Chapter 102Lexical-Head and Formal-Noun Nominalization (kur, uske, ruwe/hawe/siri/humi)

102.1 Overview

Hokkaido Ainu forms nominalized clauses by placing a formal noun after a finite predicate. The verb retains its ordinary person marking; the formal noun supplies the nominal head and converts the preceding material into an NP. The class is closed and its members differ in the semantic profile they impose — 'thing', 'person', 'place', 'appearance', 'voice', and so on — and in whether they function as relative-clause heads, complementizers, or both Nakagawa (2024: 257–262, 503–513); Satō (2008: 174–180); Bugaeva (2016: 226–228).

The allomorphs p ‘thing, person (nominalizer)’ ~ pe ‘thing, person (nominalizer)’, their participant-nominalization and complementizer functions, and the arity diagnostic that distinguishes the two readings are treated in Chapter 101 (The Nominalizers -p/-pe and -i/-hi (Participant, Event, Place, Fact)). The gap strategy, accessibility hierarchy, and postposition stranding that characterize all relative clauses are covered in Chapter 103 (Relative Clauses: The Gap Strategy and the Accessibility Hierarchy). The clause-final evidential uses of ruwe ‘the trace of’, siri ‘the appearance of’, hawe ‘the voice of’, and humi ‘the sound/feeling of’ are treated in Chapter 120 (The Nominalization-plus-Copula Evidential Schema). The present chapter addresses the morphological contrast between the concept form and the affiliative (possessed) form of these nouns, the semantic profiles that govern head-selection across the formal-noun inventory, hi ‘thing, place, time’ in its several uses, the sensory-noun formal nouns as non-evidential perception-complement heads, and the cline from clausal nominalization to lexicalized noun.

102.2 Concept form and affiliative (possessed) form

Each formal noun has two shapes described in Chapter 44 (Concept Form vs. Affiliative (Possessed) Form). The bare concept form (概念形) is the citation form of the noun, autonomous in a sentence. The affiliative form, or possessed form (所属形), marks the noun as belonging to an antecedent; it is the shape in which the sensory-noun formal nouns appear in their evidential and perception-complement uses. For the four sensory nouns the two forms are distinguished by a possessive suffix Nakagawa (2024: 258, 表66):

Satō observes that a noun in its possessed form presupposes an antecedent — something it belongs to — so that placing a sensory noun in this form after a finite clause grammatically encodes that the described situation is already held as given by the speaker Satō (2008: 135). This presupposition distinguishes evidential ruwe ('the trace of —', implicating a recognized antecedent event) from lexical ru ('a track'). Dialect evidence reinforces the contrast: southern Hokkaido (Saru, Chitose) uses the possessed forms in the clause-final evidential frame, while Tokachi and Ishikari close the clause with the bare concept forms ru, sir, haw, hum Nakagawa (2024: 110, 390); Takahashi (2013: 129–131).

102.3 Principal formal nouns as nominal heads

All formal nouns have arity −1, which allows them to terminate a relative clause (closing an arity ≥+1 gap) or a saturated complement clause (arity 0) alike Nakagawa (2024: 261–262). The table below covers those relevant to nominalization; the full catalogue with arity derivations appears in Chapter 35 (The Formal / Defective Noun Set (ruwe, hawe, siri, humi, hi, pe, kur)).

The pair usi / uske are dialectally distributed: uske is attested in Saru, Chitose, and Shizunai; the forms usi and usike are more widely spread across Hokkaido Nakagawa (2024: 258). Both head relative clauses without requiring the relational noun or that movement-case postpositions need before locative complements Bugaeva (2015: 47). Shibatani lists the nominalizers -p(e), -i, -ike, and -kur as productive phrasal and clausal nominalizating suffixes in classical/yukar Ainu Shibatani (1990: 33–34), corresponding to the word-class of formal nouns in Nakagawa's synchronic description Nakagawa (2024: 257–258).

