Chapter 125Mirativity, Evidential Scope, and Grammaticalization to Sentence-Final Particles

125.1 Overview

Three properties of the Hokkaido Ainu evidential system are treated here because they apply across all four formal-noun evidentials rather than belonging to any single term: the expression of mirativity and its structural separation from the evidential nominalizers; the scope of the evidential predicate over the TAM-marked clause it closes; and the grammaticalization cline by which ordinary sensory nouns become clause-final formal-noun evidentials. Double-evidential stacking is treated alongside grammaticalization because it reveals the hierarchical position that ruwe ne has reached on that cline. The individual evidentials are described in Chapter 120 (The Nominalization-plus-Copula Evidential Schema), Chapter 121 (ruwe ne — the Inferential / Visual-Trace Evidential), Chapter 122 (siri ne — the Direct-Perception / Situational Evidential), Chapter 123 (hawe ne — the Reportative / Hearsay Evidential), and Chapter 124 (humi ne — the Non-Visual Sensory Evidential). Bugaeva (2022)

125.2 Mirativity and the auxiliary aan

Surprise at the content of a proposition — mirativity — is not encoded by any of the four formal-noun evidentials. The mirative value in the Hokkaido corpus attaches instead to the perfective auxiliary aan and to the related form anan Hirosawa (2026: 90). Dal Corso (2018: 336) states this explicitly: Hokkaido Ainu shows no systematic evidence that the evidential paradigm itself generates mirative meanings, in contrast to some Sakhalin configurations. The mirative reading arises from aan/anan, not from the choice of evidential nominalizer.

Yoshikawa (2018) argues that aan encodes not just mirativity but also indirect evidentiality — placing the information at a remove from the speaker's direct epistemic access — and that the two meanings are inseparable in this auxiliary Hirosawa (2026: 90) ‹contested›. Dal Corso (2018: 336), working from a different theoretical framing, finds no grounds for aligning aan with the indirect evidential paradigm and treats its mirative effect as pragmatic. The debate turns on sources consulted second-hand; Yoshikawa's two studies are cited here via Hirosawa (2026: 90). What the accessible sources agree on is the descriptive fact: aan/anan is the element that contributes mirativity ‹corpus-suggested›.

Within a clause the two elements can divide the semantic labour. Hirosawa (2026: 90) (ex. 33) analyses a Saru narrative example in which aan places content outside the speaker's prior epistemic territory — the structural basis of the surprise — while a following ruwe ne reclaims that same content as established fact, producing the 'it turned out that…' reading. Hirosawa reads the sequence as a territory-conflict pattern parallel to Japanese みたい followed by のだ: outward epistemic displacement, then incorporation ‹contested›.

The same anan follows double-evidential constructions when a mirative reading accompanies an indirect evidential stacked under ruwe ne:

An early attestation of the mirative tail with the non-visual sensory evidential is preserved in Kindaichi & Chiri (1936: 192): rupus humi ne awan 'it had completely hardened and frozen', where the past-mirative tail awan (a form historically related to or identical with aan) follows humi ne. This pattern — any of the four direct evidentials followed by a mirative tail — recurs in the corpus, confirming that mirativity is an independent layer above the evidential rather than a property of any particular nominalizer.

The pedagogical grammar Ijäs (2023) presents the evidential nominalizers alongside perfective and mirative auxiliaries without foregrounding the structural distinction between them. The primary analyses Dal Corso (2018: 336); Hirosawa (2026: 90) agree that this distinction is grammatically real, even where they diverge on the semantics of aan itself.

125.3 The scope of the evidential predicate

All four formal-noun evidentials share the same scope geometry: the evidential predicate (formal noun + copula or existential verb) closes a fully finite, TAM-marked clause. The constituent before the evidential carries its own person marking, suppletive number alternations, aspect and perfective auxiliaries, and propositional negation; the evidential then takes scope over the entire embedded clause Nakagawa (2024: 392–393); Takahashi (2022).

Propositional negation — the denial of the embedded proposition — is expressed by negating the verb before the formal noun; the pattern …somo ki ruwe ne predicates the negative proposition as a recognized fact, with ruwe ne outside the scope of somo. The same holds for all four siblings: …somo ki siri ne, …somo ki hawe ne, …somo ki humi ne. The outer rhetorical frame ruwe somo ne is categorically different — it negates the evidential predication as a whole — and is discussed in Chapter 121 (ruwe ne — the Inferential / Visual-Trace Evidential) Nakagawa (2024: 393).

