Incremental sentence processing

What is incremental processing?

An extensive volume of work has been carried out to address the question of whether the human sentence processor processes linguistic inputs incrementally; numerous recent sentence processing models have supported incrementality at some level. For example, some of the most prominent aspects of incrementality in sentence processing are that the incoming linguistic inputs are processed without delay almost on a word-by-word basis, that relevant constraints are applied in parallel to the analysis of the unfolding input, and that all relevant analyses of the input are specified to some degree (e.g. Altmann & Steedman, 1988; MacDonald, Perlmutter & Seidenberg, 1994). However, not all these different aspects of incremental processing are presupposed in the psycholinguistic literature. For example, structural analyses of some inputs are delayed until a certain grammatical item is encountered (e.g. Pritchett, 1991), the utilization of certain types of information (e.g. “thematic” information) is delayed (e.g. certain two-stage models: Rayner, Carlson & Frazier, 1983; Mitchell, 1987), or detailed specification of certain grammatical relations (for example, dominance and precedence) is delayed (e.g. underspecification models: Marcus, Hindle & Fleck, 1983; Weinberg, 1993). Historically, incrementality in sentence processing has only become widely assumed in the past decade, and there are several aspects of incrementality which, although often assumed, have not been explored empirically.

However, even a class of approach to sentence processing that has been (perhaps most) widely proposed and demonstrated in recent years, constraint-satisfaction models (e.g. MacDonald, Perlmutter & Seidenberg, 1994; Trueswell & Tanenhaus, 1994), has yet to explicitly offer a satisfactory explanation for the issue.