Discrimination of complex stimuli: the relationship of training and test stimuli in transfer of discrimination.

Abstract

From a theoretical analysis of discrimination learning in terms of implicit observing responses it was predicted that transfer of discrimination training from one task to a second task would be positive when the stimuli employed were distinguished by the same property in both tasks, and that transfer would be negative when the stimuli were distinguished by different properties in the two tasks. To test these predictions, comparisons were made of discrimination performance during paired-associates learning following three different conditions of prior familiarization training-(a) training stimuli not identical to the test stimuli, (b) training stimuli not identical to the test stimuli but distinguished by the same property as the latter, (c) training stimuli not identical to the test stimuli and distinguished by a different property than the latter-and a control condition without prior training. Forty human Ss served under the four conditions. Results of analysis of variance of the number of trials to mastery under the four experimental conditions were in complete agreement with expectation: no difference was obtained between the first two familiarization conditions; under both of these conditions positive transfer was obtained relative to the control condition; and under the third familiarization condition negative transfer was obtained relative to the control condition. Results obtained with frequencies of correct responses and intrusion errors were complicated by differences among the four conditions in the total number of responses. However, a corrected measure-number of correct responses minus number of errors-gave results identical to those obtained with trials to mastery.