The ordering of the evidential with respect to other clause-level material follows a fixed hierarchy. For ruwe ne, the clause-linkers wa, hine, akusu, kusu, and nominal heads hi, p all follow the evidential; the purposive modal kusu ne takes the evidential as its complement (kusu ne ruwe ne is grammatical) but cannot follow it (*ruwe ne kusu ne is ungrammatical) Nakagawa (2024: 393). The community discussion at aynu-corpora Discord aynu-corpora Discord (2023–2026) (nukopoli, 2024-11-09) restates this as a layer-hierarchy principle — modal heads sit above the evidential, taking it as complement rather than the reverse — a formulation corroborated by Nakagawa's paradigm data ‹corpus-confirmed›. The evidential therefore occupies a position above clause-internal aspect but below the clause-linking and modal projections that take a finite clause as their argument.

Person and number agreement inside the evidential clause is independent of anything outside it. Suppletive verbal number alternations are resolved inside the evidential: the contrast íne nispa ek ruwe ne (singular ek 'come') vs. íne nispa arki ruwe ne (plural arki) turns on agreement within the embedded clause. The formal noun and copula carry no agreement of their own Nakagawa (2024: 393).

The formal status of the nominalized clause embedded under the evidential is not settled in the sources consulted directly for this grammar. Its finiteness is evident from its TAM and person marking. Dal Corso (2018: 314–315) observes that the copula ne of the direct forms expresses an internal, speaker-anchored viewpoint while the existential and posture verbs of the indirect forms express an external perspective, but stops short of a formal constituency analysis. Whether the embedded constituent constitutes a full finite nominalization, a predicate-phrase nominalization, or a construction-specific projection remains ‹speculative›. The general nominalization strategy underlying the construction 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)); the perfective auxiliary that may follow the evidential is treated in Chapter 111 (The Perfective/Anterior Particle a (a, a…a, aan)).

125.4 The grammaticalization cline

The four formal-noun evidentials derive from the possessed forms of ordinary sensory nouns: the path-noun ru ‘track, path’ yields the possessed form ru-we 'the trace of —', the appearance-noun sir ‘appearance, land’ yields sir-i, the voice-noun haw ‘voice’ yields haw-e, and the sound/feeling-noun hum ‘sound, feeling’ yields hum-i. The shift from sensory noun to clause-final marker was recognized in the earliest grammatical descriptions Kindaichi & Chiri (1936: 155–165) and is confirmed morphologically by Tamura's dictionary entries and Chiri's etymology: ruwe derives from 'there is a trace of X', shifted to 'there is the fact of X' Tamura (1996: s.v. ru, ruwe); Hirosawa (2026: 65). On the typological framework of (Hopper & Traugott 2003), the Ainu evidentials represent an intermediate point on the layered cline from lexical noun to grammatical marker.

Dal Corso (2018: 24–25) stresses that the result remains only partially grammaticalized: source-noun semantics are still active in the synchronic distribution (each form retaining a distinct sensory channel), the post-nominal predicate choice still matters for the direct/indirect distinction, and the formal nouns have lost full noun-phrase syntax — they admit no demonstrative or attributive modifier, block the extra possessive ending (*ruwehe, *humihi as evidentials), and are fixed in linear position Tamura (1996: s.v. ruwe). The Ainu system as a whole is a graded category; no Hokkaido form approaches the full grammaticalization of the Sakhalin particle manu ‹consensus›.

Within Hokkaido, the grammaticalization has advanced differently across dialects. Three stages are directly attested in the sources, yielding an intra-areal cline ‹speculative›:

In Stage I dialects the bare concept form closes the clause where Saru uses the possessed form: Tokachi iruska haw ne 'she was angry', tuk ru ne 'it emerged', alongside Saru iruska hawe ne and tuk ruwe ne Takahashi (2013: 130–131); Nakagawa (2024: 110). The allomorphs sin (for sir) and aw (for haw) are Tokachi-specific Takahashi (2013: 130–131). Stage I and Stage II represent a morphological difference — concept form vs. possessed form — rather than a semantic one; the four-way source distinction is present at both stages.

Stage III, the Horobetsu development, takes the possessed-form nominalizer a step further. Takahashi (2018: 108, 111) documents uses in which the nominalizers function not as complement-clause heads but as adverbial-clause markers — subordinating-conjunction-like elements — without a following copula. The adversative use is the clearest instance:

Takahashi (2018: 111–112) notes that this versatile adverbial use is anticipated in Kindaichi & Chiri's 1936 description of nominalizers as conjunctive particles Kindaichi & Chiri (1936: 155–160), suggesting the Stage III path is of some age within Horobetsu rather than a recent innovation. Takahashi connects this to what Noonan termed 'versatile nominalization' — a nominalization strategy that extends across both complement-clause and adverbial-clause functions.

