Chapter 106Nominalized Complements and Control/Raising

106.1 Overview

Clausal complements in Hokkaido Ainu are formed through the same nominalization strategy that builds relative-clause NPs: a finite, person-marked clause is followed by a formal noun or complementizer that turns the whole into a nominal argument of a higher predicate. The verb inside the embedded clause keeps its ordinary person affixes, polarity, and aspect morphology; no non-finite or participial form exists Bugaeva (2016: 100, 226–228); Nakagawa (2024: 503). What separates a complement clause from a relative clause is a single structural condition, the arity of the embedded clause, treated in the section that follows. The inventory of complementizers includes the abstract formal noun hi ‘thing, fact, time, place’, the sensory possessive forms siri ‘appearance’, hawe ‘voice’, and humi ‘sound, feeling’, the concrete nominalizer p ‘(thing)’ ~ pe ‘(thing/person)’, and the irrealis complementizer kuni ‘(purposive/deontic)’, together with a set of indirect-speech markers treated in Chapter 105 (Quotative Complementation and Reported Speech (sekor)). Their grammaticalization from lexical nouns and their role in evidential clause-final constructions are set out in Chapter 35 (The Formal / Defective Noun Set (ruwe, hawe, siri, humi, hi, pe, kur)) and Chapter 120 (The Nominalization-plus-Copula Evidential Schema); p ~ pe as a headless relative-clause nominalizer is treated in Chapter 101 (The Nominalizers -p/-pe and -i/-hi (Participant, Event, Place, Fact)), which cross-references the present chapter for the complementizer use.

106.2 The arity-0 diagnostic

Nakagawa's arity (アリティ) accounting provides the structural boundary between a relative clause and a complement clause. Every NP in the clause contributes −1 to a running count; every argument slot required by the verb contributes +1. A formal noun itself has arity −1. When the clause before a formal noun has total arity ≥ +1, a participant slot remains open and the formal noun fills it as the head of a relative-clause NP. When the clause already has arity 0 — every argument slot is satisfied — the formal noun nominalizes the whole clause as an event or fact; this is the complement reading Nakagawa (2024: 261–262, 504, 507).

Satō's formulation of the same distinction is categorical: a nominalizer (名詞化辞) nominalizes a whole content clause and allows no reconstruction of an underlying sentence, whereas a formal noun (形式名詞) heading a relative clause can be so reconstructed Satō (2008: 174–175). Nakagawa prefers to treat the formal nouns as a word class and the complementizer function as a syntactic role available to some members of that class Nakagawa (2024: 508). The same surface string [clause] hi or [clause] p is parsed differently depending on which reading the arity yields; examples below illustrate both.

The following illustrates the complement reading of hi. The embedded clause hekattar tura sinot=an has arity 0: sinot 'play' is an intransitive verb (+1) whose S slot is satisfied by the fourth-person suffix =an (−1), leaving no gap. hi therefore nominalizes the whole clause as a fact:

106.3 Complement-taking predicates

The matrix predicates that select a nominalized complement fall into three semantic groups: cognition, perception, and speech/directive. A verb's argument structure places an upper bound on the number of person affixes it can carry — typically one for the agent (A) and one for the object (O); the complement clause then occupies the single object slot Nakagawa (2024: 503–509). This is directly related to the two-affix cap on the verb template.

The list above is drawn from Nakagawa and Satō and is not exhaustive Nakagawa (2024: 475–480, 503–513); Satō (2008: 175–178). The complementizer preferences are typical but not always categorical: ramu, for instance, co-occurs with kuni for future and wished content, with yak for past or ongoing content, and with the sensory heads sir-i and hum-i when thought involves a quasi-perceptual quality ‹corpus-suggested›.

106.4 hi as complementizer

The formal noun hi ‘thing, fact, time, place’ is the maximally abstract complementizer. Satō describes it as carrying "almost no concrete meaning" and functioning to nominalize the entire clause before it Satō (2008: 169, 174); Nakagawa calls it a 補文標識 (complementizer) when its preceding clause has arity 0 Nakagawa (2024: 507). In this role it appears most frequently as the object of cognition predicates such as eramiskari 'not understand' and kopan 'refuse':

A nominalized clause headed by hi can also serve as the grammatical subject of a stative predication. Bugaeva cites (after Tamura) a clausal-subject construction in which a coordinated VP is nominalized by hi and predicated as 'best':

The Okuda–Nakagawa analysis of hi distinguishes a 'place where' reading (arity ≥ +1, RC head, requires a case postposition such as oro ta) from a 'time/thing/fact' reading (arity 0, complementizer, no postposition required) Nakagawa (2024: 510–511). This distinction is a tendency, not absolute: the Horobetsu example seta utar osoma hi 'the dogs' privy' (literally 'the place dogs defecate') lacks oro ta yet takes the 'place' reading Nakagawa (2024: 511) ‹contested›. The 'place' analysis rests on Okuda (1989), read only second-hand via Nakagawa here.

