補強配合の効果について

抄録

Our problem in this paper is to examine the nature of Humphreys' effect, the so-called ‘Paradox of reinforcement’. Unfortunately, there is no unified theory to account for this phenomenon. At present, the main alternatives are two: one is Humphreys expectation principle, the other is Hull's after-effect (generalization) principle. The first step in our research was directed to the examination of the former principle based on experiments.
Since the expectation principle regards the extinction process as ‘another process of learning’, it tends to solve the problem by referring mainly to discrimination or cognition of the difference between acquisition and extinction situations. Extending the above principle, however, it would become difficult to explain the other well-known effect of reinforcement, namely, the more numerous the number of trials, the greater the resistance-to-extinction in case of continuous or 100% reinforcement. In order to make this point more precise and to get some indirect but important suggestions relating to Humphreys' effect, the comparison between extinction process and ‘another process of learning’ or relearning process would be required.
The experimental arrangement used here was ot the card-game type and consisted mainly of the anticipatory responses of the human subjects. By setting the relearning series which was as simple as possible and comparable to the procedure of the extinction experiment, we studied: (1) The after-effect of mixed (or percentile) reinforcement.
(2) In case of continuous reinforcement;
a) some effects of the number of trials upon the acquisition level and relearning process.
b) the correlation between the acquisition level and resistance-to-relearning.
As a result of our study, it was revealed that the after-effect of mixed reinforcement showed Humphreys' effect dominantly both in the properties of the relearning process as a whole (α-effect) and the initial gradient of the relearning process (β-effect). However, the after-effct of continuous reinforcement showed, as we had expected, a high but negative correlation between the acquisition level and resistance-to-relearning which is a result opposite to the one derived in the extinction situation.
Accordingly it has been established that the simple mode of the expectation principle, applied to the extinction process, can hardly be supported. Moreover, even with learning and relearning situations it might not be profitable to assert the principle so vaguely. Finally, our data revealed the more following points:
(1) The relation between the learning process and individual difference.
(2) The properties of the relearning process following mixed and continuous reinforcement respectively.