Imagine you have to recount a small event — say, that John played soccer yesterday. In English you can simply say "John played soccer", and your listener has no idea whether you were on the touchline, heard cheering from the next street, deduced it from his muddy boots, or were told by a friend. Languages of Europe leave the source of information almost entirely unmarked.
In Tuyuca — an Eastern Tukanoan language with about 1,000 speakers in the Vaupés region on the Brazil–Colombia border — you simply cannot. Every finite verb must carry a small suffix indicating how you know what you are claiming. There is no neutral form. Speak without an evidential and you have not spoken a grammatical sentence.
The pioneering description is Janet Barnes's 1984 paper. She catalogued five evidential categories in Tuyuca, each realised as a different verbal ending:
- Visual — the speaker saw the event with their own eyes. Tuyuca: díiga apé-wi "he played soccer (I saw it)".
- Non-visual sensory — the speaker heard, smelled, or felt the event but did not see it. díiga apé-ti "he played soccer (I heard them)".
- Inferential — the speaker did not perceive the event but saw physical traces. díiga apé-yi "he played soccer (I see the muddy field)".
- Assumed — the speaker reasons from general knowledge: e.g. he always plays on Saturdays. díiga apé-hĩyi.
- Reported — the speaker was told. díiga apé-yigi "he played soccer (so they say)".
The evidential system is grammatical rather than optional etiquette. A Tuyuca speaker recounting a quarrel must mark whether they witnessed it, perceived it indirectly, inferred it, assumed it, or heard it reported. Omitting the required marker produces an ill-formed sentence.
Tuyuca is not unique. Evidentiality as a grammatical category is found across the Tukanoan family, in Quechua, Aymara, Turkish (the famous -miş), Bulgarian, Korean, many Tibeto-Burman languages, and across Amazonia. Alexandra Aikhenvald's typological survey (2004) collected nearly 500 languages with grammaticalised evidentials. But Tuyuca stands out for the granularity of its system — five distinctions where most languages make two or three — and for being among the very first to be carefully described.
The implications for cognition are debated. Some researchers, including Lev Vygotsky's intellectual heirs and Aikhenvald herself, argue that habitually marking the source of information cultivates a different orientation to evidence and rumour. Others note that English speakers can express the same distinctions — they just are not forced to. The current consensus is somewhere in between: obligatory grammaticalisation makes a category cognitively salient, but does not make speakers of other languages incapable of the same distinctions.
What is clear is that Tuyuca exposes a hidden assumption of much Western philosophy of language: that statements are by default unmarked claims of "fact". For a Tuyuca speaker there is no such default. Every sentence is, in effect, prefaced with the epistemic equivalent of "I saw", "I heard", or "I was told". A grammar of provenance is built into every verb.