Chapter 119ruwe ne — the Inferential / Visual-Trace Evidential

119.1 Place in the evidential paradigm

Hokkaido Ainu marks the speaker's source of information periphrastically: a finite clause is nominalized by a bound formal noun and the result is predicated with the copula ne. The language has no dedicated evidential verbal affix. Four formal nouns build the core paradigm, each the possessed form of a sensory noun: ruwe (< ru ‘track, path’), siri (< sir ‘visible appearance, land, weather’), hawe (< haw ‘voice’), and humi (< hum ‘sound, feeling’) Nakagawa (2024: 258, 392); Hirosawa (2026: 64). Mapped onto the typology of (Aikhenvald 2004: 63–65), the system is of the C1 type — visual, non-visual sensory, inferred, and reported — realized analytically, through a native nominalization-plus-copula construction Hirosawa (2026: 64); Takahashi (2013: 129). The architecture common to all four terms is set out in Chapter 118 (The Nominalization-plus-Copula Evidential Schema); the present chapter treats the central member ruwe ne, and its siblings in the situational evidential, the reportative, and the sensory evidential. The closed class of formal nouns that makes the construction possible is defined in Chapter 34 (The Formal / Defective Noun Set (ruwe, hawe, siri, humi, hi, pe, kur)).

Dal Corso resolves these four direct terms into an eight-form system whose indirect counterparts replace the copula with the existential an, the verb ki ‘do’, or as ‘stand’, organized by two parameters, source reliability and event accessibility; within it ruwe ne is the highest-reliability direct form, and the trace-noun ru forms direct evidentiality alone — the southern dialects have no indirect *ru-counterpart to siri an Dal Corso (2018: 304, 309, 337).

ruwe ne also dominates the paradigm quantitatively. In Dal Corso's Hokkaido reference texts it supplies 1508 of 2232 evidential tokens — roughly two thirds of the eight-form system, and far ahead of hawe ne, siri ne, and humi ne Dal Corso (2018: 72). The present Hokkaido corpus of 167,890 sentences makes the skew sharper. A string search returns 16,350 tokens of ruwe ne beside 796 hawe ne, 607 siri ne, and 200 humi ne among the direct forms, and 304 siri an, 395 hawe as, 210 humi as, and 218 siri ki among the indirect: some 86% of the eight forms together, and about 91% of the four direct ones (the two corpora are compared below). The string count folds in the homophonous lexical noun ruwe and the reduced final particle, so 86% is an upper bound on the evidential proper. The asymmetry is far too large to follow from that conflation, and it confirms the descriptive consensus that ruwe ne is the unmarked, most grammaticalized term of the four ‹corpus-confirmed›.

119.2 Morphology: ruwe is ru-we, ‘the trace of —’

The head ruwe is the possessed form of the noun ru ‘track, path’. In Tamura's Saru dictionary the ordinary possessed form is ruwe(he) ‘the track that — passed along, the path made by —’s passing’, and the evidential formal noun is analysed as ru-e, the noun plus a possessive ending Tamura (1996: s.v. ru, ruwe). For glossing this is decisive: the morpheme is the possessed ru-we, ‘the trace of —’, whose implicit possessor is the event described by the nominalized clause. The same possessed-noun pattern underlies sir-i, haw-e, and hum-i Nakagawa (2024: 258, 390). A subtle point distinguishes the evidential from the full noun: the lexical possessed forms may carry an extra ending (ruwehe, humihi), whereas the evidential formal nouns block it (*ruwehe is ungrammatical as the evidential) Tamura (1996: s.v. ruwe).

The construction is therefore a copular predication over a nominalized clause — [clause] ruwe ‘the trace/fact that [clause]’, closed by the copula ne treated in Chapter 94 (The Copula ne and Predicate-Nominal Clauses). It is not, however, an ordinary relative-clause head plus a copula. The head has lost the syntax of a free noun: it admits no demonstrative or attributive modifier of its own (contrast a true lexical head such as pirka menoko ‘a beautiful woman’), it cannot head a common noun (*pirka ruwe ne kewtum), and it is fixed in linear position — kusu ne ruwe ne is well formed but *ruwe ne kusu ne is not Nakagawa (2024: 393); Dal Corso (2018: 24–25). The accurate synchronic label is thus a partially grammaticalized formal-noun evidential, intermediate between a fused verbal suffix and a fully lexical noun, and it is best analysed alongside the formal-noun nominalization treated in Chapter 100 (Lexical-Head and Formal-Noun Nominalization (kur, uske, ruwe/hawe/siri/humi)) and the general clausal nominalizers of Chapter 99 (The Nominalizers -p/-pe and -i/-hi (Participant, Event, Place, Fact)) ‹consensus›.

