ABSTRACT
The purpose of this paper is to examine the nature of perceptual learning by manipulating a number of priming conditions to determine which produce the optimal perceptual learning. In addition, several models of perceptual learning are tested. A relative model which assumes learning occurs relative to baseline is superior to the usual absolute model in predicting moments of the threshold distributions. I also consider two other sources of learning in picture fragment completion: Pure guessing, which I reject on statistical grounds, and paired associate learning between a fragment and its name. The latter is shown to underlie learning when the most fragmented stimulus is used as the prime.
The relationship between recall and fragment completion is examined in the light of likely test transfer effects. Although functional dissociation is observed between recall and completion performance, stochastic dependence is observed in conditions where it is expected—that is, when the recall test precedes the completion test.
