Some philosophers argue that pure quotations can refer to different kinds of objects, including expressions, signs, pronunciations, concepts, and meanings (Washington 1992; Saka 1998; García-Carpintero 2003, 2017, 2018). In this paper, I criticise extant accounts of this feature of quotations (García-Carpintero 2017, 2018; Saka 1998, 2006; Gómez-Torrente 2011) and propose an original version of Davidson’s (1979) theory that accounts for it.

My proposal consists of two claims. First, a quotation is made up of (i) a demonstrative pronoun (surfacing as the quotation marks) and (ii) a truth-conditionally inert expression that is displayed to make itself, or an entity related to it, available for being referred to by the pronoun. Second, a quotation has a referent only relative to a linguistic context: the appropriate object of demonstration, and hence the referent of a quotation, is selected by the predicate in whose argument position the quotation occurs. I formally represent the predicate’s capacity of selecting a referent as a contextual restriction imposed by the predicate.

Additionally, I argue that my proposal allows me to defend a preliminary account of reports containing the so-called ‘quotative ‘be like’’, thus filling a significant gap in the philosophical and semantic literature on quotational contexts.