Abstract
Recently, D'AMATO et al. (2) found, opposed to their hypothesis, negative effect of extinction trials on discrimination reversal learning in the rat. Their hypothesis was that some degree of extinction of original discrimination will facilitate the subsequent reversal learning. It was deduced from the notion that for the reversal learning to take place, it is necessary, first, to extinguish the old response, and then to acquire a new one. According to D'AMATO et al. 's assumption, the retardation of reversal compared with original learning is at least in the rat, attributed to the so-called cue function of reward (3) which serves to maintain responding that is appropriate to the initial discrimination.
The purpose of the present experiment was to retest the same hypothesis that D'AMATO et al. presented. Twenty chicks received color discrimination trainings, with red positive and blue negative as stimuli, in a T-maze for 10 days (Fig. 1). The daily session consisted of 10 trials. This schedule was enough to enable all Ss to reach the learning criterion of successive two days of 90% correct responses (Fig. 2). Then they were divided into two groups of 10 each on the basis of the original scores : the one (Group I ; average number of days required to reach the criterion was 7. 7, SD= 1. 95) was given 4 day extinction followed by 6 day reversal learning, the other (Group II ; 7. 5, SD=1. 57) 10 day reversal learning only, with red negative and blue positive as stimuli. The extinction trials for Group I were performed, in order that Ss of this group could respond only to the previously positive stimulus, with correction procedure. It is different from D'AMATO et al. 's experiment in which the free response procedure was employed. The daily session now again consisted of 10 trials both in the extinction and reversal trainings.
The comparison of two groups in correct response for the reversal trainings reveals a marked tendency of the positive effect of extinction trials (Fig. 5). That is, the differences between the scores of Group I and Group II on the 5 th as well a 10 th day of reversal training (including extinction days) were statistically significant (U=85, p<. 01 : U=83. 5, p<. 01, respectively) in favor of the former. This means that the hypothesis was confirmed by the experiment. It should be added here that both the latency and running time of Group I were neither lengthened nor shortened during extinction (Fig. 6 and 7 ; solid lines lst-4 th day).
To conclude, these results appear to indicate that the critical factors which determine the direction of the effect of extinction trials on reversal learning are the procedure of extinction and also the number of extinction trials probably.
