Chapter 71The Split-Intransitive (Active–Stative) Core

71.1 What split-intransitive alignment predicts

A split-intransitive, or active–stative, alignment divides intransitive subjects into two agreement classes: subjects that pattern with transitive A, and subjects that pattern with transitive O. The conditioning factor is typically control, volition, or affectedness (Stassen 1997: 23–28); (Foley & Van Valin 1984: 29–37). The Hokkaido Ainu finite person-index system makes no such division. A predicate's transitivity class — zero-place, one-place, two-place, or three-place — fixes which indexes appear on the verb; within the one-place class, no further agentivity distinction feeds the morphology Nakagawa (2024: 124–128); Bugaeva (2012: 471–472).

The morphological architecture, including the full S/A/O paradigm and the side-encoding that makes the 1PL.EXCL and fourth-person rows tripartite, is described in the personal-affix architecture chapter. This chapter addresses the narrower diagnostic: whether the intransitive S column itself splits by agentivity. It also explains where an active-looking division genuinely occurs — in lexical classification, incorporation diagnostics, and valency-changing morphology — and why that domain does not constitute an agreement-level split ‹consensus›.

71.2 All intransitive subjects use the S set

Nakagawa's valence chapter gives the decisive test. Hokkaido intransitive verbs with an acting, controlling subject and intransitive verbs whose sole argument is a patient, experiencer, or affected entity alike take the intransitive S forms. Agentive intransitives and patientive ones are both one-place verbs; the personal-index system responds to the valence count, not to the semantic role Nakagawa (2024: 126–127). His own examples pair a zero-place ambient predicate — which bears no person index at all — with an experiential predicate whose sole argument is an affected experiencer indexed by the fourth-person S suffix =an, the same suffix an agentive subject of the same person would carry.

A force-of-nature predicate provides the same diagnostic from a different angle. In rittursere ‘be rolled by waves’, the wave element is incorporated into the stem and the human undergoer is the sole grammatical subject. That undergoer is indexed by =an; the fourth-person object form i= does not appear. A patient-like meaning does not move the index toward the O set Nakagawa (2024: 126–127).

The first singular supplies a complementary diagnostic. The 1SG row is nominative-accusative: ku= serves as both A and S, opposed to the object en= Bugaeva (2012: 472); Bugaeva (2008: 32). A 1SG agentive intransitive and a 1SG experiential intransitive look identical in their person index; the object form en= appears only when the participant is the O of a transitive verb. The identity belongs to the 1SG row of the paradigm and is not tracking control ‹consensus›.

71.3 Lexical classes, the unaccusative diagnostic, and incorporation

Within the intransitive class, a semantically motivated distinction is widely recognized. Intransitive verbs whose subject is a semantic agent — labelled S_a — behave differently from those whose subject is a semantic patient or undergoer (S_o) with respect to certain derivational and syntactic operations (Foley & Van Valin 1984: 29–37); (Mithun 1984: 854–856). In Hokkaido Ainu the distinction operates at the lexical and derivational level; it does not feed finite person-index assignment.

The clearest evidence for an S_a/S_o cut comes from noun incorporation. The accessible Shibatani source is the 1988 conference abstract, which states that the incorporable nominal is the patient subject of an intransitive clause and the object of a transitive clause — an unaccusative framing stated in terms of semantic role and grammatical relation Shibatani (1988: 210–211). The abstract contains no alignment-type term. A broader active-alignment characterisation sometimes associated with Shibatani (1990) cannot be attributed on the basis of the pages read; that book could not be independently consulted for this grammar ‹speculative›. The noun-incorporation domain, including the patient-subject diagnostic, is developed in noun incorporation.

The S_a/S_o distinction is therefore a valence and lexical-class diagnostic, not an agreement-level split. A patientive intransitive subject is eligible for certain derivational processes — notably the unaccusative-style incorporation Shibatani describes — that agentive subjects are not. Once the predicate's transitivity class is fixed, its sole argument takes the S set regardless of whether that argument is an agent, experiencer, or undergoer.

71.4 Valency-changing morphology and the S set

Valency-changing morphology moves predicates across transitivity classes before person indexing is assigned. The antipassive or indefinite-object prefix i- ‘antipassive / something’ detransitivises a transitive verb whose object is indefinite or backgrounded; the resulting predicate is intransitive, and its sole argument — which is an agent — takes the S set, formally identical to that of any intransitive experiencer Tamura (1972: 35); Nam (2021: 62–74). Reflexive yay- and reciprocal u- fill the object position and reduce transitivity in the same way. Applicatives and causatives add object positions, moving a predicate from intransitive to transitive class Nakagawa (2024: 196–199); Satō (2023: 40).

