Effects of Instructional Accuracy on a Conditional Discrimination Task

The effects of instructional accuracy on performance using a complex conditional discrimination procedure were studied in 23 humans. The task was to push 1 of the 9 number pads of the keyboard corresponding to 9 stimuli presented on a computer display randomly. In Phase 1, subjects were assigned to 1 of 3 groups with different instruction varying in degree of accuracy: accurate, partially accurate, and minimum instructions. All groups then were required to perform a conditional discrimination task without instruction in Phase 2. Both phases ended when subjects made correct responses on 18 successive trials, if subjects could not satisfy this criterion, the phase ended after 720 total trials. In Phase 1, all subjects accomplished the task within 270 trials. In Phase 2, subjects in the accurate and minimum instruction groups accomplished the task within 450 trials. However, the subjects in the partially accurate instruction group needed more trials than the other groups and 4 subjects did not complete task within 720 trials. The results indicate that the partially accurate instruction interfered with acquisition of an optimal responding under new contingencies, suggesting that performance depends upon not only current contingencies, but also the history of the subject.