幼児の過剰訓練逆転効果におよぼす刺激次元数の効果

Abstract

The present experiment was designed to test a hypothesis that magnitude of the overtraining reversal effect (ORE) is positively related to number of stimulus dimensions in the discrimination problem. This hypothesis was deduced from attention theories in discrimination learning (Lovejoy, 1966; Mackintosh, 1965), which assume that discrimination learning involves the acquisition of mediating attention responses to the relevant dimension and selective responses to the positive stimulus.A 2×3 factorial design (i.e. 6 groups) was used, which incorporated degrees of original training (criterion and overtraining) and number of stimulus dimensions (1, 2, and 3). The Ss were 144 preschool children with a mean age of 5 years and 7 months, and they were assigned to one of the 6 groups equating in age and sex. Half of the Ss were trained on a 2-choice simultaneous discrimination problem to a criterion of 10 successive correct responses and the rest were given additional 50 overtraining trials after reaching the criterion. Immediately after completing the original discrimination, they were given a reversal discrimination to a criterion of 10 successive correct responses or 100 training trials. Throughout the experiment, S's responses were to point out one of the 2 stimuli on each card, and reinforcements were provided by E's saying “Hit” or “Miss”. The Ss were trained on one of the 3 discrimination problems: (1) one-dimensional problem consisting of color, form, or size dimension, (2) 2-dimensional problem consisting of color-form, form-size, or size-color dimensions, and (3) 3-dimensional problem consisting of color-form-size dimensions (Fig. 1). The positive stimulus and relevant dimension were counterbalanced within each problem.The main results were that (a) the original discrimination was more difficult with increasing the number of stimulus dimensions (Table 1), and (b) for the reversal discrimination, a significant interaction was obtained between the degree of training in originall discrimination and the number of stimulus dimensions (Table 2). This interaction indicated that (a) the magnitude of the ORE was positively related to the number of stimulus dimensions, and (b) the performance difference due to the number of dimensions was statistically significant when Ss were trained to reach the criterion in original discrimination, but there was no significant difference for the overtrained groups.The results supported the present hypothesis. The role of irrelevant dimensions in the ORE was emphasized in relation to the attention theory, and the relationship beween problem difficulty and the ORE was discussed.