Tuesday March 17, 14:30-15:30

Episodic Influences on Word Identification

Michael E. J. Masson

(Department of Psychology, University of Victoria, Canada)

A single processing episode involving a familiar word can substantially increase the fluency with which that word is encoded on a subsequent occasion. In keeping with the principle of transfer appropriate processing, it is assumed that the increase in processing fluency created by a prior presentation will be realized to the extent that current processing invokes operations similar to those applied during the earlier encounter with the word. This assumption is used to explain results from a forced choice version of the masked word identification task (Ratcliff &McKoon, 1997): there is a bias in favor of a previously studied alternative, relative to a nonstudied alternative, but only when the two alternatives are orthographically similar (e.g., card cord). On the fluency account proposed here, the bias arises from a difference in the fluency with which the two alternatives are encoded--the studied word is encoded more fluently and this fluency provides evidence in support of that word over the nonstudied alternative. It is further proposed that when alternatives do not share any letters (e.g., card lost), subjects tend not to encode the alternatives as words, but rather analyze the alternatives with respect to presence or absence of particular letters detected during the brief display of the target. Experiments are reported that show a robust bias effect occurs only when prior study involves an encoding task that increases the fluency of subsequent word reading. Additional experiments are reported that were designed to (1) replicate the bias effect by artificially enhancing (without prior exposure) the processing fluency of one alternative in the forced choice task and (2) generate a dissociation between the free report and forced choice variants of the masked word identification task that is predicted by the fluency account proposed here but not by either the counter model (Ratcliff & McKoon, 1997)or the REM model (Schooler, Shiffrin, & Raaijmakers, in press).