Chapter 16Evidentiality

16.1 The reportative manu

manu signals that the speaker holds the information through someone else's words. Its core uses are hearsay, where the verbal source is left unspecified, and quotative, where the original speaker is named or the report accompanies a quotation (Dal Corso 2018: 295–298). Both are at home in conversation:

In direct discourse the bare reportative is rare; speakers prefer to reinforce it lexically with nu ‘hear’ ‘so I heard’ or nah ye ‘so they said’, which spell out the hearsay or quotative reading that manu alone leaves implicit (Dal Corso 2018: 297–298):

The letters of Sentoku Tarōji show the reportative at work outside folklore, in the east-coast variant manuy. We count seventeen tokens of manuy (beside one manu) across the three letters: writing real news, Sentoku systematically separates what he vouches for from what he passes on.

The third use is the one that floods the narrative corpus: manu as a marker of traditional knowledge. In tales the original purveyor of the information is either irretrievable or irrelevant; the teller presents the content as community knowledge transmitted by word of mouth, and manu in this function never takes a lexical reinforcement (Dal Corso 2018: 296–299). The formula opens nearly every one of Asai Take's tales:

16.2 The grammar of manu

Murasaki listed manu among her eighteen ‘auxiliary constructions’ (Murasaki 1976: 53–64), but its syntax sets it apart from the aspectual periphrases and auxiliaries of Chapter 15. Dal Corso ran the whole class through five tests — relativization of the construction as one predicate, independent use of the second element, obligatoriness of a syntactic linker, nominal particles on the first element, and nominalization with -hV — and manu fails them all: a clause with manu cannot be relativized or nominalized, manu never predicates on its own, nothing links it to its host, and it carries no morphology whatsoever (Dal Corso 2018: 176–187). He concludes that it is a final particle, like the polite kanne or emphatic naa: an invariable word that closes the sentence and contributes only procedural meaning. It accordingly takes no person marking and co-occurs freely with every person, including the indefinite of narration, and it follows the entire aspectual complex — kusu an manu, teh an manu, wa isam manu are the normal narrative cadences (Chapter 15). Its origin is opaque; the only published speculation derives it from hum an ‘there is the sound’, which Dal Corso reports as unverifiable (Dal Corso 2018: 77). The east-coast variant manuy and lengthened manuu are phonological, not morphological.

One extension is morphological in appearance only: manuyke, the reportative fused with the converb ike ‘and (then)’ (Chapter 23). It keeps the narrative under report while linking the clause to what follows; with 260 tokens in Asai Take's tales against 2,357 of plain manu(u) (our count), it is the regular way to continue a reported chain without closing the sentence.

16.3 The inferential ruwehe an

The inferential is built on the possessed form of ruu ‘trace’: a clause followed by ruwehe (reduced ruhe, ruuhe) plus an ‘exist’. The construction is an erstwhile possessive construction: the nominalized clause is the possessor and the sensory noun the possessee, so that re monimahpo an ruwehe an below is literally ‘the trace of three girls being there existed’. Synchronically the possessed noun has been pseudo-incorporated into an — the two form one phonological word, no modifier may intervene except the possessor clause, and neither piece can be displaced — while still bearing nominal morphology (Dal Corso 2018: 147–156). In the typology of Eurasian noun-modifying clauses the pattern is a gapless one: the scope clause relativizes no argument of its own but simply fills the possessor slot of a perception noun, on the template [possessor clause]–[possessed head]–predicate (Bugaeva 2017: 220–221), in contrast to the gapped relative clauses of Chapter 21. The development presupposes the Sakhalin extension of possessive marking to clauses described in Chapter 20 (Bugaeva 2016: 104–113). The speaker asserts an event he did not witness, on the strength of its perceptible results:

Beside ruwehe an the east coast has ruwehe ne(e) with the copula, unattested in the west (Dal Corso 2018: 72). The copular variant presents the inference as processed and certain — the content has been thought through and brought into the speaker's own dimension — and it is the only inferential that combines with the personal-knowledge forms (§16.6) (Dal Corso 2018: 300–301). What conditions the choice between the two anchors — viewpoint, person and dialect — is analysed from the copula's side in §18.6:

Person interacts with the construction in a way that confirms the clausal-possessor analysis: a first- or second-person subject of the scope verb is never cross-referenced on the evidential noun (Piłsudski's otanuye e-kii ruwehe an ‘you must have drawn on the sand’ shows bare ruwehe, not *e-ruwehe), because the possessor is the whole nominalized clause and hence third person (Dal Corso 2018: 154–155). The one piece of agreement the noun does accept is the nominal plural -hcin, optionally echoing a plural subject of the scope verb (§16.4).