One dialect fact complicates a simple Stage I → II reading of the cline. In Asahikawa and Shiranuka the trace-form ru enters the existential frame as an inferential (ru an), apparently in complementary distribution with the siri an of the southern dialects Dal Corso (2018: 326). This gives ru an indirect-evidential function that the southern varieties lack for this stem; it may reflect convergence toward the Sakhalin existential pattern (ruwehe ʔan), but the sources are thin and Dal Corso flags the comparison as requiring a wider-dialect study ‹speculative›.

The grammaticalization trajectory also bears on the frequency asymmetry within the direct paradigm. Yoshikawa (2020) characterizes ruwe ne as semantically unmarked among the four direct forms, capable of summarizing and marking scene transitions much as Japanese のだ does — a discourse function consistent with its being the most grammaticalized direct member Hirosawa (2026: 73). The community discussion at aynu-corpora Discord aynu-corpora Discord (2023–2026) (nukopoli, 2024-03-17) summarizes the same Yoshikawa 2020 characterization and notes the high frequency of ruwe ne in narrative and its summative/transitional function, a point corroborated by Dal Corso's corpus counts Dal Corso (2018: 72) ‹corpus-confirmed›. Genre-specific evidential patterning is treated in Chapter 155 (Narrative TAM and Evidential Patterning by Genre).

125.5 Double-evidential stacking

Double evidentials — two formal-noun evidential constructions stacked within a single clause — are more frequent in Hokkaido than in Sakhalin Dal Corso (2018: 331). Dal Corso (2018: 331–333) documents three functional types from the Kayano, Tamura, Nakagawa, and Bugaeva corpora:

In Type A the outer ruwe ne functions as a ratifier. An indirect form — typically hawe as (reportative indirect) or siri an (visual indirect) — presents the content from a character's perspective, and ruwe ne closes the utterance from the narrator's perspective as traditional, shared knowledge Dal Corso (2018: 331). Attested Type A combinations include hawe as ruwe ne, siri an ruwe ne, and siri ki siri ne Dal Corso (2018: 331–333).

The outer form in Types A and B is systematically ruwe ne and not siri ne, hawe ne, or humi ne. This asymmetry supports the analysis developed in Chapter 121 (ruwe ne — the Inferential / Visual-Trace Evidential): ruwe ne functions as an epistemic incorporation operator — a form that can ratify any already-evidentially-marked proposition as established knowledge — rather than as a fourth parallel sensory channel. Among the direct members it is the one whose source-noun semantics ('the trace of —') has bleached sufficiently to generalize across all sources of evidence once they have been assimilated by the speaker.

Type B uses ruwe ne as an epistemic upgrade. A direct evidential of relatively lower reliability is reinforced by a final ruwe ne asserting the speaker's overall confidence in the proposition. Dal Corso (2018: 332) (ex. 276) cites a narrator's utterance of the form hawe ne nankor ruwe ne — the modal nankor 'probably' attached to hawe ne is then closed by ruwe ne, which does not eliminate the uncertainty introduced by nankor but frames the compound as an incorporated fact from the speaker's perspective. This use is consistent with ruwe ne's position at the top of the source-reliability hierarchy in Dal Corso's analysis.

Type C, same-stem iteration, occurs in narrative contexts where the speaker presents both a subjective sensation and its arising as separate evidential moments:

Dal Corso (2018: 326–327) adds a distributional argument bearing on the siri forms. When the indirect visual siri an appears in a double-evidential construction, the ratifying outer form is consistently ruwe ne and never siri ne. Were the two simply co-equal visual channels, this selection would be unmotivated; it follows naturally if ruwe ne occupies a higher position in the evidential hierarchy, one at which its source-noun meaning has bleached enough to allow ratification of any other evidential type. The distributional gap that Dal Corso's sense-hierarchy account left open Dal Corso (2018: 311–312) is therefore partially resolved by the double-evidential distribution itself ‹corpus-confirmed›.

The stacking found in mirative contexts (§2 above) should be distinguished from double-evidential stacking. In siri an ruwe ne anan, the final anan is a mirative auxiliary, not a third formal-noun evidential; the stacking is between two evidential predicates and a mirative auxiliary in sequence, not among three co-equal evidential forms. Conflating the two construction types would obscure the structural difference between the evidential system (formal noun + copula or existential verb) and the mirative system (a perfective auxiliary above the evidential layer) Hirosawa (2026: 90); Dal Corso (2018: 331, 336).