Evidentiality is a morphosyntactic system that obligatorily or frequently marks the speaker's source of information: direct perception, inference, reported speech, or other epistemic access. Languages with rich evidential systems (e.g., Quechua, Turkish, Korean) require speakers to grammatically encode how they know what they are asserting.
Study a language with obligatory evidential marking to observe morpheme placement and interaction with tense and aspect. Then examine languages where evidentiality is optional or expressed through lexical adverbials (reportedly, obviously) to understand the spectrum of grammaticalization.
You already know from pragmatics that language does more than convey propositional content — it encodes the speaker's relationship to that content. You also know from morphology that grammatical categories can be obligatorily marked through bound morphemes or free grammatical words. Evidentiality sits at the intersection of these two domains: it is a grammatical category that encodes not *what* is being claimed but *how the speaker knows it*.
In everyday English, you can optionally signal your epistemic source through adverbs and hedges: "reportedly," "obviously," "I heard that," "apparently," "I saw with my own eyes." But these are lexical choices — you can leave them out. In languages with obligatory evidential marking, leaving the source unspecified is not an option. Quechua, Turkish, and Korean, for instance, require speakers to morphologically commit to a source category on every assertion. The major distinctions typically encoded are: direct evidence (the speaker personally witnessed the event), inferential (the speaker is reasoning from visible results — you see smoke and conclude there was a fire), and reportative (the speaker heard about it from someone else). Some systems add finer distinctions: hearsay from a named source versus general rumor, inference from physical evidence versus inference from general knowledge.
Consider a Turkish example. If you say "geldi" (came-PAST), you're asserting direct knowledge — you saw the person arrive. If you say "gelmiş" (came-INDIRECT.PAST), you're signaling you weren't a direct witness: you inferred it or heard it from someone else. The morpheme *-miş* is not a past tense — Turkish has two separate past tenses, one for witnessed and one for unwitnessed events. This is not a subtle stylistic choice; it is grammatically obligatory. Misusing the evidential marker in Turkish carries social weight — asserting direct knowledge of something you only inferred is a form of dishonesty, not just an imprecision.
What makes evidentiality theoretically fascinating is its relationship to epistemic modality, which you'll encounter in the builds-toward topic. Modality encodes degrees of certainty (*might*, *must*, *probably*); evidentiality encodes source of evidence. The two often interact — inferential evidentials frequently carry an implication of lower certainty — but they are grammatically distinct in languages that encode both. English blurs this distinction by using modal verbs to do some evidential work ("She *must have* left already" = inference from evidence), but the grammaticalization is incomplete. Studying languages with richer evidential systems reveals that these categories are grammatically real, universally relevant to human communication, and simply expressed differently across languages — not absent in some and present in others, but variously grammaticalized along a continuum from obligatory morphological marking to optional lexical signaling.