direct evidentials, case, tense and aspect in tibetan: evidence for a general theory of the semantics of evidential

tibetan has three different morphemes expressing direct evidentiality. only two of the three have been described in any detail, and the distinctions among these morphemes are described in quite different ways by different authors. we argue that careful study of these morphemes reveals that evidentials do not encode evidence type per se. instead, they encode a relation between the situation being reported by the speaker and the situation within which evidence was acquired. this approach turns out not only to provide an accurate and systematic characterization of the different tibetan direct evidentials, but also to predict a number of seemingly unrelated restrictions on their syntactic distribution. the distribution of these direct evidentials hence provides strong support for the proposal of speas (2010) that evidentials of all categories encode relations among situations. since tibetan evidentials operate at the illocutionary level, our analysis further suggests that illocutionary force is best modeled not as a feature of situations per se, but rather as a relation between relevant situations.