Repetition priming: The roles of instance-based and abstract processing in word recognition

Repeated exposure to a word makes it easier to identify in reading (the repetition priming effect). At times, repetition priming effects are moderated by memory for the details of past experience (e.g., visual detail). When the visual appearance of a word changes across presentations, repetition priming effects are sometimes smaller than when appearance is constant across presentations. This decrease in repetition priming due to changes in visual appearance will be referred to as the visual specificity effect. At times, however, the visual specificity effect does not occur. The present experiments investigated possible mechanisms responsible for the visual specificity effect. In Experiment 1, nonsignificant specificity effects were obtained in a word pronunciation task. That is, changing form from upper case (e.g., IODINE) to alternating case (iOdInE) or vice versa did not influence the magnitude of the specificity effect. In Experiment 2, the possible role that changes in task demands may play in the specificity effect was explored (Carr, Brown, & Charalambous, 1989). Changing task demands did not produce specificity effects, contrary to the view of Carr et al. (1989). However, visual specificity effects occurred when lexical decision was performed on both presentations, but not otherwise in lexical decision. Specificity effects did not occur at all in pronunciation. Experiment 3 explored a possible reason for this result, that the visual specificity effect results from a post-identification spelling verification that occurs more often in lexical decision than pronunciation. In Experiment 3 the likelihood of a spelling check was manipulated in lexical decision by changing the difficulty of the word/nonword decision--difficult decisions should be more likely to employ a spelling check to help verify the accuracy of the lexical decision. Visual specificity effects increased with the time it took to make the decision, supporting the notion that such specificity effects result from the operation of a post-identification spelling check. The possible role of a spelling check in normal reading was also discussed