Aparicio , H. . , Kennedy, C. . & Xiang, M.
University of Chicago
Referential Contrast Effects (RCEs) involving adjectivally modified Noun Phrases (e.g., the tall glass) have been attributed to Gricean quantity-based reasoning, which penalizes the use of redundant modification (Sedivy et al. 1999; Sedivy 2003, 2004). We argue that RCEs are further shaped by the grammatical properties of the different adjective classes, which determine their degree of context-sensitivity, and hence the speed at which contextual information is integrated during online processing. We present results from a Visual World (VW) eye-tracking study testing Color (CAs, e.g., blue), Relative (RAs, e.g., tall), and Maximum/Minimum Standard Absolute Adjectives (MaxAAs/MinAAs, e.g., empty and spotted). The instructions used in the VW-study were rated for their perceived degree of informativity in each of the visual contexts tested. CAs, RAs and MaxAAs were deemed too informative when used redundantly compared to the condition testing contrastive uses (all p<0.004). No such effect was observed for MinAAs (p>0.3). Critically, however, the online eye-tracking data shows that the RCEs for CAs, RAs and MaxAAs present different time-course patterns, suggesting that perceived informativity is not the sole source of the RCEs. We discuss how the lexical semantics of the different adjective classes also contributes to RCEs.