The antipassive i- is the case most directly relevant to the split-S question, because applying it to a semantically transitive stem produces an intransitive whose subject is an agent yet whose S index is formally identical to that of an intransitive experiencer. The prefix must also be distinguished from the fourth-person object pronoun i=, which is an O-form of the indefinite/fourth series alongside A-form a= and S-form =an; the two are kept apart by the transitivity class of their host, not by the meaning of the argument Nam (2021: 62–64); Bugaeva (2021: 235–239). The antipassive and its origin debate are treated in the detransitive i-.

71.5 Zero-place predicates and the valence boundary

The boundary between zero-place and one-place predicates marks the outer limit of S-set assignment. Ambient weather predicates such as sirsesek ‘be hot (weather)’ and mean ‘be cold (weather)’ are zero-place: they index no participant Nakagawa (2024: 127–129). Felt-temperature predicates such as sesek ‘feel hot’ and merayke ‘feel cold’ are one-place and carry a person index on their experiencer subject. The zero-place/one-place boundary separates argument-free from argument-bearing predicates; it is a valence distinction, not an agentivity one.

Ijäs treats weather expressions as avalent and places ambient forms in that category; Nakagawa agrees on the category but leaves apto as 'rain falls' analytically uncertain, since apto 'rain' is an autonomous noun that can function as a subject in other constructions Nakagawa (2024: 128–130); Ijäs (2023) ‹contested›. Property-concept predicates — pirka ‘be good’, poro ‘be big’, hure ‘be red / become red’ and related items — are one-place intransitives in the baseline Hokkaido analysis. Ijäs notes that descriptive verbs also carry inchoative readings, so a single lexeme spans a stative and its onset without changing the person index. Both uses index the sole argument as S. Their predication syntax is treated in Chapter 99 (Property-Concept (Adjectival) Predication without a Copula).

71.6 Ijäs's L2–L3 claims in this domain

Ijäs's learner grammar organises valence through the metaphor of verbal "pockets": the verb carries participant information obligatorily, while external noun phrases are optional cross-references. Analytically this is the head-marking architecture of (Nichols 1986), used also by Satō Satō (2025: 361). Her Lesson 2 states that intransitive verbs take one argument and Lesson 3 that transitives take two; these arity classes align with Nakagawa's zero/one/two/three-place distinctions and Satō's valence discussion Nakagawa (2024: 124–128); Satō (2023: 37–47); Ijäs (2023).

Two Ijäs formulations that bear on the split-S question carry caveats. Her Lesson 9 account of i- as an antipassive filling the object position with a generic 'thing/place' meaning aligns in substance with the Bugaeva line; Satō uses a derivational-prefix label without the term 'antipassive' Ijäs (2023); Satō (2008: 232–233); Bugaeva (2021: 235–236) ‹contested›. Her Lesson 3 formulation that the indirect object falls between subject and object in ditransitives is a pedagogical word-order simplification; Nakagawa's ditransitive examples show no fixed order for the two non-subject noun phrases Nakagawa (2024: 363–365); Ijäs (2023) ‹contested›.

71.7 The alignment label and the broader debate

For Hokkaido person indexing, the secure description is mixed morphological alignment by person row, with no productive split-S agreement. Bugaeva's formulation remains the standard account: neutral alignment in second and third person, nominative-accusative in 1SG, and tripartite in 1PL.EXCL and the indefinite/fourth person Bugaeva (2012: 472, 498); Bugaeva (2008: 32). Nakagawa's valence chapter blocks an agentivity-conditioned reading of the S column Nakagawa (2024: 126–127). Satō frames the same system as head-marking and argues for a zero third-person affix on the basis of the two-index cap and ditransitive restrictions; Nakagawa denies a third-person exponent. That dispute affects the model of the zero row but produces no active/stative split among the overt S forms Nakagawa (2024: 169); Satō (2023: 39, 47); Satō (2025: 369) ‹contested›.

Sakhalin work sharpens the typological contrast. Dal Corso rejects a tripartite component for ‹SA› Sakhalin, proposing a mixed nominative-neutral-inverse analysis, and explicitly cautions that his diachrony and hierarchical account may not apply to Hokkaido Dal Corso (2025: 187–188, 200, 213) ‹contested›. The tripartite, hierarchical, and inverse positions are weighed against one another in Chapter 72 (Tripartite, Nominative-Accusative, and the Central Alignment Debate); the Sakhalin contrast leaves the Hokkaido picture intact.

The productive-split question has one remaining open edge. The literature does not yet provide a verb-by-verb inventory of whether any Hokkaido intransitive verb alternates between A-like and O-like S marking according to control or discourse construal. Such an inventory would require a raw-token study across the corpus; the normalised frequency layer strips personal-affix clitics and cannot supply per-affix counts for individual S forms. No example in the material consulted shows an intransitive undergoer indexed by the object forms en=, un=, or i=; the available evidence supports the uniform-S generalization as the correct synchronic description ‹speculative›.