Chapter 104Headless, Internally-Headed, and Noun-Modifying Clause Analyses

104.1 Overview

Hokkaido Ainu builds noun-modifying clauses prenominally: a finite verb phrase precedes a head noun with no relative pronoun and no change to the verb's morphology; the relativized argument's role is recovered from a gap in argument structure and from any postpositions retained inside the clause Nakagawa (2024: 503); Bugaeva (2015: 47); Shibatani (1990: 39). When no lexical noun is available or intended as the head, a member of the closed class of formal nouns fills that role, supplying a generic semantic type — 'person', 'thing', 'people', 'place', 'time' — while the clause itself names the referent. In continuous Hokkaido Ainu text these formal-noun heads appear more often than common-noun heads in relativization contexts Bugaeva (2015: 47).

Two further analytical questions organize the later sections. The first is whether any Hokkaido Ainu constructions approach the internally-headed pattern of noun modification — in which the notional head is expressed inside the modifying clause rather than outside it — and whether that pattern involves a form of noun incorporation. Dal Corso raises this for southern Hokkaido Ainu indirect evidential constructions and proposes a classificatory-incorporation analysis Dal Corso (2020: 47, 51–52). The second is whether the perception-complement constructions built with possessed sensory formal nouns are biclausal — an external formal noun heading a relative clause — or whether they have grammaticalized toward monoclausal complex predicates. Both questions remain contested in the literature ‹contested›.

The p ‘thing (nominalizer)’ ~ pe ‘thing (nominalizer)’ formal noun in its RC-head and complementizer uses is treated in Chapter 101 (The Nominalizers -p/-pe and -i/-hi (Participant, Event, Place, Fact)); the gap strategy, the accessibility hierarchy, and postposition stranding in Chapter 103 (Relative Clauses: The Gap Strategy and the Accessibility Hierarchy); lexicalized forms that crystallized from headless relative clauses in Chapter 102 (Lexical-Head and Formal-Noun Nominalization (kur, uske, ruwe/hawe/siri/humi)); and the evidential clause-final uses of the sensory formal nouns in Chapter 120 (The Nominalization-plus-Copula Evidential Schema) and in Chapter 121 (ruwe ne — the Inferential / Visual-Trace Evidential).

104.2 The formal-noun inventory for headless nominalizations

The formal nouns that serve as generic heads form a compact, closed set. Nakagawa lists them as a distinct word class defined by weak lexical autonomy — they follow another element to build an NP — and shows that they carry arity −1, on a par with common nouns in their citation form Nakagawa (2024: 257–258, 261–262). This arity value allows a formal noun to fill the open argument slot of a clause with total arity ≥ +1 and produce a headless relative-clause nominal Nakagawa (2024: 261–262, 504).

Among person-type heads, p ~ pe is maximally generic and extends to human referents with a pejorative or collective edge in those contexts; kur (< kuru ‘person’) selects specifically for a personal referent and carries a polite register. Shibatani describes -kur as a phrasal nominalizing suffix in Classical/yukar Ainu, productive alongside -p(e) Shibatani (1990: 33–34):

The utar head provides person semantics with plurality. Bugaeva's accessibility-hierarchy study records it as a headless head for clauses with a plural human referent:

The p ~ pe formal noun extends to human agents when context provides that reading. Bugaeva attests a pair of parallel headless nominals whose open argument is the subject of a motion verb:

104.3 Headless nominalizations across the accessibility hierarchy

The argument positions available to headless nominalizations are the same as those available in externally-headed relative clauses. Under the noun-phrase accessibility hierarchy of (Keenan & Comrie 1977), Hokkaido Ainu relativizes subjects, agents, objects, and a range of obliques through the gap strategy; the alienable possessor and the standard of comparison resist this because of island constraints and the absence of an accessible comparee position respectively Bugaeva (2015: 47–49). These exclusions carry through to the headless type: there is no headless relativization of an alienable possessor or a comparison standard in Bugaeva's data.