119.3 Meaning: fact from knowledge, not raw perception

Four framings of ruwe ne compete, and they are best read as describing one form from different angles. The oldest, the mood tradition, classed ruwe as the 確説法/事実法 ‘assertive / fact mood’, with Chiri deriving it diachronically from ‘there is a trace of X’ to ‘there is the fact of X’ Kindaichi & Chiri (1936); Hirosawa (2026: 65–68). The evidential framing, now standard, retains a source semantics: Nakagawa defines ruwe ne as marking that the speaker recognizes the clause content as fact Nakagawa (2024: 392), and Dal Corso treats it as a direct evidential of reasoning or knowledge (glossed dir.rsn / dir.knw), the highest-reliability direct source. This reliability is a matter of source, not of epistemic modality: it names the psychophysical or cognitive channel of acquisition and says nothing about the speaker's degree of certainty, and the original stimulus is typically backgrounded once the information has been assimilated into personal knowledge Dal Corso (2018: 305, 327, 335, 339).

Satō's question data refine what that channel is. Among the four nominalizers, ruwe applies where the relevant content is an abstract state of affairs, recognized only indirectly through the exercise of understanding — the domain left once hawe takes an utterance, siri a visible ongoing scene, and humi a sensation Satō (2008: 134–135). He further notes that the very fact that these nominalizers stand in the possessed form presupposes an antecedent — a possessed noun needs something to belong to — so the form grammatically encodes that the speaker already has the relevant situation in hand Satō (2008: 135). The diagnostic case is a predicate whose truth no sense can deliver:

The fourth framing is Hirosawa's, who relocates ruwe ne into information structure. He argues it marks discourse-new information much as Japanese のだ does, and more generally declares the content to lie inside the speaker's territory of information (情報のなわ張り); since ruwe references processed knowledge (知識), this reconciles its two apparently opposite uses — introducing brand-new information and ‘summing up’ (前方統括) information already developed in the discourse Hirosawa (2026: 75–81, 89). The self-identification below involves no inference and no perceptual source at all; the sea-god simply supplies his identity to the protagonists as new information:

The same mechanism lets a narrator assert events of the story world that she could not have witnessed, presenting them under ruwe ne as traditional, shared knowledge — the use that makes it the default predicate of the narrative frame-text Dal Corso (2018: 331); Hirosawa (2026: 75):

These accounts are not interchangeable, and at one point they genuinely conflict: Hirosawa builds his analysis on territory-of-information, whereas Dal Corso explicitly argues that the territory-of-information theory is inapplicable to Hokkaido direct evidentiality Dal Corso (2018: 335–337); Hirosawa (2026: 89) ‹contested›. The synthesis adopted here is that ruwe ne grammatically encodes epistemic incorporation: the clause is predicated as a possessed trace (ru-we, ‘the trace of —’) whose possessor is the described event, so that the proposition is offered as content already taken into the speaker's knowledge and thereby made assertable. This single device derives both Dal Corso's observation that the stimulus is backgrounded once information enters personal knowledge and Hirosawa's that the form marks new information and claims it for the speaker's territory: incorporation is exactly what lets a freshly perceived, inferred, remembered, or just-narrated state of affairs be presented as a fact the speaker now owns. The possessed morphology — which, as Satō notes, presupposes an antecedent — is the formal anchor of this account, and it explains why the older ‘fact mood’ label, the modern source label, and the discourse-new label each capture a true facet of the same form ‹original-needs-review›.

119.4 The ruwe ne / siri ne puzzle

The hardest boundary in the system is between ruwe ne and siri ne, because both can rest on what the speaker sees. Dal Corso states plainly that a sense-channel hierarchy does not predict the split: if both are ‘visual’, it fails to say why Ainu so often chooses the trace-form where sight is available, and he leaves the question open, suspecting a driver beyond elicitation Dal Corso (2018: 311–312, 335–336). Hirosawa's response is to separate raw sight from processed knowledge — siri profiles the visible situation, ruwe the knowledge the situation yields — but he offers this as an explicit hypothesis (仮説), not a settled result Hirosawa (2026: 81, 89) ‹contested›.

