1979 年 50 巻 2 号 p. 89-96
The direction of feedback in the selection strategy situation has traditionally been out of control. In the present paper, the probability that the selected instance is positive was redefined so as to control the direction of feedback, based on Johnson's non-predetermined solution technique, in terms of the average rate of hits denoted by τ. To evaluate the effectiveness of τ, the effects of three conditions (τ=0, 1/2, 1) upon the strategy-learning process were experimentally examined. The results showed that the selection distributions varied, as the problems proceeded, toward the optimal strategies defined under each value of τ.