Ness, T. & Meltzer-Asscher, A.
Tel Aviv University
The "Active Filler" hypothesis for the processing of long-distance dependencies assumes active maintenance of the filler and search for a gap site. Thus, the processing of these dependencies is predicted to incur increased working memory (WM) demands between the filler and gap sites. However, most online effects of the processing costs of dependencies were demonstrated at the gap site, hence reflecting retrieval costs rather than filler maintenance costs, which were not demonstrated behaviorally. In the current study, we increased WM load by using similarity-based interference, in order to identify filler maintenance costs.
Using a self-paced reading paradigm to measure effects on online reading times (32 participants, 24 sets), we compared object relative sentences, in which a filler is maintained between the filler and gap sites, with NP-ellipsis sentences, in which there is no indication for the dependency prior to the retrieval site, and hence no filler maintenance. Interference was manipulated by the addition of an interfering NP (similar to the filler in grammatical and semantic features). A significant interaction was found between sentence type and interference, driven by significantly increased reading times due to interference only in the object-relative condition. These results offer behavioral evidence for filler maintenance costs.