Chapter 87Middle, Anticausative, and the (Non-)Passive Question

87.1 The ci- middle prefix

Among the arity-decreasing voice prefixes, ci- ‘middle prefix (object-resultative)’ presents an event from the affected side and yields an intransitive verb whose sole argument is the entity undergoing the action Nakagawa (2024: 190–191). Nakagawa labels the output 中動態 (middle voice) and gives the underlying semantics as "expressing, from the affected side, an action one thing performs on another" Nakagawa (2024: 190). The Handbook glosses the same prefix object-resultative (RES), emphasising the affected argument's endpoint state Bugaeva & Kobayashi (2022: §4.2.4). Both labels address the same morphosyntactic fact: the base transitive loses its agent position, arity drops by one, and the base object surfaces as the intransitive subject. The terminological difference between "middle" and "object-resultative" is not analytical ‹contested›.

The prefix reduces arity by exactly one. cari ‘scatter something’ (transitive, +2) gives ci-cari 'be scattered' (intransitive, +1); sina ‘tie something’ (transitive) gives ci-sina 'be tied'. The prefixed form takes suffixal S-argument indexes (fourth-person =an, first-plural exclusive =as) in place of the A-argument prefixes of the transitive base — the standard transitivity diagnostic of Chapter 73 (Valency, Voice, and Noun Incorporation: Overview). The following table gives a selection of attested pairs.

Typologically, middle voice marks spontaneous, patient-initiated, or affectee-oriented events across many of the world's languages — the semantic space where a passive would otherwise be expected but no agent is expressed Shibatani (1990: 41–48). In Hokkaido Ainu, ci- occupies that niche: the source event is agentive (someone scatters, someone ties), but the prefix profiles its outcome from the affected participant's perspective, suppressing the agent position entirely.

87.2 Person restriction and the ci= / ci- homophony

The middle prefix ci- is restricted to third-person subjects. The form *ku=ci-montapire ('I was informed') is ungrammatical; a first-person affected participant instead requires the indefinite-agent construction with objective cross-referencing: a=en=montapire 'the indefinite agent informed me as object' Nakagawa (2024: 190). This restriction distinguishes ci- sharply from the reflexive yay- ‘REFL’ and the antipassive i- ‘antipassive’, which are available across persons.

The prefix is homophonous with the first-person plural exclusive A-index ci=, and both reduce arity by one — the index by filling the subject slot with a specific 1PL.EXCL argument, the middle by suppressing the agent entirely. Nakagawa treats them as "probably related but synchronically distinct" and notes a shared phonological behaviour: before a, e, o the i drops, yielding surface ca-, ce-, co- (ci-opirasa → copirasa) Nakagawa (2024: 190–191). Disambiguation in context relies on person morphology and arity: a middle verb takes suffixal S-indexes, while a first-plural A-prefix takes the O-indexes typical of transitives.

Several ci- forms have lexicalized as uninflected bases. The noun cikoykip ‘game animal, prey’ — 'thing that is hunted' — freezes the middle of koyki ‘hunt, catch something’, no longer perceived as a productive derivational composite by speakers Nakagawa (2024: 191). Lexicalized middles of this type document the prefix's historical productivity; the set of active, synchronically transparent pairs is smaller.

87.3 Community analysis: ci- as a productive middle

Discussion in the aynu-corpora Discord has proposed that ci- functions as a systematically productive middle/anticausative across a wider set of lexemes than the published literature has catalogued (nukopoli, aynu-corpora Discord, 2023–2024) aynu-corpora Discord (2023–2026). The proposal analyses several opaque-looking forms as transparent middle derivations: copirasa 'spread out, widen (intr)' as ci=opirasa (from transitive opirasa ‘spread something out’); ceosmare 'dash forward, plunge in' as ci=e-osma-re (from an applicative-causative 'cause to thrust into'); and cikarkarpe as ci=kar-kar-pe 'things being made' aynu-corpora Discord (2023–2026) ‹contested›.

The same line of analysis was applied to ci-koykarkar. On this reading the ci- in hattoho ci-koykarkar is the middle prefix, yielding a passive-like meaning 'restrictions are imposed on [us]', rather than the first-plural A-index ci=, which would give 'we impose restrictions together' (nukopoli, aynu-corpora Discord, 2024) aynu-corpora Discord (2023–2026) ‹contested›. The disambiguation requires pragmatic and discourse context; surface form alone does not decide it.

