The study attempted to minimize nonspecific response strategies which supposedly mask the positive effect of perceptual pretraining on initial discrimination learning within the predifferentiation paradigm. The subjects were 44 first- and second-graders. Experimental-group subjects received rules learning (RL), pretraining, initial discrimination (ID), and reversal shift (RS) phases. Rules learning consisted of a two-choice discrimination using stimuli varying on shape and orientation-of-line, and was designed to teach subjects to associate a specific dimensional value with the correct response. During pretraining, subjects made "same-different" judgments on pairs of stimuli varying on height and brightness. The ID and RS phases again involved a two-choice discrimination with a subset of the stimuli used in pretraining. Three additional groups controlled for the effects of RL, pretraining, and RL plus pretraining. The results showed that subjects given RL learned the ID phase more rapidly than subjects not given RL (p less than .01) Pretraining did not facilitate ID learning. Neither RL, pretraining, nor a combination of RL and pretraining facilitated RS learning. Similarities between the present study and learning-set literature, the failure of perceptual pretraining to facilitate ID learning following RL, and the failure of RL, pretraining, or both to affect RS learning are discussed. (Author/CS)