抄録
In order to test our two-factor theory of discrimination learning, five groups of rats were given a simple position discrimination learning while motivated by electric shock. The degree of the initial training was different from group to group. As soon as the respective criterion of learning had been achieved, the rats were given reversal learning.
As was expected from our theory, a maximum negative transfer appeared when the prior task had been well learned (Group III). The negative transfer decreased in either direction. The general trend was identical irrespective of varying degrees of learning criterion in the second reversal task. Thus UNDERWOOD's hypothesis was not confirmed, which predicts a maximum interference when the two tasks are equally learned.
The present results also disagree with CAPALDI & STEVENSON'S theory in terms of generalization decrement, since the negative transfer increased with the initial increase of amount of the prior training. On the contrary, our findings confirm our previous studies with human subjects and have reinforced our position that specific S-R connections are predominant in early stages of learning but more general set or attitude is acquired during later overlearned 'mature' stages. Thus our theory is shown to be closely related to HEBB's schema of the development of the learning capacity (2).
It was also shown that reversal learning was easier for the fast learners than the slow learners of the first task. Further research is necessary to clarify this point.