16.4 The sensory evidentials

Three further nouns enter the same construction, each naming the channel of perception: siri ‘appearance’ for what is seen, hum ‘sound’ for what is felt or heard as noise, haw ‘voice’ for what is heard as speech or cry. The set matches the Hokkaidō indirect evidentials siran, humas, hawas, with which it shares its origin in sensory nouns — Ainu being one of the typologically rare languages whose evidentials are noun-based (Dal Corso 2018: 73, 156), (Bugaeva 2016). The choice among the forms follows a sense hierarchy: vision outranks the other channels, so ruwehe an (inference from visible traces) and sirihi an (direct visual impression) encode more reliable access than humihi an and hawehe an (Dal Corso 2018: 272–295).

The corpus distribution is sharply skewed. In Asai Take's 54 tales we count 53 tokens of hawehe an, 15 of ruwehe an, 10 of humihi an and none of sirihi an; the visual form is a Rayciska, west-coast resource (17 tokens in the Fujiyama Haru texts, none in Piłsudski's east-coast corpus by Dal Corso's count, where instead ruwehe ne appears) (Dal Corso 2018: 72). The auditory form earns its frequency from narrative technique: scenes are routinely perceived through walls and doors, as in the last three examples. Beside its sensory use, hawehe an shades into hearsay — ‘there is the voice (saying)’ — bridging towards the reportative; the corpus formula nah yee hawehe an manu ‘a voice was heard saying so’ (text 19) stacks voice-perception, quotation and report in one sentence.

Two evidentials may co-occur on one predicate. Personal knowledge -hV combines only with ruwehe ne/an and sirihi an, the high-reliability inferentials, and there it retains only its epistemic force; and manu, in its traditional-knowledge function, closes constructions of every type, always last (Dal Corso 2018: 300–302):

The reportative supplies a possible third layer, chaining onto the outside of whatever stack the sentence has built. In the following Piłsudski sentence a personal-knowledge nominalization sits under the copula and the whole assertion is then relayed by manu — three evidential values in strict continuum order, the most reliable innermost:

16.5 Evidentials as relative tense

Sakhalin Ainu has no tense morphology (Chapter 15), and in narrative much of the temporal load falls on the evidentials. The mechanism is relative tense of the Reichenbachian kind. A sensorial construction asserts that the stimulus — trace, sight, sound, voice — is there for the speaker now, so it pins the reference time to speech time; where the event itself sits is then read off the aspectual class of the predicate in its scope (Dal Corso 2018: §7.3.9). Because a trace can outlast the event that left it, ruwehe an is the one form that genuinely alternates: a telic scope predicate places the event before its still-perceptible result, yielding a past inference, while an atelic predicate lets event and trace run concurrently, yielding a present one.

The rest of the set does not alternate. With ruwehe ne(e) the copula draws the inferred content into the speaker's here and now, collapsing event, reference and speech time and fixing a present reading whatever the telicity of the scope predicate; and since the sight, sound or voice of sirihi an, humihi an and hawehe an exists only while the event produces it, those three default to the present as well (Dal Corso 2018: §7.3.9) (on the two anchors see §18.6).

Dal Corso left one cell open: motion and possession verbs resist a lexical telicity value, so the past/present prediction for ruwehe an appears undetermined over exactly the verbs it most often scopes. For motion the west-coast record decides. A ruu is the paradigm of a resultant-state object — what a completed displacement leaves behind — and when a tracker reads a blood-trail in Fujiyama Haru's telling, oman ‘go’ under ruuhe an can only mean that the going is over:

We do not take this as evidence that oman is lexically telic. Rather the construction coerces its scope: only a bounded, completed motion can deposit a trace, so the trace-based inferential imposes a telic, result-leaving construal on whatever predicate it takes, and the past reading follows from the construction, not from the verb. Possession resists the same diagnostic for a principled reason — a having that has lapsed leaves nothing behind to read — and we find no token of ruwehe an over koro ‘have’; that sub-case remains genuinely open.

