Zhou, Y. , Todd, R. & Chow, W.
University College London
Much research has shown that, upon processing a displaced element (a "filler"), comprehenders posit a gap at the first possible site without waiting for unambiguous evidence ("active gap-filling"). However, recent findings by Wagers and Phillips (2009; 2014) suggest that this is not the case when the dependency spans a long distance (e.g., "The {mixture/stream} which the friendly hostess described that the builder skillfully plastered..."). If these results hold, they would constitute a clear exception to a long-standing generalization about how filler-gap dependencies are processed. Further, they may be taken to suggest that the anticipatory processes involved in computing such dependencies are compromised or inhibited when the filler has to be maintained in memory for longer. We attempted to replicate these findings in an eye-tracking study with a larger set of carefully controlled materials, manipulating both the plausibility and length of the dependency. Contrary to Wagers and Phillips' findings, our results (n=30) showed clear plausibility effects on multiple eye-tracking measures in both the critical and post-critical regions and, crucially, no significant interactions with dependency length in any measure or region. Such robust and immediate sensitivity to the dependency's plausibility suggests that filler-gap dependencies are computed actively regardless of dependency lengths.