Two types of inference in evidentials: Efficacy vs. doxastic worlds

Abstract

In this paper we propose that Korean evidential -te- requires the inference from directly perceived evidence, and this inference should be based on efficacy (Copley and Harley, 2015). Then we show that this proposal can explain the apparent variable evidentiality of -te- between direct evidentiality and inferential evidentiality (Chung, 2007; Lee 2013; a.o.) without assuming relative tense (Lee, 2013; Smirnova, 2013; Koev, 2017; a.o.). We also argue that this proposal can easily account for the differences between -te- and other closely related categories such as epistemic modals and predictive futures. In addition to this, given that -te- introduces he speaker’s weak commitment due to its inference, we try to explain this weak commitment in terms of the scoreboard semantics (Farkas and Bruce, 2010; Malamud and Stephenson, 2015; a.o.). We also discuss the implication of our proposal, especially with respect to the typology of evidentials.