ABSTRACT
Many investigators working on language processing are coming around to the idea that the field should adopt a more naturalistic and ecological approach to psycholinguistics (Pickering & Garrod, 2004; Trueswell & Tanenhaus, 2005). The trick is to hold on to the insights and rigorous methodology associated with the formal, representation-based psy-cholinguistic tradition while expanding the range of phenomena to include those found particularly in naturalistic speech and human conversation. For instance, people gesture while they talk, but there is little work on the way that linguistic representations are built and coordinated with the production of gestures (for an important exception, see Levelt, Richardson, & La Heij, 1985). Similarly, when people talk, they are often disfluent (Clark, 1996): They produce fillers such as uh and um, and they repeat words, backtrack, abandon utterances, start them over, and so on. In our recent work, we have begun to investigate how disfluencies affect the parser’s structure-building operations (Bailey & Ferreira, 2003, 2005; Ferreira & Bailey, 2004). We have not only demonstrated that disfluencies have a systematic effect on parsing, but we have also shown that it is possible to study a phenomenon of the “wild” like disfluencies in a way that retains the assumption that formal linguistic structures are built during processing, and in real time.
