Abstract
The purpose of the present study was to investigate the effects of overtraining on the subsequent discrimination shift learnings by using existence or nonexistence of an antecedent relevant dimension in the subsequent discrimination learning as an experimental variable in a concurrent discrimination situation, in which subjects were concurrently given training of a discrimination task A on some trials and task B on other trials.
A 5×2 factorial design was used, incorporating shift types (W, P, HEDS-I, HEDS-II and EDS) and degree of overtraining (NOT and OT). The subjects were 60 Kindergarten children with a mean age of four years and one month, and a mean IQ: II5.3. They were assigned to one of ten groups of some age, sex and IQ. Group NOT was trained on two independent discrimination tasks to a criterion of five successive correct responses. Group OT were given an additional 24 overtraining trials after reaching criterion. Immediately after completing the original learning, all subjects of every group were divided then into five subgroups: W, P, HEDS-I, HEDS-II and EDS. Group W was run under whole reversal shift learning condition where both _??_ discrimination task and _??_ discrimination task were reversed. Group P was run under partial reversal shift learning condition where _??_discrimination task was reversed, and _??_ discrimination was unreversed. Group HEDS-I Was run half extradimensional shift learning condition-I where _??_discrimination task was unreversed and _??_discrimination task was replaced by a new discrimination task (_??_=discrimination task). Group HEDS-II wag run under half extradimensional shift learning condition-II where _??_discrimination was reversed, and _??_ discrimination task was replaced by a new discrimination task (_??_=discrimination task). Group EDS was run under extradimensional shift learning condition where both _??_discrimination task and _??_ discrimination task were replaced by new discrimiantion tasks (_??_=discrimination task and discrimination task). Training of the shift phase continued to a criterion of five successive correct responses. Reinforcement was provided by yellow light and buzzer for correct responses or the experimenters saying “Hit”.
The following were the main results.
(1). Group W showed a tendency for the reversal learning to be facilitated by overtraining, but the effect of overtraining did not reach the statistically significant level of 5%. The reversal learhing of Group P was delayed by over training, but the reversal learning of Group HEDS-II was facilitated by overtraining.
(2). The extradimensional shift learning of Group HEDS-II was facilitated by overtraining. Group HE DS-I showed a tendency for the extradimensional shift learning to be facilitated by overtraining, but the effect of overtraining did not reach the statistically significant level of 5% (see FIG. 3). The extradimensional shift learning of Group EDS was neither facilitated nor delayed by overtraining.
(3). The learning process on _??_ discrimination task in Group W was significantly different from the same task in Group P under overtraining condition, but there was no difference in the learning process on the same task between Group W and Group Punder nonovertraining condition. There were no differences in the learning Process on _??_= discrimination task among Groups HEDS-I, HEDS-II, EDS under both overtraining and nonovertraining conditions.
These results demonstrate that the existence or nonexistence of an antecedent relevant dimension in the subsequent discrimination shift learning is an important factor as a determinant of the effect of overtraining on the subsequent discrimination learning, and that overtraining facilitates not only areversal shift learning but also an extradimensional shift learning when an antecedent relevant dimension remains as a relevant dimension in a subsequent discrimination learning.
