Competition processes and proactive interference in short-term memory

Abstract

A competition model of proactive interference in short-term memory would predict that if subjects are forced to emit low-confidence responses, there will be an increase in both the probability of a correct response and in the number of intrusions. Dillon and Thomas ( Journal of Verbal Learning and Verbal Behavior, 1975 , 14, 603–615) recently performed such an experiment by forcing subjects to guess and found a very modest increase in the probability of a correct response at the cost of a large increase in intrusions. They suggest that these data contradict a competition model. We first argue that their results are not discrepant with the predictions of any realistic competition model and then perform an experiment which tests this same prediction in a slightly different situation. In an experiment using single-word items with conditions arranged so that the probability of a correct response is near one, subjects are run under three different speed-accuracy trade-off conditions: high accuracy, intermediate, and high speed. A competition model would predict that when subjects are forced to respond quickly, there will be an increase in errors, and these errors will tend to be from recent past items. This prediction was confirmed.