Chapter 103Relative Clauses: The Gap Strategy and the Accessibility Hierarchy

103.1 Overview

The Hokkaido Ainu relative clause is prenominal and head-final: the modifying clause precedes its head noun, there is no relative pronoun, and the relativized participant is simply absent from the clause — a zero gap Nakagawa (2024: 503); Satō (2008: 168); Shibatani (1990: 39); Bugaeva (2015: 47). Stative property verbs modify a following noun directly and without any special marking: hure suwop 'red box'. Any verb, stative or dynamic, can head the modifying clause; the verb inside the RC carries the same person prefixes, number-sensitive suppletion (singular ek ‘come’ vs plural arki ‘come (PL)’), polarity, and clause-internal morphology as it would in a main clause Shibatani (1990: 39–40); Satō (2008: 168). Subordination is signalled by the structural position of the clause before its head, not by any morphological change to the verb Bugaeva (2016: 100); Nakagawa (2024: 503).

This architecture — a full finite clause immediately before its head — is shared by the headless relative (where a formal noun such as p ‘thing, person’~pe ‘thing, person’ supplies the head) and the complement clause (where a formal noun nominalizes a saturated clause). The boundary between these constructions is governed by Nakagawa's valency (arity) principle, discussed in the following section and with worked examples in the chapter on the nominalizer p~pe. The headless variant is treated in Chapter 104 (Headless, Internally-Headed, and Noun-Modifying Clause Analyses); lexicalized RC-headed nominalizations such as aep ‘food’ in Chapter 102 (Lexical-Head and Formal-Noun Nominalization (kur, uske, ruwe/hawe/siri/humi)).

103.2 The arity ≥ +1 diagnostic

Nakagawa's arity accounting provides the formal test that separates the relative clause from the complement clause. Each verb contributes a positive arity value equal to the number of NP arguments it requires; each NP present inside the clause subtracts one; a formal noun has arity −1. When the summed arity of the embedded clause plus its formal-noun head is still positive (≥ +1), one argument slot remains unfilled — that is the gap — and the whole unit functions as a relative clause. When the arity reaches zero, the clause is saturated and the formal noun nominalizes the whole clause as a complement Nakagawa (2024: 261–262, 504, 507).

The surface string [clause] p/pe is identical in the relative and the complementizer readings; only the arity count distinguishes them. Satō's rule of thumb for the same distinction is whether the original sentence from which the head noun was extracted can be reconstructed: for a relative clause it can, for a nominalizer it cannot Satō (2008: 174–175).

103.3 The gap strategy and postposition retention

The gap is a zero element: no pronoun, resumptive or otherwise, marks the relativized position inside the RC Bugaeva (2015: 47); Dal Corso (2020: 51). When the head noun fills a core argument position (S or O) in the embedded clause, this gap is sufficient, because the verb's argument structure and the remaining NPs identify the role of the missing participant. When the head fills an oblique position — instrumental, comitative, locative, dative — the postposition that governs that oblique remains inside the RC, moved to a position at or near the left edge of the embedded clause. This retained postposition is what marks the semantic role of the gapped head when that role is not recoverable from the verb's argument structure alone Bugaeva (2015: 47–49); Nakagawa (2024: 504).

The practical consequence is that a postposition's presence inside an RC signals an oblique head. Conversely, removing the postposition collapses the arity to zero and makes the relative-clause reading unavailable. Nakagawa's minimal pair makes this visible: nea menoko ani turepta ni is grammatical as an instrumental RC ('the stick with which that woman dug uba-lilies'), but the postposition-free string *nea menoko turepta ni has no gap and cannot relativize the instrument Nakagawa (2024: 505–506).

Only postpositions that are capable of appearing independently with a zero-anaphoric NP in a main clause — ani ‘with (instrumental)’, turano ‘together with (comitative)’, eun ‘to it (dative)’, pakno ‘up to, as far as’ — undergo this retention directly. For locative and allative postpositions that require a case-marking particle (ta, un), the relational noun oro ‘place, there’ must be inserted before the particle; it serves as the NP host for the postposition and as the covert correlate of the gapped head Bugaeva (2015: 49); Satō (2008: 172); Nakagawa (2024: 510–511).

103.4 The Keenan–Comrie accessibility hierarchy

Bugaeva maps Hokkaido Ainu relativization against the noun phrase accessibility hierarchy of (Keenan & Comrie 1977) and finds that HA relativizes all core and most oblique positions; the only positions that resist are the alienable possessor and, directly, the standard of comparison Bugaeva (2015: 45, 47–49). These findings are confirmed for southern Hokkaido data by Dal Corso (2020: 51).