The double-evidential distribution gives a positive argument that resolves part of the puzzle. When two evidentials stack, the indirect visual siri an is reinforced specifically by ruwe ne, and not by siri ne Dal Corso (2018: 326–327, 331–333). Were ruwe and siri two co-equal visual channels this would be unmotivated; it follows naturally if ruwe is not a fourth sense parallel to sight, sound, and speech, but a higher knowledge-from-trace operator that ratifies the output of another evidential as established fact. On this reading Dal Corso's unresolved ‘why ruwe and not siri when sight is available’ Dal Corso (2018: 311–312) and Takahashi's Tokachi finding that individual-level — cognitively stable — properties take ru while transient scenes take sir Takahashi (2013: 131–133) become two faces of one contrast: siri profiles the accessible appearance, and ruwe the cognized fact extracted from it ‹original-needs-review›. The boundary with hawe is similarly non-rigid — Hirosawa cites near-identical sentences alternating hawe ne and ruwe ne, which fits a form that can absorb information of any original channel once it has become knowledge Hirosawa (2026: 81).

119.5 Clause-final frames and illocution

ruwe heads a small family of sentence-final frames that vary the following predicate and particle. Nakagawa arranges them into a paradigm shared by all four formal nouns: NMLZ + ne declarative, NMLZ + an content question (with a question word), bare NMLZ polar question and exclamation, and NMLZ + un affirmative answer Nakagawa (2024: 392–396). The bare polar reduction is an instance of insubordination — a once-dependent nominalized clause used as a main clause — analysed for the Tokachi cognate by Takahashi (2013: 132). These interrogative uses belong with the wider interrogative system in Chapter 142 (Interrogative Strategies, Question Particles, and Evidential Questions), and the existential frame with the existential an.

The content question replaces the copula with the existential and is opened by a question word, the formal noun staying in place:

Negation belongs inside the evidential: the proposition is denied by negating the verb before the formal noun (the pattern … somo ki ruwe ne ‘it is a fact that [one does not …]’), which leaves ruwe ne itself outside the scope of negation. The rhetorical ruwe somo ne is a separate outer frame, a high-scope convictional construction distinct from this inner propositional negation Nakagawa (2024: 393). The interaction of the evidential with the sentence-final negation and TAM apparatus on a copular predicate is treated in Chapter 98 (TAM, Evidentiality, and Negation on Nonverbal Predicates).

119.6 Scope over TAM and the finiteness of the embedded clause

The constituent before ruwe ne is a finite, clause-sized predication. It carries its own person marking, number-sensitive verb choice (singular ek vs. suppletive plural arki ‘come’, so íne nispa ek ruwe ne ‘four men came’), aspect and perfective auxiliaries, and negation; the evidential then scopes over that whole TAM-marked clause Nakagawa (2024: 393). This is also why ruwe ne can close a non-verbal predicate, as in the copular kamuy a=ne … ruwe ne example above. Ordering confirms the layering: the evidential may be followed by aspectual and clausal material (a, nankor, na, wa, hine, kusu), and it sits above clause-internal aspect but below the clause-linkers, so kusu ne ruwe ne is grammatical and *ruwe ne kusu ne is not Nakagawa (2024: 393).

Dal Corso ties the choice of post-nominal predicate to perspective: the copula ne expresses an internal (speaker-anchored) viewpoint and so hosts the direct forms, while the existential and posture/action verbs an, as, and ki express an external viewpoint and host the indirect forms Dal Corso (2018: 313–315). For ruwe the paradigm is asymmetrical — in the southern dialects there is no indirect *ru form — so ruwe ne functions as the direct, personal-knowledge anchor of the whole system Dal Corso (2018: 309, 326). What the accessible sources do not settle is the formal size of the nominalized clause: that it is finite is clear from its TAM and person marking, but no read source provides a constituency account that would decide between a finite nominalization, a nominalized predicate phrase, and a construction-specific evidential projection ‹speculative›. The complementation side of the same formal-noun strategy — ruwe and hi clauses under perception and cognition verbs — is taken up in Chapter 104 (Nominalized Complements and Control/Raising).

119.7 Predication type: the Tokachi constraint

Takahashi's Tokachi study adds a factor outside the reach of both the mood tradition and a pure source typology. In Tokachi the visual nominalizer sir is restricted to imperfective, stage-level (transient) predication: it presents a passing scene. For an individual-level (permanent) property — even one known by sight — Tokachi instead uses ru, which is unmarked for information type. Of the four, only sir carries this restriction; hum, haw, and ru work at both levels Takahashi (2013: 131–133). The constraint surfaces on the nominalizer itself, which Takahashi parallels with the ban on individual-level complements of English perception verbs.