These community analyses are leads rather than established findings: they extend the attested middle use productively to forms the peer-reviewed descriptive literature has not formally analysed under this label. The core claim — that ci- is a productive middle prefix — is supported by Nakagawa and the Handbook; the community contribution is the extension to a broader set of lexemes and the identification of middle ci- in forms previously treated as first-plural-exclusive or as unanalysed compounds. A further diachronic hypothesis that ci- derives from an earlier passive-like *ti, distinct in origin from the person index, was proposed but disputed on phonological grounds — the affricate [t͡s] realised as c also appears in non-palatalizing environments, casting doubt on a straightforward palatalization account (nukopoli and aomidori, aynu-corpora Discord, 2023) aynu-corpora Discord (2023–2026) ‹speculative›.

87.4 Anticausative semantics and the si- overlap

The reflexive prefix si- ‘AUTOC/ANTIC/REFL’ covers a semantic domain that overlaps partially with the middle. Under Satō's analysis, adopted by Nakagawa and the Handbook, si- marks the subject's indirect involvement — through another agent, through circumstance, or through a force outside the subject's direct control Satō (2007: 31–39); Nakagawa (2024: 188–189); Bugaeva & Kobayashi (2022: §4.2.2). The minimal pair yay-pusu 'float up volitionally (by swimming)' against si-pusu 'float up of itself' illustrates the anticausative semantics of si-: the event is profiled as spontaneous, with an affected participant rising without an expressed agent Satō (2007: 31–32). Similarly, the transitive turi ‘lay something down, stretch something’ yields si-turi 'stretch, lie down (of itself)', which Tamura analyses as a pure intransitivizing derivation Tamura (1996: 622). Kobayashi groups such pairs under "anticausativization" — a transitivity alternation in which the causer is suppressed and the event presented as spontaneous Kobayashi (2015: 33).

The morphological and person-marking properties of ci- and si- remain distinct despite the semantic overlap. ci- suppresses the agent position entirely and is restricted to third-person contexts; si- fills the object slot with a self-participant coreferent with the clause subject and is available across all persons. Spontaneous or agent-backgrounded events with a first- or second-person participant therefore surface with si-, not ci-. The full treatment of the yay- / si- opposition belongs to Chapter 84 (The Reflexive yay-) and Chapter 85 (The Reflexive-Possessive si- and the yay-/si- Division); the present section records only the semantic intersection with the anticausative domain.

Bugaeva's diachronic study of anticausatives in Ainu concentrates on two markers: the primary anticausative / intransitivizing suffix -ke, and si-, whose path is REFLEXIVE > AUTOCAUSATIVE > ANTICAUSATIVE Bugaeva (2025: §1.2, §3.1.2, §5). The inchoative-causative interface — where a stem such as rewke ‘be bent’ / rewe ‘bend something’ pairs a spontaneous intransitive with a transitive — is treated in Chapter 75 (The Transitivizer -ka and Lexical Valency Pairs). The prefix ci- is therefore best kept as a separate MID/RES resultative prefix: it suppresses an agent and profiles an affected result, but it is not the same historical problem as Bugaeva's si- anticausative path.

87.5 Body-region prefixes he- and ho-

Two lexical prefixes of noun origin, he- 'head, tip' and ho- 'base, bottom', also reduce arity by one by filling the object slot with the subject's own body-region Nakagawa (2024: 186–187); Bugaeva & Kobayashi (2022: §4.2.4). puni ‘lift something’ (transitive) yields he-puni 'raise one's head' (intransitive) and ho-puni 'get up, fly up' (intransitive); kiru ‘turn something’ yields he-kiru 'turn around'; tuye ‘cut something’ yields ho-tuye 'shout' (semantically opaque) Nakagawa (2024: 186–187).

These prefixes are the arity-decreasing (−1) counterparts of the homophonous affiliative e- 'head of' and o- 'base of', which are arity-neutral (0) and fill a slot while requiring their own possessor. The minimal pair makes the arity contrast concrete: he-puni is intransitive and takes S-indexes, while affiliative e-puni 'raise the head/end of something (belonging to someone)' remains transitive Nakagawa (2024: 192–193).