For oblique participants, the governing postposition stays inside the relative clause when the head is a formal noun, as it would with a common-noun head. The place-type formal noun usi ~ uske heads clauses with a locative, dative, or directional gap:

Nakagawa's arity accounting governs the boundary between a headless relative-clause nominal and the complementizer use of the same formal noun. When the clause preceding p ~ pe or hi is already saturated (arity 0), the formal noun nominalizes the event or fact as a whole rather than filling a participant gap Nakagawa (2024: 261–262, 507). The complementizer use is treated in Chapter 106 (Nominalized Complements and Control/Raising) and for p ~ pe specifically in Chapter 101 (The Nominalizers -p/-pe and -i/-hi (Participant, Event, Place, Fact)).

104.4 The appositive construction and the GNMCC question

Shibatani identifies a distinct pattern in Classical/yukar Ainu that he calls the appositive construction: a clause followed by a bare sensory noun in its concept form — without the possessive suffix — where the noun names the perceptual character of the event described by the clause Shibatani (1990: 41):

In the standard description of the Saru/Chitose perception-complement construction, the sensory noun carries the possessive suffix — haw-e 'the voice of—', sir-i 'the appearance of—', hum-i 'the sensation of—' — and the suffix cross-references the preceding finite clause as the implicit possessor Satō (2008: 175–178); Nakagawa (2024: 508). On that analysis the construction is an externally-headed formal-noun relative clause: the head carries arity −1, the possessed suffix makes the semantic link between head and clause overt, and the possessed formal noun is then an argument of the matrix perception verb or the copula.

Bugaeva's relativization study and Dal Corso's subsequent work both observe a distributional difference when the sensory noun appears in its bare (non-possessed) form alongside an existential or posture verb: hum as 'there is a sensation of', siri an 'there is an appearance of' Bugaeva (2015: 100); Dal Corso (2020: 52). In this variant the possessed link that sustains the external-RC analysis is absent. Bugaeva reads the distribution as showing characteristics of a general noun-modifying clause construction (GNMCC) — a type in which a nominal element classifies the clause referent from inside rather than standing as an external head:

The possessed and non-possessed variants have different distributional profiles: the possessed head functions as an argument of a matrix perception verb (e.g. nukar ‘see’, nu ‘hear’), while the bare-noun variant appears as a predication in its own right, typically clause-final. Whether the distributional difference warrants a distinct structural analysis — GNMCC versus external RC — or reflects a stage of bleaching along the same grammaticalization cline is not settled by the sources accessible here ‹corpus-suggested›.

104.5 Dal Corso's internally-headed RC and classificatory-incorporation proposal

Dal Corso develops the GNMCC observation into a formal structural claim specific to the indirect evidential constructions of southern Hokkaido Ainu. He argues that siri ki, siri an, hawe as, and humi as are instances of internally-headed relative clauses (IHRCs) in which the sensory noun is incorporated into the event clause as a classifier rather than serving as an external head Dal Corso (2020: 47, 51–52). The proposal comprises two components: (i) an IHRC claim — the notional head is expressed inside the modifying clause — and (ii) a classificatory noun incorporation (CNI) claim — the mechanism is a type of noun incorporation in which the incorporated noun classifies the event described by the clause Dal Corso (2020: 52).

The argument turns on the contrast between direct and indirect evidential pairs. The direct forms (siri ne, hawe ne, humi ne) carry the possessive suffix and close with the copula ne; the possessive suffix cross-references the preceding clause as the possessor, sustaining an external-RC analysis. The indirect forms replace ne with existential or posture verbs (an, as, ki), and the sensory noun appears without possessive morphology; Dal Corso reads the absence of possessive marking as a diagnostic that the sensory noun is incorporated into the clause as a classifier rather than standing as an external head outside it Dal Corso (2020: 51–52).

