Probing the Ignorance of Epistemic Indefinites: A (Non)-Familiarity Constraint

Epistemic Indefinites (EI), with an added ‘ignorance’ component not seen with ordinary indefinites, are licensed in broadly two contexts. In one, there is ignorance of the witness of the existential claim among a plurality of referents in the domain. In the other, the witness of the existential claim can be identified, but there is ignorance about certain aspects of this witness. The first context is analysed as domain widening of some sort (Alonso-Ovalle and Menendez-Benito 2003, et seq). The second context is analysed as domain shifting (Aloni & Port 2010), with two identification methods at play, one required for knowledge (the ignorance component) and one used for specifying of the EI. Available identification schemes are naming, description, & ostension. We focus on the second context, and show that the Telugu EI [eed-oo]/[evar-oo] (which-DISJ/who-DISJ) ‘some-thing/body’ can be used even when the speaker has access to all 3 methods of identification, which should not be possible going by the domain shifting account. We propose a solution along the lines that if the speaker has recognized the person, [evar-oo]/[eed-oo] cannot be used. Similar to the familiarity theory of definiteness with discourse referents (Karttunen 1976) and ‘file cards’ (Heim 1983), we formalise this notion using mental referents and mental files (Recanati 2013, 2016). untranslate