Evidentiality

How evidentiality works

An evidential is a grammatical morpheme — a suffix, clitic, particle, or auxiliary — that obligatorily marks the source of the speaker's information. The key word is obligatorily. Languages without grammatical evidentiality, like English, can express the same notions optionally with adverbs and parentheticals; languages with grammatical evidentiality cannot leave the source unspecified in the relevant clause types. If you tell someone in Tariana that it rained last night, the morphology of the verb forces you to commit to whether you saw the rain falling, heard it on the roof, inferred it from a wet street in the morning, assumed it because clouds had been gathering, or were told by a neighbour.

The Quechua family illustrates a typical three-way system. The suffixes -mi, -si, and -cha attach to the focused element of the clause:

• Para-sha-n-mi — "It is raining" (direct evidence: I see the rain).
• Para-sha-n-si — "It is raining (they say)" (reportative: someone told me).
• Para-sha-n-cha — "It must be raining" (inferential: I see wet umbrellas).

Martina Faller's dissertation on Cuzco Quechua (2002) shows that these three suffixes are not interchangeable: a speaker who says -mi is committed to having direct evidence, and the hearer can challenge the assertion on those grounds. This is the diagnostic for genuine grammaticalisation — failing to use the right evidential is treated as a category error, not a stylistic lapse.

Tariana — the canonical five-way system

Alexandra Aikhenvald's fieldwork on Tariana, an Arawakan language with roughly 100 fluent speakers in the Vaupés region of Brazil, produced the most detailed grammar of evidentiality ever written (Aikhenvald 2003, A Grammar of Tariana from Northwest Amazonia). Tariana fuses tense and evidential into a single portmanteau suffix on the verb. For the simple past of na-keta "he found it":

Suffix	Type	What the speaker is committing to
-ka	Visual	I saw him find it
-mahka	Non-visual sensory	I heard / smelled / felt the finding
-nihka	Inferential	I see the result and infer the finding (he is holding the object)
-sika	Assumed	I have general reason to believe it (he was searching all day)
-pidaka	Reported	Someone told me he found it
(none)	Ungrammatical	You cannot omit the suffix in past-tense statements

Tariana children acquire the system early. By age three they distinguish at least visual versus reported. Aikhenvald notes that mismatches — using -ka for events one did not actually witness — are treated as a moral failure on a level with overt lying.

Grammatical evidentiality vs English hearsay adverbs

	Grammatical evidentiality (Tariana, Quechua)	English lexical hedges
Obligatory?	Yes, in the relevant tense/mood	No, always optional
Form	Bound morpheme on the verb or focus	Free adverbs, parentheticals, "I heard" complements
Number of distinctions	2 to 6 fixed grammatical categories	Open lexical class — "allegedly", "reportedly", "they say", "apparently"
Default if omitted	Sentence is ungrammatical	Sentence is grammatical, source unspecified
Mismatch (wrong source)	Treated as lying	Treated as imprecise wording
Acquisition order	By age 3-4 (Papafragou et al. 2007 on Korean)	Continues into adolescence
Interaction with question	Marks the source the questioner expects in the answer	Mostly absent; rises in indirect questions only

The functional load is different too. In English, "allegedly" cues distance and is mainly a journalistic register marker; in Tariana, the evidential is as routine as tense in English and carries no special stylistic colour.

Worked examples

Bulgarian (two-way: indicative vs renarrated). The renarrated mood (sometimes called the preizkazno naklonenie) is built from old perfect forms. Compare:

• Ivan dojde. "Ivan came." (Indicative; I witnessed it.)
• Ivan dosul. "Ivan came (they say)." (Renarrated; reported information.)

Newspapers and historical narrative use the renarrated form heavily; using the indicative for unwitnessed events sounds presumptuous (Friedman 1986).

Turkish (two-way: -DI direct vs -mIş indirect). The suffix -mIş covers reported, inferred, and surprise meanings:

• Ali geldi. "Ali came." (Witnessed.)
• Ali gelmiş. "Ali came, apparently / they say / it turns out." (Indirect — speaker did not directly observe arrival.)

Lars Johanson (2003) treats -mIş as an indirective rather than a pure evidential, because the cluster of meanings (reported, inferred, mirative surprise) tracks "the speaker is not the immediate source" rather than picking out one source type.

Eastern Pomo (Pacific Northwest, four-way). Sally McLendon (2003) describes four suffixes: -ink'e (non-visual sensory), -ine (inferential), -le (hearsay), and -ya (direct/visual). A statement about a person burning themselves picks a different suffix depending on whether the speaker saw it, smelled the burned skin, deduced it from a bandage, or was told by someone else.

Comparing evidential systems cross-linguistically

Language	Family	Distinctions	Marker type	Obligatory?
Tariana	Arawakan	5 (visual / non-visual / inferred / assumed / reported)	Verbal suffix (fused with tense)	Past + present
Tucano	Tucanoan	5	Verbal suffix	Past + present
Cuzco Quechua	Quechuan	3 (direct / reported / inferential)	Enclitic on focused constituent	Most assertions
Aymara	Aymaran	3	Suffix on verb	Most assertions
Turkish	Turkic	2 (direct -DI / indirect -mIş)	Verbal suffix	Past tense
Bulgarian	Slavic (Balkan)	2 (indicative / renarrated)	Periphrastic mood	Often, in narrative
Korean	Koreanic	4-5 (direct, reported, inferential, retrospective)	Sentence-final ending	Many speech levels
Eastern Pomo	Pomoan	4 (visual / non-visual / inferred / hearsay)	Verbal suffix	Past, present
English	Germanic	0 grammatical (lexical hedges only)	Adverbs, "I heard", "reportedly"	No

Variants and edge cases

• Mirativity. A related category marking surprise or new information; the Tibeto-Burman mirative (DeLancey 1997) often shares morphology with non-visual evidentials. Turkish -mIş doubles as a mirative.
• Egophoric systems. Tibetan and many other Sino-Tibetan languages mark whether the speaker is an "intentional source" of the action — a related but distinct category (Tournadre 2008).
• Evidential strategies. Languages without dedicated evidential morphology may co-opt other categories — perfect tenses, passives, complement-taking verbs — for evidential-like work. German sollen ("is said to") and wollen ("claims to") are evidential strategies, not grammatical evidentials.
• Reported speech as quasi-evidential. Direct quotation is universal; what matters for grammatical evidentiality is whether bare assertion can be reportative without a quotative frame.

Common pitfalls in describing evidentiality

• Confusing evidentiality with epistemic modality. Source is not the same as confidence. A speaker can be highly confident about a reported event (the news is reliable) and uncertain about a witnessed one (it was foggy). Aikhenvald (2004) is firm that the two categories are formally distinct in languages where they are independently expressed.
• Treating English "reportedly" as evidential. It is an evidential strategy, not a grammatical evidential — it is optional, free-standing, and lexical. Authors who call English an evidential language are using the term loosely.
• Assuming evidentials always mark uncertainty. The Tariana visual evidential signals high confidence; non-visual and inferred can be associated with lower confidence, but reported is sometimes used precisely because the source is authoritative.
• Glossing -mIş as "past tense". A common error in Turkish pedagogy. The -DI/-mIş contrast is not primarily temporal; both can be past, but they encode different sources.
• Ignoring the question paradigm. In Quechua, the evidential in a question marks the source the questioner expects in the answer (conjunct/disjunct alignment). This is a major Aikhenvald-Faller diagnostic.