Transfer of Discrimination Learning in Animal Psychology

This article consists of two parts; the first concerns the problem of generalization, and the second the problem of experimental extinction. I. Generalization of discrimination a) Transposition. In the beginning, we have discussed the problem of transposition. The object of this discussion is to determine whether or not the animal respond to the positive or negative stimuli, in absolute or relative manner, during the original discrimination learning. From recent experiments in this field, we may expect the following conclusions : 1. When the pattern of the stimuli in the original training is similar to the pattern of the test stimuli, the animals will respond in relative manner. 2. When the pattern of the stimuli in the original training is not similar to the pattern of the test stimuli, the animals will respond at random or in absolute manner. 3. In the case of human subjects, the response of older subjects depends mainly on the pattern of the test stimuli, while the response of younger subjects depends mainly on the pattern of the original stimuli. Of this the following interpretation may be given. When the subject is higher in the developmental scale, they can respond in both relative and absolute manner. So, they respond in relative or absolute manner, according to the pattern of test stimuli. As K. Murofushi and R. Motoyoshi (55) have shown, this conclusion is in agreement with the mathematical view proposed by James (34). 4. a) The relation between language and transposition has been studied by Kuenne (39) and Erenfreund (11) etc. But unfortunately they have not yet reached any clear result. b) The generalization of single stimulus. At present, the knowledge about the exact shape of the generalization of stimulus is not yet very clear. c) Next, we have discussed the problem of generalization of the stimuli, which has two aspects : the genralization built up separately to each particular stimulus and the generalization of the pattern of these stimuli (Nissen 57). In this case, it will be concluded that the degree of generalization depends on the degree of training and the dynamics generated between two generalizations. II. Extinction of discrimination learning a) Reversal learning. Experimental extinction of discrimination learning has been studied by many scholars in reversing the positive and negative stimuli. The results could not always prove that the numbers of reinforcements in order to master the reversal learning were inversely proportional to the numbers of reinforcements in the original learning. According to Reid (66), the overlearning group has learned more rapidly than the group which just reached the criterion in acquisition process in reversal learning, though the former persisted in responding to the previously positive stimulus longer than the latter. We think two factors might be at work : 1) a factor which inhibits the position habit, 2) a factor which establishes a pattern of cues or the distinctiveness of cues during the original learning. Further studies are needed in order to determine which factor is more dominant. b) Repeated reversal learning. Hitherto the interproblem improvement has been interpreted in terms of the habit dominance theory, of the perseverative cue theory and the acquired distinctiveness of cue theory (North 60, 61). But we consider that the interproblem improvement might be interpreted rather in terms of the formation of learning sets, as reported by Harlow (26) in regard to the discrimination learning of monkey. Here also, further studies are needed to determine how the learning set can have the effect on generalization and extinction of discrimination learning