The analysis has not been adopted by Bugaeva, Satō, or Nakagawa in their accounts of Hokkaido Ainu ‹contested›. Bugaeva's relativization study treats the same perception complements as noun-complement structures where the possessed nominal head cross-references the clause as a whole, and characterizes the non-possessive variant as an emerging GNMCC pattern rather than a settled IHRC structure Bugaeva (2015: 79, 92, 100). Nakagawa and Satō describe the sensory formal nouns as complementizers of the perception domain without positing an IHRC subcase Nakagawa (2024: 508); Satō (2008: 175–178). The CNI claim further intersects with the broader question of what constitutes noun incorporation in a language where the nominalization-to-lexicalization cline already accounts for many apparent incorporation effects; the incorporation dimension is taken up in Chapter 88 (Object (Patient) Noun Incorporation).

The wider biclausal-versus-monoclausal question — whether [clause] ruwe ne, [clause] hawe ne, and their kin are biclausal (an external formal noun heading the preceding clause, the copula predicating the result: Tamura (2000: 227); Satō (2008: 175)) or are grammaticalizing toward monoclausal complex predicates (Bugaeva (2016: 306–315); Dal Corso (2018: 24–25)) — is addressed with its evidence in Chapter 101 (The Nominalizers -p/-pe and -i/-hi (Participant, Event, Place, Fact)) and Chapter 121 (ruwe ne — the Inferential / Visual-Trace Evidential). No constituency proof in the sources available here settles the question ‹contested›.

104.6 The nominalization-to-lexicalization cline

Headless relative clauses sit at one end of a cline whose other end is a lexical noun. Bugaeva documents a set of Hokkaido Ainu nouns that crystallized historically from [VP]–p or [VP]–i nominalizations Bugaeva (2016: 240–251):

These forms illustrate Bugaeva's Stage II of Hokkaido Ainu nominalization development: a nominalizing word follows a full finite clause and the result gradually freezes into a noun, the clause eventually compressing into a stem Bugaeva (2016: 243, 385–397). The productive synchronic counterpart is the headless RC with p ‘thing (nominalizer)’ ~ pe ‘thing (nominalizer)’; the line between productive construction and lexicalized noun is determined by morphological productivity Shibatani (1990: 33–34).

Nakagawa's diagnostic distinguishes an active relative-clause nominalization from a compound noun: the nominalization retains the postposition or applicative that marks the head's semantic role inside the clause, while a compound leaves the relation between its members lexically implicit. The contrast with instruments is sharpest: a=e-suke-p 'kitchen utensil' retains the applicative e- that licenses the instrument, whereas the plain compound sukemakiri 'kitchen knife' does not Nakagawa (2024: 506–507). The lexicalized inventory is treated in Chapter 102 (Lexical-Head and Formal-Noun Nominalization (kur, uske, ruwe/hawe/siri/humi)).

Ijäs observes that place nouns such as ipeusi ‘restaurant’ derive historically from headless relative clauses with usi ‘place’ as the head, and that lexicalized usi-type nouns need no accompanying relational noun when used synchronically, unlike the productive headless-RC construction Ijäs (2023); Nakagawa (2024: 258). The observation fits Bugaeva's Stage-II account and the productive/fossilized distinction Nakagawa draws Bugaeva (2016: 245–251).

Dal Corso's CNI proposal connects the cline back to the IHRC discussion. If the bare sensory noun in the indirect evidential constructions is an incorporated classifier, it occupies an inside-clause position analogous to the nominalizing element in the lexicalization cline, rather than the external-head position of the possessed formal noun. The proposal would extend the incorporation-nominalization interface already visible in the lexicalized nouns above to a productive grammatical construction, making classificatory incorporation a third strategy alongside externally-headed and headless relative clauses. Whether this extension is warranted by the southern Hokkaido data, or whether the indirect evidential constructions are better handled as partial grammaticalization of the external-RC pattern, is among the questions Dal Corso raises without fully settling Dal Corso (2020: 52) ‹contested›.