EFFECT OF DEGREE OF ORIGINAL LEARNING AND TIME INTERVAL BETWEEN TWO TASKS ON RESPONSE SHIFT

Abstract

An experiment was designed to determine the effect of degree of original learning and time interval between two tasks on response shift.
Ninety-six boys, aged from 15 to 19 years old, were trained on a simple two-choice size (or color) discrimination problem (Fig. 1) to reach either one of the three performance levels; 3 successive correct responses, 5 successive correct responses, and 10 successive correct plus 20 correct responses. Immediately after reaching the respective learning criteria, half of Ss was given the shifted, color (or size) discrimination problem under the same stimulus sets, without added instructions. The other half was given the same shifted discrimination problem after a 30-minute rest. This is, what is called, a ‘nonreversal’ shift in which the relevant cue was reversed to a different dimension. The learning criterion in the shifted problem was 10 successive correct responses for all Ss.
Ss made almost no errors in the original problem after 5 successive correct responses had been attained (Table 1). Thus, three different criteria in the original problem seem to correspond respectively to the early, asymptotic, and overlearning stages in the original learning. Response shift was easier in the 30 minute-rest groups than in the no-rest groups (Table 2 and Fig. 2). The findings agreed with Stevenson and Weir's results. However, the effect of a 30-minute rest was significant only when Ss were trained on a criterion of 5 successive correct responses on the original problem. The superiority of response shift in the 30 minute-rest groups may be explained by weakening of S-R connections in the original problem during rest (Table 4, a), and expectancy of change in the relevant cue. Ease of response shift in the no-rest groups indicated a ∩ shaped function of degree of original learning, and confirmed our two-factor theory of learning; that is, specific S-R connections in eary stages of the original learning have negative effects, while discrimination or learning sets acquired in later stages have positive effects on the shifted discrimination learning. Based on the data obtained from the overtrained groups (10+20) some elaborations of our theory were attempted, and relationships of the discrimination set to Wyckoff's observing responses were discussed. It was indicated that relationships of resistance to extinction to the number of acquisition trials were similar to those of response shift to degree of original training.