Chapter 21Relative clauses

21.1 The gap strategy

Any clause may modify a noun. The simplest case is a bare intransitive verb — the ‘adjectives’ of Chapter 6 (Word classes), as in Furukawa's kurasno seta ‘black dog’, are exactly this construction with a one-word clause. With fuller clauses the pattern is unchanged:

The head noun fills its ordinary slot in the main clause, and the relative clause may in turn contain the person prefixes that its own arguments require. When the head corresponds to the object of the modifying clause, the clause-internal A marking stays in place:

21.2 Accessible positions

Measured against the accessibility hierarchy, Ainu relativization reaches nearly the whole way down: subjects, objects, obliques, and inalienable possessors can all head a relative clause, and only the alienable possessor and the standard of comparison are excluded (Bugaeva 2015: 43). Within that wide reach the corpus distributes its heads very unevenly. S and O relativize freely and account for nearly all heads; (1) and (2)–(3) illustrate each, and existential relatives of the shape cise ohta an N ‘the N who was in the house’ are among the most frequent of all (§18.3). Heads corresponding to A are structurally unproblematic — the modifying clause simply retains its overt object — but they are rare in running text, which we attribute to discourse rather than grammar: A referents are typically topical protagonists, person-marked on the verb and not in need of restrictive modification. The pattern matches the cross-linguistic skew of relativization frequency along the accessibility hierarchy rather than any Sakhalin-specific constraint. The lower positions — obliques, possessors, and the standard of comparison — each recruit extra machinery, and the next three sections take them in turn.

21.3 Obliques: retaining the role marker

Below the core roles the bare gap stops working: verb agreement registers arguments, not adjuncts, so an oblique gap would leave its role unrecoverable. The Ainu answer, family-wide, is retention — the postposition that would have marked the constituent stays behind inside the relative clause, orphaned of its complement, and signals the role of the head. In Hokkaidō the device is textbook material; Satō's paradigm pair shows the instrumental ani left at the clause edge once its complement makiri ‘knife’ is promoted to head (Satō 2008: 263–264):

Not every marker can be orphaned. Bugaeva delimits the strandable set as the instrumental ani, the comitative tura, the dative eun, and the terminative pakno; the plain locative ta and allative un cannot strand, and a spatial adjunct relativizes instead through a relational noun such as oro ‘place of’, which remains in the clause carrying the locative marker (Bugaeva 2015: 48, 50). The Sakhalin counterparts of the strandable set are ani ‘with, by means of’, tura ‘together with’, and the terminative pahno ‘until, as far as’ (§11.1), the dative work of Hokkaidō eun falling to onne.

The one direct Sakhalin token of an oblique relative obeys the constraint exactly. In (6) a goal of motion has been relativized, and the clause does not strand bare ta: it retains or-ta ‘at the place (of)’ — the relational-noun bridge — which Dal Corso analyses as the cue that keeps the role of the gapped constituent recoverable (Dal Corso 2025b: 53):

Spatial relativization also has a route that needs no retained marker at all: a place noun as head. In Sentoku's letters the locus argument of an ‘exist’ ‘be (at)’ is gapped under the head toko ‘place’, whose own semantics supply the spatial role; the headed phrase then behaves like any noun, here as complement of the terminative (the same token quoted in §11.4):

Retention competes, finally, with the applicatives of §14.2. An oblique promoted to object by e-, ko-, or o- relativizes by the plain gap, like any other object; for Hokkaidō, Bugaeva shows the applicative route taken wherever the lexicon provides it, with retention as the fallback for roles no applicative covers (Bugaeva 2010: 760). The Sakhalin texts give us no secure relative clause headed by an applied object, but the feeding relation costs nothing: once e-tuhse-ka ‘kick with’ has made the instrument its object, nothing distinguishes that object from any other O under relativization.

21.4 Possessors

An inalienable possessor relativizes with nominal rather than adpositional tracking material: the possessum stays inside the clause in its possessed form, and the retained possessive suffix is what identifies the head as possessor (Bugaeva 2015: 49). Sakhalin is structurally well prepared for this, having generalized the possessed form across nearly the whole noun lexicon (§7.1). The texts show the pattern in phrasings built from clauses like maci-hi isan ‘his wife was away’ (§17.4): in macihi isan horokewpo-type phrases, the possessive morphology on maci-hi is all that marks horokewpo ‘young man’ as the one whose wife is meant. We analyse this as the regular gap strategy applied to a clause whose subject is a possessed noun — no further machinery is needed. The alienable possessor, by contrast, is one of the two genuinely inaccessible positions (Bugaeva 2015: 43): ‘the man whose dog…’ has no direct rendering, and the texts paraphrase through the appositive possession strategy of §7.4.

21.5 The standard of comparison

The other inaccessible position is served by a conventional work-around rather than a relative clause proper. The standard of a comparison (‘than X’) cannot itself be relativized; what relativizes instead is a degree adjunct in the terminative pakno ‘up to, as far as’, stranded like the other members of the strandable set, with the negative existential isam ‘not exist’ as the clause's predicate: ‘a Y [up to whom no Y exists]’, that is, ‘a Y without equal’ — a superlative (Bugaeva 2015: 52–53). Both the idiom and the equative comparison underlying it are on record from Hokkaidō (Saru):

No Sakhalin token of the complete frame survives, but each of its ingredients is in place: the terminative pahno ‘until, as far as’ is frequent (§11.4) and takes a gap-headed relative as its complement in (7); isam ‘not exist’ is the everyday negative existential (§17.4); and degree adjuncts belong to the strandable set. We accordingly read the absence as a gap in the record rather than in the grammar — though, as elsewhere, confirmation would take a token of the shape [X pahno Y isam] Y ne, which the surviving texts do not provide.

21.6 Headless relatives

With the bound nouns pe ~ p ‘one, thing’ and hi ‘thing, time, place’ as heads, relativization yields the participant and event nominalizations of §20.1 — ‘the one that…’, ‘the fact that…’. These heads also combine with the possessive suffix, as in an=oyra-pe-he ‘the things we had forgotten’ (Murasaki (1979: 51); (Bugaeva 2016: 106)). The chain noun > bound noun > nominalizer is the same grammaticalization that produced the negator hanneh(ka) and the topic particle neampe (§17.1; Chapter 24).

21.7 Relative clauses in the noun phrase

Within the noun phrase, the relative clause follows demonstratives and numerals and immediately precedes the head (§19.3); demonstrative plus relative is common in narrative for re-identifying referents (neya cise ohta an monimahpo-type phrases; §9.3). The corpus gives no evidence of a restrictive–non-restrictive contrast in form; both readings are carried by the same construction and resolved in context. Hokkaidō Ainu allows relativization to interact with noun incorporation in ways that have attracted theoretical attention (Dal Corso 2020); the Sakhalin corpus offers parallel ingredients — compare the incorporated modifier in (1) — but no studied cases, and we leave the interaction as an open question: settling it would require elicitation or a larger text base than survives.