106.5 The sensory nominalizers as perception complements

The four possessed-form nominalizers ru-we, sir-i, haw-e, and hum-i function primarily as clause-final evidential heads (treated in Chapter 120 (The Nominalization-plus-Copula Evidential Schema) and Chapter 121 (ruwe ne — the Inferential / Visual-Trace Evidential)), but Satō shows that three of them — sir-i, haw-e, and hum-i — also nominalize embedded clauses as perception-object NPs under nukar 'see' and nu 'hear/ask', with a semantic specialization: sir-i marks an event in progress at reference time (visual perception), haw-e a speech event heard, and hum-i a non-visual sensation Satō (2008: 175–178); Nakagawa (2024: 508).

In this use the possessed-form nominalizer takes a saturated clause as its implicit possessor: the 'appearance/voice/feeling of [the embedded event]'. The whole [clause] sir-i or [clause] haw-e then fills the O slot of the perception verb:

Nakagawa observes that these four evidential nominalizers constitute "one type of complementizer" alongside the non-evidential complementizers; the clause-final 文末句 use (where they are closed by the copula ne or existential an) is the same construction extended to main-clause predication Nakagawa (2024: 508). The Dal Corso debate about whether these structures are biclausal or partially grammaticalized monoclausal predicates applies equally here and is taken up in the constituency section below.

106.6 kuni as complement marker

The complementizer kuni ‘(purposive/deontic)’ nominalizes the preceding clause and expresses content that is future-oriented, inferred, desired, or directed: what the matrix subject says must happen, thinks should happen, or intends to happen Nakagawa (2024: 476). Nakagawa classifies kuni as a conjunction with formal-noun-like use (接続助詞の形式名詞的な用法); Bugaeva analyses the complex kuni-p as deontic 'what should be' Bugaeva (2016: 294). The purpose and adverbial uses of kuni are treated in Chapter 107 (The kuni Complementizer and Purpose Complements); the focus here is on the complement-of-speech and complement-of-cognition uses.

With ye 'say', kuni marks a directive or an anticipated future event reported in speech:

With ramu 'think', kuni expresses a wished or expected state of affairs:

The combination kuni p forms a deontic or purposive nominal Bugaeva (2016: 294). The embedded clause is nominalized by kuni and then a second formal noun p further nominalizes the resulting complex as 'a thing one should / will do':

The corpus count of kuni is 6,362 tokens (rank 38), a figure that conflates the complementizer, purposive, and adverbial uses; no per-function count is derivable from word-frequency data alone ‹corpus-suggested›.

106.7 Other indirect-speech complementizers

Beyond kuni, four elements function as complementizers in indirect speech and thought contexts, each encoding a different epistemic or temporal profile Nakagawa (2024: 475–480); Satō (2008: 176–177):

A shared diagnostic: all of these elements nominalize the preceding VP and make it the object of the matrix transitive. The form yakne is not an indirect-speech marker — it is a conditional conjunction, distinct from the complementizer yak Nakagawa (2024: 477). Final particles of the type used in main clauses cannot appear inside these embedded complements, and person marking shifts from the original speaker's viewpoint to the narrator's Nakagawa (2024: 475). The kunak and yak complementizers, and the quotative sekor, are treated in detail in Chapter 105 (Quotative Complementation and Reported Speech (sekor)).

The embedded question complementizer ya introduces content questions and polar questions as the object of cognition predicates such as eramiskari 'not know'. Nakagawa distinguishes the polar-question use ('whether') from the content-question use ('what/where/when'), both employing ya Nakagawa (2024: 510).

106.8 Bare complementation

A small residue of complement constructions in the corpus involves no overt nominalizer: the embedded clause appears without a following formal noun, directly as the argument of the matrix predicate. Bugaeva identifies this as Stage I of her diachronic typology of Ainu nominalization — the oldest stratum, in which a bare finite clause functions nominally without a nominalizing word Bugaeva (2016: 196–200, 229). Refsing similarly notes cases in Shizunai Ainu that she tentatively treats as "ellipsis of the nominalizer" Refsing (1986). The construction is residual in the Hokkaido attested record and not systematically documented for modern Saru or Chitose ‹corpus-suggested›. Its relationship to compounding and to the lexicalized nominalization cline (where p ~ pe has contracted into lexical items such as aep ‘food’) is treated in Chapter 101 (The Nominalizers -p/-pe and -i/-hi (Participant, Event, Place, Fact)) and Chapter 102 (Lexical-Head and Formal-Noun Nominalization (kur, uske, ruwe/hawe/siri/humi)).