103.5 Unattested and blocked positions

103.6 RC heads: common nouns and bound nouns

The external head of an HA relative clause may be a common noun or a formal (bound) noun. Among the formal nouns, kur ‘person (honorific)’, utar ‘people’, p ‘thing, person’~pe ‘thing, person’, uske ‘place, time’, and hi ‘thing, time, place (abstract)’ are, in Bugaeva's account, "even more common" as RC heads than common nouns Bugaeva (2015: 47). The bound noun supplies a generic semantic category — person, people, thing, place — while the RC clause provides all the identifying content.

When the formal noun is p~pe, the construction shades into the headless relative. That continuum is the subject of Chapter 101 (The Nominalizers -p/-pe and -i/-hi (Participant, Event, Place, Fact)) and Chapter 104 (Headless, Internally-Headed, and Noun-Modifying Clause Analyses). The lexicalized end — former RC-plus-formal-noun combinations frozen into compound nouns such as aep ‘food’ (a=e-p, 'thing one eats') — is treated in Chapter 102 (Lexical-Head and Formal-Noun Nominalization (kur, uske, ruwe/hawe/siri/humi)). The interaction of relativization with noun incorporation, where a patient argument merges into the verb stem and changes the arity count, is treated in Chapter 88 (Object (Patient) Noun Incorporation).

103.7 Ijäs Lesson 11 — graded claims

Ijäs's English-language grammar guide covers relative clauses in Lesson 11 Ijäs (2023). Her examples are constructed teaching material; none are corpus-attested. Her core grammatical claims — prenominal head-final order, absence of a relative pronoun, gap strategy for core arguments, postposition retention for obliques (including the oro-insertion rule for locatives/allatives) — align precisely with the primary sources ‹consensus›. Her demonstration that ku=kor cep 'my fish' is itself a relative clause ('the fish that I have') follows the Handbook's alienable-possession analysis Bugaeva (2022: ch. 1) ‹consensus›. And her identification of the ditransitive subject as a problematic head position matches Bugaeva's "no examples attested so far" Bugaeva (2015: 48); she is consistent in flagging it as an open question rather than asserting ungrammaticality ‹corpus-confirmed› (as gap).

Two of her claims need a hedge. Her report that the ordering of multiple stative modifiers is free (poro retar seta ~ retar poro seta 'big red dog, red big dog') is her own inference from the absence of counter-examples; she acknowledges this explicitly. Modifier stacking is attested — Shibatani gives ampene pirka sukup aynu from yukar data Shibatani (1990: 39) — but a free-order generalization across dialects exceeds what the sources establish ‹corpus-suggested›. Her inability to locate attested examples of ablative wa(no) or mutative ne as RC heads is shared by all sources read here; the forms she cautiously suggests for these positions should not be treated as established Ijäs (2023) ‹corpus-suggested›.

103.8 Analyses and open questions

All major descriptive accounts converge on the externally-headed, prenominal gap strategy: Shibatani (1990: 39–42), Refsing (1986: 259–264), Satō (2008: 168–173), Nakagawa (2024: 503–511), and Bugaeva (2015: 47–49) all describe the construction in these terms. The differences among them are terminological — Refsing calls the process "adnominalization", Nakagawa uses "関係文", Bugaeva "gapped prenominal relative clause" — rather than empirical.

Dal Corso 2020 proposes that certain southern Hokkaido indirect-evidential constructions (siri ki, hawe as, humi as) show a layout reminiscent of internally-headed relative clauses, instantiated through what he calls classificatory noun incorporation Dal Corso (2020: 51–52). This proposal covers the indirect-evidential constructions specifically; it has not been adopted by Bugaeva, Satō, or Nakagawa; and its claim of classificatory NI remains unconfirmed in the wider literature ‹contested›. The relativization system described in this chapter is the externally-headed construction; the IHRC/CNI question bears on the evidential construction and is addressed in Chapter 104 (Headless, Internally-Headed, and Noun-Modifying Clause Analyses).

Three gaps in the accessible evidence remain open. First, whether the ditransitive subject position is grammatically excluded or merely unattested is not decided: the absence is systematic — reported independently by Bugaeva 2015: 48 and by Ijäs — but no source provides a grammaticality judgment supported by speaker elicitation ‹contested›. Second, ablative and mutative RC heads are unconfirmed in any source read here ‹corpus-suggested›. Third, the accessibility-hierarchy findings rest primarily on Saru and Chitose data Bugaeva (2015); whether all Hokkaido dialects relativize the same positions has not been systematically tested ‹speculative›.