16.6 Personal knowledge, person, questions and negation

The personal-knowledge evidential asserts that the information sits in someone's own knowledge, whatever the original channel of acquisition. Its exponent is the nominalizing -hV of Chapter 20 used sentence-finally, beside a zero alternant; Murasaki described the overt form as an assertive or ‘compassionate question’ ending (Murasaki 1979: 72), and Bugaeva showed that all its uses share presupposed, given content (Bugaeva 2016: 110–113).

Person constrains this evidential less than one might expect. Dal Corso's survey finds -hV and -Ø distributed almost evenly across first, second and third persons and across affirmative and negative sentences; what varies with person is the pragmatics (Dal Corso 2018: 238–241). In declaratives the source of the knowledge is the speaker. In questions it is the addressee: the speaker uses the personal-knowledge form to signal a presupposition that the hearer knows the answer at first hand (Dal Corso 2018: 237, 252–254). Hence the interrogative is this evidential's natural habitat — Chiri already noted that questions prefer the -hi form (Chiri 1942: §40):

Where no answerer is present at all, the same form builds the rhetorical question with the final hetaneya: hemata kusu enan si hura an-i-hi hetaneya? ‘why on earth is there such a smell of dung?’ (Piłsudski 1912: 160), (Dal Corso 2018: 252–253). The other evidentials behave asymmetrically here. The inferential and sensory constructions occur freely in content questions inside their scope clause but are not themselves questioned; and the reportative is barred from questions altogether — in Asai Take's 3,852 corpus sentences manu never occurs in a genuine interrogative (our count), as expected of a final particle that marks the assertion as relayed.

Under negation the ordering is rigid: negation is built inside the clause and the evidential closes over it, so the report or inference is about a negative state of affairs, never the reverse. The negators of Chapter 17 precede manu (five corpus tokens of the type V ka hankii manu, our count), and we find no token in which an evidential noun is negated by isam or by a preverbal negator — ‘there is no trace that…’ is simply not how the system works. We analyse this as a scope fact: in the attested material evidentiality in Sakhalin Ainu consistently scopes outside polarity, attaching at the illocutionary level, and a counter-evidential reading would require the lexical strategy (e.g. nukara ka hankii ‘did not see it’).

16.7 Genre, narration and the reportative

Evidentiality is the grammatical backbone of Sakhalin narrative genre. In the tuytah (folktales) of Asai Take, 1,848 of 3,852 corpus sentences — 48% — carry manu(u), 2,357 tokens in all, plus 260 of manuyke (our counts, (Murasaki 2001)). Whole pages run with the particle closing every sentence: it is no longer reporting an utterance but framing the tale as traditional knowledge. Dal Corso tracks this shift from reportative proper to direct evidential of shared knowledge across dialects and finds it most advanced exactly in Rayciska and Ussoro, where manu is hardly ever a plain hearsay marker any more, while east-coast dialects keep the reportative use alive and disambiguate it with lexical reinforcements; the geography of the shift suggests an areal diffusion through adjacent west-coast villages (Dal Corso 2018: 298–299). Throughout the shift the marker itself never changes: one invariable manu covers both functions, and only distribution and reinforcement tell them apart. Hokkaidō Ainu makes the same move overtly, exchanging the hearsay predicate hawas for ruwe ne when narration presents settled knowledge, which throws the Sakhalin invariance into relief (Dal Corso 2018: 298–299). Whether the west-coast state continues an older east-coast one cannot at present be tested: Piłsudski's turn-of-the-century materials and Murasaki's of the 1970s were gathered from different speakers, genres and elicitation settings, and the apparent diachrony may be an artefact of the record.

The reportative frame interlocks with a person convention. In folktales the narrator-protagonist is cross-referenced by the indefinite person an-, -an, i- (Chapter 13): the ‘I’ of the tale is not the teller, and the use of indefinite and first-person-plural marking for the speaker is a fixture of folklore (Dal Corso 2024: 68). Piłsudski's incipit above (§16.1) shows the pattern: an-kor henki-hi i-reske manu, indefinite person under the traditional-knowledge reportative. The letters invert both settings at once: Sentoku writes of himself with an= and ci=, vouching for his news in unmarked clauses, and reserves manuy for the items he has at second hand (§16.1). Genre, person marking and evidential choice are thus a single package: indefinite narration under manu for inherited tales, first-person assertion with occasional reportatives for lived report. The same division explains why the inferential and sensory forms cluster inside tales at moments of restricted perception — the protagonist hears crying through a wall, finds a cold hearth — while the personal-knowledge forms cluster in dialogue, where speakers negotiate who knows what.