The purpose of this talk is to provide a semantic analysis of the mirative evidential element ra'e of Guaraní (Tupian; Paraguay, Argentina, Bolivia and Brazil). Ra'e, described as “non-expected evidential marker” by Tonhauser (2006) and considered to be various types of past in earlier sources, has inferential evidential, mirative, and [relative] past senses, often overlapping.
We claim that ra'e is essentially a mirative operator. Mirativity is descriptively defined as indicating that a proposition is new or surprising (DeLancey 1997), and has been formalized (cf., e.g., Rett and Murray 2013) as containing exclamative illocutionary force and contradicting a previous expectation. In such an approach, miratives are roughly an exclamation. We claim that the mirativity of Guaraní ra'e is better approximated by a sentence of the form “it turns out that P”, which lacks exclamative force, and simply encodes acquisition of the evidence that allows the speaker to believe a proposition.
Such formulation allows us to account for the fact that the speaker's discovery can be anchored in a time other than the present, which, in line with previous work on Bulgarian (Smirnova 2013) and Korean (Lee 2012) evidentials, could be called Evidence Acquisition Time (see also Fleck 2007 for an application of a similar notion to Matsés inferential evidentials).
In fact, ra'e is such that the evidence that changes the speaker's belief has to be indirect (at least with punctual eventualities). However close this makes ra'e to an indirect evidential, the mirative component of ra'e is necessary to explain the fact that it can't be used as a pure narrative evidential. We derive the inferential flavour ra'e by claiming that this particle lexically encodes a requirement of non-simultaneity between the event itself and the acquisition of the evidence for the event, something that is also responsible for its past tense interpretations.