Kush, D. 1, 2 & Van Dyke, J. A. 2
1 NTNU, Trondheim
2 Haskins Laboratories
How do pronouns identify their antecedents among items in memory? Unrestricted Access (UA) models (e.g. Foraker & McElree, 2007) assume that the parser identifies NPs as potential antecedents primarily on the basis of low-level feature-match with a pronoun, irrespective of the structural prominence of those NPs. In contrast, Ordered Access (OA) models assume that NPs are considered as potential antecedents in an order (at least partially) determined by their structural prominence (e.g. Gordon & Hendrick, 1999). OA predicts that prominent NPs should be accessed more rapidly than non-prominent NPs, while UA predicts no speed advantage for prominent antecedents. We used the MR-SAT paradigm to test whether participants were faster to access a feature-matching NP as the antecedent of critical pronoun when it was the main subject of the preceding sentence than when it occupied a less prominent position. We found that prominent antecedents were more available for retrieval, as measured by asymptotic d' (Prominent: d'=2.65; Non-Prominent: d' =1.76, p<.001). Consistent with OA predictions (contrary Foraker & McElree, 2007) we found that prominence had a reliable effect on speed of access: prominent antecedents were accessed more quickly than their non-prominent counterparts (p<.05).