For the present form this matters in two ways. First, the ru domain is broader than ‘inference’ narrowly understood: it is the unmarked choice where sir is blocked by predication type, which is consistent with the knowledge-from-trace reading developed above. Second, it shows the Hokkaido system to be organized partly by the ontology of the predicate — eventive, stage-level, individual-level — alongside source of information; this does not overturn Dal Corso's reliability/accessibility account but bounds its domain, marking where the grammar of predication independently constrains the available evidential ‹consensus›.

119.8 Mirativity and double evidentiality

ruwe ne is often rendered ‘it turned out that …’ in surprising contexts, but the surprise is not contributed by ruwe. Mirativity in Hokkaido Ainu is carried mainly by the perfective auxiliary aan and by anan; ruwe ne then packages the resulting proposition as fact Hirosawa (2026: 90). Dal Corso is explicit that Hokkaido shows no evidence of the evidential paradigm itself systematically generating mirative meanings — so mirativity may co-occur with ruwe ne but is not its value Dal Corso (2018: 336) ‹contested›. The system-level treatment of the evidential-to-mirative interface, evidential stacking, and the cline to sentence-final particles is given in Chapter 123 (Mirativity, Evidential Scope, and Grammaticalization to Sentence-Final Particles).

Double evidentiality, more frequent in Hokkaido than Sakhalin, confirms the special standing of ruwe ne argued for above. Two patterns recur: a character's indirect source layered under the narrator's ruwe ne as shared knowledge (… hawe as ruwe ne), and ruwe ne reinforcing a low-reliability direct form (… hawe ne nankor ruwe ne) Dal Corso (2018: 331–333). In both, ruwe ne is the member that can ratify an already evidentially marked proposition as established fact, behaving as a knowledge operator that scopes over the other evidentials.

119.9 Diachrony, dialect microvariation, and a Sakhalin contrast

The diachronic path is transparent and is stated already in the earliest grammars: the noun ru ‘track, path’ > the possessed formal noun ru-we ‘the trace of —’ > the clause-final evidential [clause] ruwe ne, part of the larger cline by which sir, hum, haw, and ru all move from sensory nouns to nominalizers to evidentials Tamura (1996: s.v. ruwe); Hirosawa (2026: 65). Dal Corso stresses that the result is only partially grammaticalized — the lexical source stays visible and the choice of post-nominal predicate still matters — so grammaticalization here is a matter of degree Dal Corso (2018: 24–25); Takahashi (2018: 111).

The endpoint of that cline is reached unevenly across Hokkaido, and the differences are morphosyntactic: a single semantic field realized by varied forms, which cautions against writing one morphology for all of Hokkaido ‹contested›:

• Saru, Chitose, Biratori — the southern core of Dal Corso's data — keep the possessed forms ruwe / siri / hawe / humi and the eight-form direct/indirect system Dal Corso (2018: 304–337).
• Tokachi (and frequently Ishikari) close the clause with the bare concept forms ru / sir / haw / hum instead of the possessed forms Nakagawa (2024: 110, 390); Takahashi (2013: 129–131).
• Horobetsu grammaticalizes the possessed-form nominalizers a step further, into adverbial-clause and conjunction-like markers — a ‘versatile nominalization’ use not reported for well-described Saru Takahashi (2018: 107–112).
• Asahikawa and Shiranuka have an inferential ru an, apparently in complementary distribution with the siri an of the south — a genuine exception to the generalization that ru forms direct evidentiality alone Dal Corso (2018: 326).

The Sakhalin reflex. The Sakhalin varieties split the trace-form where the southern Hokkaido system has a single member. Beside the copular ruwehe ne(e) they carry an existential ruwehe ʔan, and the two divide one inferential field by reliability: the existential encodes inference from a less-processed visual stimulus, the copular form inference from a stimulus already worked through, the more reliable of the pair Dal Corso (2018: 281–284). Southern Hokkaido has only ruwe ne and admits no existential *ruwe an; the ru an of Asahikawa and Shiranuka is the one Hokkaido locus where the trace-noun enters the existential frame, and Dal Corso reads it as the Hokkaido counterpart of Sakhalin ruwehe ʔan Dal Corso (2018: 326, 336). The split sits within a wider reorganization of nominalization in the Sakhalin variety (Bugaeva 2016).