Structural and lexical constraints on filling gaps during sentence comprehension: A time-course analysis

Reaction time and speed-accuracy trade-off procedures were used to examine when different linguistic constraints were operative in processing sentences with filler-gap dependencies. Experiments measured the time to assess the acceptability of structures with anomalous filler-gap dependencies stemming from violations of configuration (syntactic) constraints or local lexical constraints. Full time-course data indicate that configurational (island) constraints were operative 200-400 ms before local lexical constraints (i.e., subcategorization and thematic role restrictions). These results suggest that filler-gap assignments are determined by processes that appeal first to general syntactic information and only later to specific lexical information. The time-course data indicate that the parser does not posit potential gap sites within syntactic islands, which in turn motivates a restricted version of a first-resort model of filler-gap processing.