The Elimination of Irrelevant Information in a Multidimensional Stimulus Classification Task

Abstract

Two experiments were conducted to investigate the effect of providing stimulus control on classification performance. In the first study, subjects were given control of various stimulus dimensions during a series of seven sessions. There was no difference in classification accuracy between the group having stimulus control and the group having no stimulus control. A significant positive correlation was obtained between the amount of information eliminated and classification accuracy. This finding suggested that subjects were able to filter some stimulus information overtly while increasing classification accuracy. The second study was designed to determine if the elimination of information produced a more “efficient” stimulus display. Subjects viewed either a “filtered” display created by subjects in Experiment 1, or the original “nonfiltered” display. Results showed that subjects viewing the “filtered” displays were significantly more accurate in classifying stimuli during the first session. Further, there was a significant correlation between the accuracy of subjects producing the “filtered” display and the subjects utilizing it. The results were discussed as a procedure for isolating the necessary information parameters for decision making.