Chapter 22Complementation

22.1 Bare complements of modal and aspectual predicates

The desiderative rusuy, potential easkay and its negatives koyaykus and eaykah, and completive hemaka take a bare verb as their complement — a zero-nominalization, the oldest layer of the system (Bugaeva 2016: 99, 113) (Chapter 15 for their semantics). Furukawa's Rayciska data show the diagnostic property: the person marking of the whole sentence migrates to the modal predicate, leaving the notional verb bare — her transformational rule reorders ku-oman koyaykus into the usual

The complement may be flagged with the additive particle ka ‘even’, and with the lexically negative predicates of §17.4 this yields fixed constructions: V ka erameskari ‘have never V-ed’ (experiential), V ka etunne ‘not want to V’, V ka koyaykus ‘be unable to V’ (Dal Corso 2025b: 40–41). The negated light-verb construction of §17.2 — nominalized verb + ka + hanki — has exactly this complement syntax, with ‘do’ as the matrix predicate.

22.2 Irrealis complements with kun(i)

Unrealized, intended, or obligatory events are embedded with kun(i), which we gloss COND following its conditional-irrealis value; it combines with the nominalizers as kun-i and kun pe (§20.1). Under verbs of thinking and intending the kun-i form supplies the complement:

With the adverbializer no, the same form makes purposive complements and adjuncts, ‘so that, in order to’; the oldest West-coast texts show it under negation:

The same texts show the assertive future complex nankoro pe ne ‘it will (probably) be the case that…’, where the dubitative, the bound noun pe, and the copula stack over the clause — complementation grammaticalized into modality (Chapter 15):

22.3 Quotation and reported speech

Speech and thought are reported as direct quotation: the quoted clause stands first, the quotative particle nah closes it, and a verb of saying or doing follows — nah yee ‘say so’, nah kii ‘go like this’ (of speech and sounds), nah ramu ‘think so’. Inside quotations the deixis is that of the original speaker, and in folktales the quotative frame carries the logophoric use of the indefinite person (§13.2; §16.7):

That logophoric convention deserves spelling out, because it is the closest the language comes to a dedicated reported-speech grammar. When a character reports speech in which he or she figures, the quoted speaker is encoded with the indefinite (fourth-person) prefixes rather than with the first person: the “I” inside the quote is grammatically held apart from the “I” of the narrator. Chiri noticed the effect and remarked that the Sakhalin indefinite inside quotations comes close to a third-person value — so close that he was inclined to call it a third-person affix outright (Chiri 1942: 764) — citing a cry from Piłsudski's East-coast texts in which the quoted speaker's ‘me’ is the indefinite object prefix:

Sakhalin Ainu thus has no logophoric pronoun series of the West African type — no form reserved for the reported speaker and nothing else. What it has is the independent fourth-person category recruited for the logophoric function, disambiguating the speech-internal from the speech-external participant without new morphology. The recruitment is the diachronically expected one: indefinite and fourth-person categories are a recognized source of logophoric marking cross-linguistically, and the Sakhalin texts show the source construction still transparent, quasi-logophoric rather than grammaticalized (§13.2).

Indirect speech as a distinct construction is not attested: even reports embedded under nah ramu ‘think’ keep direct-speech deixis. The reportative manu that ends narrative sentences is the same strategy scaled up to the whole text — the tale is one long quotation (Chapter 16).

22.4 Perception and knowledge complements

Perception verbs take a nominalized clause as object. The commonest frame in the narrative corpus is the possessive nominalization plus topic particle, V-POSS neampe, literally ‘(as for) my seeing — …’, which serves at once as a perception report and a temporal pivot ‘when I looked’:

We analyse the construction as a nominalized complement that has been topicalized out of its matrix clause: the perceived content is then asserted as a fresh main clause, exactly parallel to the presuppositional main-clause nominalizations of §20.5. Where the perceived content is expressed with the evidential nouns — ‘hear the voice of X V-ing’, ‘see the sight of…’ — the complement structure and the evidential construction coincide (§16.4; (Bugaeva 2016: 101–102)). Knowledge predicates pattern with the modals: wante ‘know’ and its negative erameskari take bare or ka-marked zero-nominalized complements (§22.1), as in the experiential construction V ka erameskari ‘have no experience of V-ing’ (Dal Corso 2025b: 40–41). The East-coast texts show that the cognition class also reaches the possessive nominalization: Piłsudski's yay-ikoramopi-hi kayki an-eramuskari ‘I do not even know what (she) thinks of me’ embeds a V-POSS form as the object of the negative knowledge verb (Piłsudski 1912), so the strategy is not confined to the innovating West coast in its embedded uses (§20.5).

22.5 Evaluative predicates and clausal subjects

A clause can also be the subject of a higher predicate. The selecting class is the one-place evaluative and veridicality predicates — pirka ‘be good’, wen ‘be bad’, sunke ‘be untrue, lie’ — which predicate something of a state of affairs rather than of a participant. On the West coast that state of affairs is packaged as the possessive nominalization: in ku=araka-ha pirika ‘my being sick got better’, the clause ku=araka-ha occupies the subject slot of pirika (the example is given in full in §20.2), and in ku=ye-he sunke ‘what I said is untrue’ the nominalized clause is what the veridicality predicate judges false (Chapter 7; Murasaki (1979: 95, 141); Bugaeva (2016: 108)).

The construction is the structural opposite of the modal class of §22.1. There the complement is stripped bare and the person marking migrates onto the matrix predicate; here the complement keeps its own person marking and carries full nominalizing morphology, while the matrix contributes nothing but a plain one-place predicate. The two classes show no attested overlap: no evaluative predicate is attested with a bare-verb complement, and no desiderative or potential with a V-POSS one. The attested evaluative class is admittedly small — only pirka and sunke are secure in the corpus as matrices of a nominalized subject clause. Its semantics, however, defines a natural class, and we predict that its remaining members, most obviously the antonym wen ‘be bad’, select the same V-POSS subject clause; a larger conversational sample would test the prediction.

22.6 The selectional system

Set side by side, the strategies of this chapter are not an arbitrary list: which one a matrix predicate selects follows from how tightly its meaning binds it to the complement event.

The gradient is the iconic one of Givón's binding hierarchy (Givón 1980): the tighter the semantic bond between matrix and complement, the less nominalizing morphology the complement carries. The subject-sharing desideratives and modals, whose two events are in effect one, take a stripped verb welded into a single complex predicate; the evaluatives, which assert something about an independent but presupposed event, take a fully nominalized clause; and the utterance verbs, whose complement is an autonomous speech act, take an unreduced direct clause merely flagged by nah. The cognition verbs straddle the loose end of the scale, sharing the quotation frame with yee and the nominalizations with the evaluatives — as expected of predicates whose semantics ranges from quoted inner speech to settled knowledge. Within Noonan's typology, Sakhalin Ainu thus runs its entire complementation on sentence-like (quoted) and reduced or nominalized complement types (Noonan 2007); the finite indicative complement that English builds with that simply has no Sakhalin counterpart. The chief residual uncertainty is the size of the evaluative class (§22.5).