THREE EXPERIMENTS ON DISCRIMINATION SHIFTS

Experiment I To test a hypothesis that the nature of mediating responses would change from a stimulus-specific mediator to a dimension-specific mediator with increasing the amount of training, 180 kindergarten Ss were given a two-dimensional, form and color, discrimination task to one of the three training levels: 4 successive correct responses, 9 out of 10 successive correct responses, and 9 out of 10 correct plus 30 overtraining trials. After reaching the level, 5s were trained on an intradimensional (IN) shift task. Three levels of similarity in the positive stimuli were set up by varying color or form in the relevant dimension of shift discrimination: red (the same in original task), orange, and dark blue in the color relevant task, and circle (the same in original task), flower-like form, and triangle in the form relevant task. The negative cue was blue for color dimension and star for form dimension in all pairs of discriminative stimuli. The main results were that (a) the ease of IN shifts was not related to the similarity of positive stimulus values, and (b) the ease of shift learning was not dependent upon interactive effects of the degree of training and the similarity of positive cue. Thus, the present hypothesis failed to be confirmed, and the findings were discussed in relation to the training level and stimulus preference. Experiment II The same hypothesis as that in Experiment I was examined by comparing a reversal (R) shift with an IN shift. It was predicted that if an IN shift is faster than an R shift, Ss would have used stimulus-specific mediating responses, and if both shifts are learned at the same rate, Ss would have used dimension-specific mediating responses. 96 kindergarten Ss were trained on one-dimensional, color or form, discrimination task to one of the following criteria: 4 successive correct responses, 8 successive correct responses, and 8 correct plus 30 overtraining trials. After reaching the criterion, Ss were given an R shift task consisting of the same stimulus values as in original task, or given an IN shift task consisting of new stimulus values. The main results were that (a) the ease of R shift did not differ significantly from that of IN shift, (b) both shift performances were facilitated in terms of overtraining trials, and (c) no significant interaction was found between type of shift and degree of original training. It was concluded that the nature of mediating responses was dimension-specific without regard to the training level, and the change in the nature of mediating responses was not confirmed. Experiment III It was hypothesized from the observing response theory by Zeaman and House that with increasing the amount of original training, an IN shift would become increasingly easier whereas an EX shift would become increasingly more difficult. 120 preschool children were trained on two-dimensional, color and form, discrimination task to one of the following levels: 4 successive correct responses, 9 out of 10 successive correct responses, and 9 out of 10 correct plus 30 overtraining trials. Half of the Ss in each training level was given an IN shift task and half was given an EX shift task. The shift task consisted of different stimuli from those in the original task. The main results were that (a) the IN shift was easier than the EX shift, and (b) no significant interaction was found between type of shift and amount of original training. The results were inconsistent with the presnt hypothesis and were discussed with refererence to previous studies and the observing response model