakagawa (2024: 190–191). Both operators are arity −1, but i- foregrounds the Agent as S while leaving the Patient unindexed, while ci- foregrounds the Patient as S while leaving the Agent unindexed. The two are therefore the A-perspective and P-perspective members of a single arity layer, and the Handbook glosses the latter object-resultative (RES) where this chapter's analysis uses ANTIP Bugaeva & Kobayashi (2022: §4.2.4).

A community-proposed analysis identifies a productive series of middle-voice derivatives in which ci- takes a transitive or causative base and produces an intransitive middle/anticausative: cikarkarpe 'made thing', copirasa 'spread (intr)', ceosmare 'rush forward' (nukopoli, aynu-corpora Discord 2023-12-08) aynu-corpora Discord (2023–2026). This extends the picture in Bugaeva & Kobayashi (2022: §4.2.4) and is graded ‹contested› pending systematic corpus verification. The ci-/i- split documented here aligns with the reflexive split treated in Chapter 84 (The Reflexive yay-) and Chapter 85 (The Reflexive-Possessive si- and the yay-/si- Division): multiple arity-decreasing operations target different argument roles, and the arity calculus in Chapter 73 (Valency, Voice, and Noun Incorporation: Overview) provides the unifying framework.