Letters are functional in word identification.

Are words identified as if they are single-unit patterns, or does letter processing mediate word identification? In the main, accuracy experiments have been found to support the mediated model, whereas recent reaction time experiments have been interpreted in favor of the single-unit pattern hypothesis. In the latter experiments, subjects could search for a target word in a test word as quickly as for a target letter in a test letter. The analysis of these results, however, did not consider the role of parallel processing of individual letters, lateral masking, limited capacity, and similarity between the target and test alternatives. In the present experiment, the advantage of parallel processing of letters in words was attenuated by manipulating similarity. The reaction times showed a processing advantage for single letters over words. This result as well as all of the previous results can be described in terms of processing mechanisms of the mediated model, the same mechanisms that have already been utilized to describe the accuracy of letter and word identification in tachistoscopic experiments.