The present study reports three experiments designed to investigate the effects of drive on performance in selective learning.
In Exp. I, twelve rats were run to reversed goal each day, receiving 5 trials per day for 20 days in a T-maze. After 10-day training Ss were satiated and then given drive-test for 10 days. On the first day of the drive-test Ss received satiated trials, next 7 days they were deprived of all food except rewards in the goal and on the last two days satiated again. During the drive-test the percentage of correct responses was an inverted U-function of drive (Fig. 2). The performance under 48-hrs of food deprivation was significantly better than under both 0-hr and 96-hrs. The mean log running time was computed separately for positive and negative runs to confirm Reynolds' R-hypothesis (Fig. 3). It appeared that the differences between the positive and negative mean log running time varied with the percentage of correct responses.
In Exp. II, high drive group of 24 Ss and low drive group of 23 Ss received 5 trials per day for 10 days using a T-maze. Each drive group was divided into two groups: one given a choice cue and the other not. The difference between two drive groups was significant in the mean log running time but not in the number of correct responses. The choice cue had no effect on mean log running time but some effect on the choice (Table 1, 2 and Fig. 5). The drive effect on performance seemed to be greater under the no-cue condition than under the cue condition. The differences in mean log running time between the positive and negative runs were shown in Fig. 4, revealing some relation between these differences and z-values of the percentage correct.
In Exp. III, two drive groups of 22 Ss (18-hrs and 42-hrs of food deprivation) were run for 14 cycles of 12 trials each, using a Y-type discrimination apparatus in which the positive side had a curtain but the negative side did not. The number of positive and negative responses was equated by means of forced trials. Curves of percentage of correct choices showed some difference in favor of the 18-hr group (Fig. 6), but the drive effect was not significant (Table 5). The 42-hr group was significantly superior in speed to the 12-hr group and the interaction of drive with discriminanda (curtain) was significant (Fig. 6 and Table 4). The z-values of the percentage of correct choices were positive functions of the differences in the mean log running time again (Fig. 7).
Thus, drive effects on choice behavior were related to the effect of drive on response speed. And these results of three experiments were interpreted as supporting the R-hypothesis. However, the interference of high drive with choice performance was explained by the drive-disinhibition hypothesis, proposing that drive released negative response tendency from conditioned inhibition. After all, many conflicting results of drive/learning experiments would be explained by both the R-hypothesis and the drive-disinhibition hypothesis.