106.9 Participant identity across the clause boundary

Because complement clauses in Hokkaido Ainu are finite and carry their own person morphology, the embedded subject is always expressible: the verb inside the complement bears its own person affixes independently of the matrix clause. No PRO or empty-category mechanism is posited in the sources consulted, and no formal control or raising analysis of HA complements appears in the literature surveyed here.

Nevertheless, patterns of participant identity are observable. With cognition predicates (ramu, eramiskari), the person marking of the embedded clause in the corpus tends to match the matrix subject: the one who thinks or fails to understand is also the agent of the thought or the one about whom the proposition holds. The examples above show this — ku=ramu 'I think' with an embedded clause about a state of affairs the speaker cares about, a=eramiskari 'one does not understand' with an embedded clause about what that same person ought to say or do. The regularity, however, follows from the semantics of cognition verbs, not from a grammatical control mechanism, and divergent persons are possible where the content demands it ‹corpus-suggested›.

With the directive use of ye + kuni, the pattern runs in the opposite direction: the matrix subject (the one who speaks) addresses the matrix object (the addressee), and the embedded clause's agent is typically that addressee, not the speaker. The corpus example from Kawakami Matsuko above — 'when I told them to enter' — shows the fourth-person speaker (a=ye) directing a different group (the addressees) toward an action (ahun 'enter') whose agent goes unmarked in the embedded clause because it is recoverable from discourse ILCAA Ainu materials (1976) ‹corpus-suggested›. The indirect-speech analysis that best captures this is given in Nakagawa's treatment of indirect speech perspective shifts Nakagawa (2024: 475–476).

The term raising (movement of an embedded subject to a matrix position) has no attestation in the analytical literature on Hokkaido Ainu clausal complementation, and the available data supply no diagnostic that would motivate such an analysis. The embedded clause remains a nominalized finite clause whose internal arguments are expressed by the embedded verb's own person affixes ‹speculative›.

106.10 The constituency of the nominalized clause

Whether the structure [clause] NMLZ ne is biclausal or a grammaticalizing monoclausal complex predicate is a live debate that cuts across complement clauses, evidential constructions, and the formal noun system generally. The two positions are:

The biclausal external-RC analysis treats the nominalized clause as a full NP whose head is a possessed formal noun. The copula ne then closes the outer predication. Tamura, Okuda, and Satō take this position Tamura (2000: 227); Satō (2008: 175).

The grammaticalization-to-monoclausal analysis holds that the erstwhile noun is considerably bleached and the construction is moving toward a complex predicate. The diagnostic is that a fully lexical noun can take demonstratives and attributive modifiers (pirka menoko 'a beautiful woman'), whereas the evidential formal noun cannot: *pirka ruwe ne kewtum is ungrammatical Bugaeva (2016: 306–315); Dal Corso (2018: 24–25). No read source provides a constituency proof — a bracket test or movement diagnostic — that decides between these ‹contested›. The same unresolved question is noted in the dedicated treatments of ruwe ne in Chapter 121 (ruwe ne — the Inferential / Visual-Trace Evidential) and of p ~ pe in Chapter 101 (The Nominalizers -p/-pe and -i/-hi (Participant, Event, Place, Fact)).

For the complement constructions treated in this chapter, the practical consequence is that the formal nouns function synchronically as clause-final markers that package a finite clause as a nominal argument. The diachronic source in lexical nouns remains transparent (Bugaeva's nominalization-cline stages, (Bugaeva 2016: 229–304)), and partial grammaticalization explains why the same forms build RC heads, complement clauses, evidential assertions, and — at the furthest end of the cline — conjunctions and adverbial-clause markers Nakagawa (2024: 515–516); Takahashi (2018: 107–112).

Ijäs's pedagogical grammar introduces relative clauses and notes that formal nouns such as uske 'place/time' are best understood as nouns that have historically originated from relative-clause constructions, without committing to a formal analysis of the biclausal-vs-monoclausal debate Ijäs (2023). This aligns with the historical-cline view and is graded as consistent with the consensus on diachrony ‹consensus›.