Interactions of length and overlap in the TRACE model of spoken word recognition

AbstractWhat determines degree of competition among phonologically similar words? One proposal is that proportion of overlap predicts competition independently of word length. We argue that proportion of overlap may provide descriptive adequacy, but does not provide an explanation. We show that TRACE correctly predicts patterns previously attributed to proportion of overlap. In additional simulations, with independent manipulations of word length and proportion of overlap, proportion of overlap fails to predict the full pattern of results. We discuss how competition dynamics in TRACE modulate competition as word length and proportion of overlap change. These results have implications for theories of human spoken word recognition, and will motivate experiments to test these new TRACE predictions.

Return to previous page