Associative Interference as a Function of Degree of Prior Training in Paired-Associate Learning

Abstract

Our previous studies (Iwahara & Sugimura, 1958; Sugimura & Iwahara, 1959) showed that in learning two discrimination problems, interference on the second (color) problem increases in direct proportion to amount of prior training on a (position) problem, but decreases when the first problem has been overlearned. Maximum negative transfer appears when the initial prob-' lem has been learned to a fairly stable level of performance. Similar results were obtained in an unpublished study in our laboratory in which relevant cues on the first discrimination problem were reversed on the second problem. These findings were explained in terms of a two-factor theory of discrimination learning, according to which the initial negative transfer is attributed to specific S-R connections of Hullian type acquired in the first discrimination learning. The later facilitative effect is due to a general 'discrimination set,' learned during later overlearning stages of the initial problem. The purpose of the present experiment was to test the two-factor theory in paired-associate learning. An A-B, re-paired A-B paradigm was used because this may be considered analogous to reversal discrimination learning.