Bugaeva and Kobayashi group he- and ho- together with the ci- middle under the object-resultative heading, noting that they "function similarly to O-incorporation" — a body-region semantics fills the object slot and makes the verb intransitive Bugaeva & Kobayashi (2022: §4.2.4). Many surface forms are semantically opaque: ho-tuye 'shout' transparently contains tuye 'cut' with ho- 'base', yet the compositional meaning is not recoverable Nakagawa (2024: 187).

87.6 The indefinite-agent construction: a= as a functional passive

When a transitive event is presented with an unspecified or backgrounded agent, Hokkaido Ainu uses the fourth/indefinite person prefix a= ‘fourth-person / indefinite agent’ — or the dialectal variant an= in Ishikari and some other varieties — as the agent argument of the transitive verb. The patient stays in object position; no promotion to subject occurs. The construction is valency-neutral: the verb retains its transitive argument frame, with the indefinite agent filling the A slot and the patient cross-referenced as O Nakagawa (2024: 428–429); Bugaeva & Kobayashi (2022: §1).

A demoted agent is expressed with the postpositional phrase or(o) wa 'from', taking the NP that would normally appear as the transitive subject Nakagawa (2024: 428–429). The construction is frequent with unintended or inanimate causers and in polite speech where direct naming of the agent is avoided. The full morphology, referential range, and dialect distribution of the fourth/indefinite person are treated in Chapter 67 (The Indefinite/Fourth Person: Forms, Reference, and Agent-Defocusing).

Satō's subjecthood diagnostic makes the non-promotion of the patient explicit. If the patient were a true intransitive subject, a 'want to be loved' predicate should take bare rusuy ‘want to, desire’ with subjective marking. Instead the construction requires the nominalizer hi together with objective cross-referencing on the verb Satō (2008: 205–206):

The distinction between the ci- middle and the a= construction also emerges clearly here: in a ci- clause the affected participant surfaces as the intransitive subject and no agent may be expressed at all; in the a= clause the patient is syntactically the object, an agent may appear in or wa, and the verb is transitive by all formal tests Nakagawa (2024: 429–430).

87.7 Does Ainu have a true passive?

Hokkaido Ainu has no subject-promoting passive Nakagawa (2024: 428); Bugaeva & Kobayashi (2022: §1). Nakagawa states it directly: それにあたるものはアイヌ語には無い 'there is nothing in Ainu that corresponds to it' Nakagawa (2024: 428). The two constructions that carry the functional load a passive would otherwise bear are structurally distinct from each other and from any classic subject-promoting passive.

The ci- middle moves the base object into subject position — in that respect it is the nearest Ainu analogue to a valency-decreasing passive — but it categorically blocks agent expression Nakagawa (2024: 430). A classic passive combines patient-as-subject with an optional by-phrase; no Ainu construction combines these two properties. Nakagawa adds that the middle has a verse-register periphrastic use with the causative — ci-ra-na-ran-ke (from ran ‘descend’, stretched to five syllables) — where meaning is unchanged; this verse device is not available for the a= construction Nakagawa (2024: 430).

The Handbook labels the a= / an= construction a "valency-neutral impersonal passive" to capture its agent-defocusing function, while recognizing that it lacks subject promotion Bugaeva & Kobayashi (2022: §1) ‹contested›. Satō's preferred formulation — "a half-active, very peculiar passive structure" — likewise foregrounds the mismatch with cross-linguistic passive prototypes Satō (2008: 205) ‹contested›. Neither label implies that the construction is a passive in the structural sense; both describe the same functional fact from different angles.

The functional distribution that results follows from the structural properties. For third-person affected participants where no agent is at issue, ci- presents the event from the affected side, and the patient-as-subject reading is the natural one. For first- or second-person patients, or wherever the agent may need to be expressed, a= / an= is used instead, with the patient retaining object status and the verb its transitive morphology Nakagawa (2024: 429–430). A by-phrase that co-occurs with patient-as-subject is unattested in the Hokkaido corpus ‹corpus-confirmed›.

Ijäs's learner grammar does not treat the passive question or the ci- middle Ijäs (2023). The nominalizer hi that anchors Satō's subjecthood diagnostic connects this discussion to the nominalization system of Chapter 101 (The Nominalizers -p/-pe and -i/-hi (Participant, Event, Place, Fact)). The antipassive i-, which defocuses the object from the subject's side, is the mirror-image operation and is treated in Chapter 83 (The Antipassive / Detransitive i-). The productive causative -re / -te / -e, which introduces an agent argument rather than suppressing one, is in Chapter 74 (The Morphological Causative -re / -e / -te).