102.4 Division of labour across the inventory

The formal-noun head selected for a nominalized clause is determined primarily by the semantic category of the reference being expressed. p ~ pe and hi cover the widest range but along different dimensions: p ~ pe profiles a concrete entity — a thing or person that can be encountered or handled — while hi is maximally abstract, covering fact, time, and place readings alike Nakagawa (2024: 257); Satō (2008: 174). Where the referent is specifically a person and an honorific register is required, kur (from kuru ‘person’) is used; utar extends this person head to a group. The place nouns usi(ke) and uske head nominalizations that refer to a location or time without the concrete/abstract ambiguity of hi.

Satō's diagnostic sharpens the contrast between the relative-clause use and the complementizer use for any formal noun. A formal noun functioning as a relative-clause head heads a clause from which a participant has been gapped; the original sentence can be reconstructed by reinserting that participant. A formal noun functioning as a complementizer nominalizes the whole clause as a proposition, and no such reconstruction is possible Satō (2008: 174–175). Both p ~ pe and hi fill either role depending on the arity of the clause they follow; the sensory-noun heads serve both functions as well, though their complement use is semantically restricted to perception contexts.

Person formal noun kur produces nominalizations whose referent is a specific individual. In classical and epic texts these regularly lexicalize: kotan kor-kur 'chief' (lit. 'person who has a village'), ray-kur 'the dead' (lit. 'person(s) who die') Shibatani (1990: 34). Compound nominalizations of this type remain analytically transparent even after lexicalization, illustrating the continuity between productive RC nominalization and the lexicalized stratum treated in the final section.

102.5 hi — semantic range and the place/time split

hi ‘thing, place, time’ is the most abstract formal noun in the inventory; its gloss covers 'thing, matter, fact', 'time when', and 'place where'. After a consonant it surfaces as -i — as in the lexicalized ape-o-i 'hearth', treated in the next section Nakagawa (2024: 257); Satō (2008: 169).

In the 'thing/fact' and 'time' readings, hi nominalizes a saturated clause (arity 0) and functions as a complementizer. The nominalized clause may serve as the grammatical subject of a matrix predicate:

Or as the object of a complement-taking predicate. In the following example, hi appears twice: as the object-head of a=ye 'say' (arity +1, 'what to say') and of a=kar 'do' (arity +1, 'what to do'). In each case the clause before hi has an open slot, so hi fills the head rather than the complementizer role:

The 'place' reading calls for a different structure: the governing postposition is retained inside the relative clause, marking hi as the locative head (arity +1 because the location slot is gapped):

With the time and abstract-fact readings, the clause is saturated and hi nominalizes it whole. The postposition that follows hi is not part of the RC but a postposition governing the entire nominalized clause:

Nakagawa frames the place/non-place split as a tendency. The 'place' reading is more concrete and closer to an ordinary noun, so the grammatical relation requires overt expression; the 'time/fact' reading is abstract and freely nominalizes the whole clause. Okuda (1989), quoted and refined by Nakagawa (2024: 510–511), first systematized this contrast. An attested exception — seta utar osoma hi 'the dogs' privy', from 知里幸恵's kamuy yukar collection (Horobetsu), which uses a 'place' reading without a retained postposition — shows the constraint does not hold categorically Nakagawa (2024: 511) ‹corpus-suggested›.

The locative formal nouns usi and uske offer an alternative to hi for the 'place' reading. Because they are inherently locative, they require no retained movement-case postposition. The directional postposition eun 'toward' is instead retained for the oblique head:

Ijäs observes that usi-headed compound nouns — such as ipeusi 'restaurant' (lit. 'eating-place') — are historical RC nominalizations, and that usi-words require no relational noun or before movement-case postpositions Ijäs (2023) ‹consensus›. A locative relative clause with a plain common noun head does require the relational noun:

102.6 Sensory-noun formal nouns as perception-complement heads

Apart from their clause-final evidential function (treated in Chapter 120 (The Nominalization-plus-Copula Evidential Schema) and the inferential ruwe ne), the possessed-form sensory nouns ruwe, siri, hawe, and humi function as nominal heads of complement clauses whose matrix predicate is a perception or cognition verb. In this intra-sentence use, the matrix verb follows the formal noun rather than closing the clause Satō (2008: 175–178).

