Listeners' interpretations of referring expressions are influenced by referential precedents-temporary conventions established in a discourse that associate linguistic expressions with referents. A number of psycholinguistic studies have investigated how much precedent effects depend on beliefs about the speaker's perspective versus more egocentric, domain-general processes. We review and provide a meta-analysis of visual-world eyetracking studies of precedent use, focusing on three principal effects: (1) a same speaker advantage for maintained precedents; (2) a different speaker advantage for broken precedents; and (3) an overall main effect of precedents. Despite inconsistent claims in the literature, our combined analysis reveals surprisingly consistent evidence supporting the existence of all three effects, but with different temporal profiles. These findings carry important implications for existing theoretical explanations of precedent use, and challenge explanations based solely on the use of information about speakers' perspectives.
Keywords: PERSPECTIVE-TAKING, META-ANALYSIS, PRAGMATICS, CONVERSATION, EYE TRACKING REFERENTIAL PRECEDENTS IN SPOKEN LANGUAGE 3Referential precedents in spoken language comprehension: A review and meta-analysis One of the central questions in research on spoken language communication concerns how listeners derive context-specific interpretations from the linguistic form of spoken utterances. Explaining how this derivation works is challenging because of there is a many-to-many mapping between linguistic utterances and a given speaker's communicative intention in a particular setting. This complexity is evident in referential expressions such as the small candle, which could be used by a speaker to refer to many different individual candles of varying size. The linguistic content of such an utterance provides information only about the type of entity that is being referred to, leaving it up to the listener to decipher which particular token the speaker has in mind. How do listeners identify relevant contextual information, and how is this information brought to bear on language processing?Given that understanding a speaker's intended meaning is of paramount importance in dialogue, some theorists have argued for a central role of mutually shared information or common ground (Clark & Carlson, 1981;Clark & Marshall, 1981). Common ground is a special kind of shared information that is not only shared, but is also known (or believed) to be shared. According to this view, listeners employ various co-presence heuristics to identify common ground, which include perceptual co-presence, linguistic co-presence, and community membership. For example, listeners can know that a particular candle is the referent of small candle because (1) it is the smallest candle that the interlocutors can see (perceptual co-presence); (2) it was previously mentioned in the current discourse in the presence of the current interlocutors (linguistic co-presence); or (3) the interlocutors are membe...