Abstract
64 subjects were required to learn a concept identification task. One half of the subjects learned a task in which one of the irrelevant dimensions was highly (p = .75) redundant with the relevant dimension and one-half of the subjects learned a task in which all irrelevant dimensions were normally (p = .5) redundant. One half of each of these groups received additional pretraining. The major findings were that mean response latencies were longer and acquisition was slower for groups trained with a highly redundant irrelevant dimension than for groups trained with normally redundant irrelevant dimensions. These results were interpreted as supporting an hypothesis that sampling sets remain large for groups trained with a highly redundant irrelevant dimension and are the basis for slower acquisition.