Each formal noun is selected by a specific subset of perception predicates: nukar 'see' takes siri (the visual scene) or hi (an abstract perceived content); nu 'hear, ask about' takes hawe (verbal content), humi (a non-verbal sensation), or hi; cognition predicates eramiskari 'not know' and eraman 'understand' take hi or p ~ pe Satō (2008: 175–178). The complement-taking predicates and the structural properties of these complements are treated in Chapter 106 (Nominalized Complements and Control/Raising).

Within the visual domain, Satō distinguishes the two available heads: siri applies to an action or situation in progress at reference time — a scene being witnessed — while ruwe applies to an already-existing state or non-progressive action asserted as recognized fact Satō (2008: 178). In the following example the speaker uses siri to frame the visible absence of large trees as the scene that prompted anger:

The shift from this intra-sentence perception complement (matrix verb follows siri) to the clause-final evidential frame (copula ne follows) instantiates Bugaeva's Stage II grammaticalization path: the nominal head that was an argument of a perception predicate becomes the terminal element of an evidential construction Bugaeva (2016: 229–304). The two uses remain synchronically productive and distinct; the evidential use is the more grammaticalized endpoint ‹consensus›.

The constituency of the whole construction — whether it is biclausal (a possessed noun heading the preceding finite clause, closed by the copula) or is grammaticalizing toward a monoclausal complex predicate — is unsettled. The biclausal analysis is held by Tamura, Okuda, and Satō, and is endorsed by Nakagawa under the label 補文標識 'complementizer' Nakagawa (2024: 508); Satō (2008: 175). Bugaeva argues that the erstwhile nouns are considerably grammaticalized and the construction is moving toward a monoclausal complex predicate, supported by the fact that the formal noun no longer heads a full NP (*pirka ruwe ne kewtum is ungrammatical — no determiner or attributive modifier can attach) Bugaeva (2016: 306–315); Dal Corso (2018: 24–25) ‹contested›.

102.7 Lexicalized formal-noun nominalizations and the RC–compound boundary

A relative-clause NP of the form [(NP…) V nmz] can crystallize into a single lexical noun when the constituent parts are reproduced as a unit with a conventional, fixed meaning. Bugaeva traces a grammaticalization cline from productive RC nominalization to lexicalized noun in the Hokkaido data, placing Hokkaido at Stage II of a broader areal typology (Stage I = zero-nominalization; Stage II = nominalizing word; Stage III–IV in Sakhalin = possessed verbal forms) Bugaeva (2016: 240–251, 385–397). The lexicalized products are morphologically transparent but synchronically function as ordinary nouns.

Two items illustrate the pattern with allomorphs of different formal nouns:

The allomorph split across these two items reflects the same head-selection principle visible in productive clausal nominalizations: -p ~ -pe yields a concrete entity ('thing killed', 'thing cooked'), while -i (allomorph of hi) yields a location ('place where fire enters'). The semantic specialization survives lexicalization.

A relative-clause NP and a compound noun are superficially similar — both consist of a predicate-like element preceding a noun — but the arity test separates them Nakagawa (2024: 506–507). The compound tureptani 'uba-lily digging-stick' juxtaposes turep 'uba-lily' and tani 'digging-stick' without marking the semantic relation between them; the RC reading 'the stick with which that woman dug uba-lilies' requires the instrumental postposition ani 'with' to be retained inside the clause, making the instrument role explicit:

The diagnostic generalizes: in an RC nominalization the semantic relation between the formal-noun head and the predicate is grammatically specified — via a retained postposition, the verb's argument structure, or an applicative prefix. In a compound noun, that relation is implicit. Lexicalized RC nominalizations such as ci=ronnu-p and ape-o-i were once transparent RC structures; the grammatical marking has become inert but remains recoverable from the morphology.

Ijäs's observation that usi-words such as ipeusi 'restaurant' ('eating-place') descend historically from RC structures aligns with Bugaeva's Stage II account and with Shibatani's finding that -p(e), -i, -ike, and -kur function as both phrasal and lexical nominalizers Ijäs (2023); Shibatani (1990: 33–34); Bugaeva (2016: 240–251) ‹consensus›. The productive and lexical levels are not sharply bounded; what counts as a new RC formation versus a lexicalized item depends on the frequency and semantic conventionalization of the combination.