Semantic expectancy effects on word access

Summary Access to word-representations in memory was studied in an experiment that used a lexical decision task: Ss had to decide whether a string of letters was a word or a nonword. Independent variables were context-similarity and semantic expectancy. The former is defined in terms of the categorical relationship that holds between a set of context words and the subsequently presented test word. The second refers to the cue-value given to subjects as to the likelihood with which related or unrelated test words would be presented. Cue-values did not provide information on the likelihood with which word or nonword decisions were required, but only on the most probable semantic relationship holding among context and test words. In theory, Meyer and Ellis' (1970) race model for word access was extended for the present purpose. It is shown that the two search processes assumed to mediate lexical access interfere on the basis of limited processing capacity, at least under the special conditions prevailing during the experiment. Furthermore, it is shown that Ss effectively manage to control the point in time at which they decide to issue a nonword response. Cuing for related test words tends to produce faster nonword decisions than cuing for unrelated words.This paper reports work that was conceived and begun at Stanford University, Calif., USA, where the author spent a year as a post-doctoral fellow. This fellowship was made possible by the Deutsche Forschungsgemeinschaft, Bad Godesberg, Federal Republic of Germany, Grant Schm 350/1. I am grateful to R.C. Atkinson, S. Monsell, P. Matthews, and D. Vorberg for supporting this research in a number of ways.