The relation between cue salience and learning rate in concept identification. A test of the Trabasso-Bower model

Abstract

In this investigation, Trabasso and Bower's Hypothesis that the cue sampling probability of the relevant cue in a concept identification task determines the rate of learning, was tested. Cue sampling probabilities were estimated on the basis of a cue preference test. These estimates were used to predict learning rate. There were six groups of subjects: (1) concept identification followed by the cue preference test; (2) concept identification preceded and followed by the preference test; (3) definition tasks followed by the preference test; (4) definition tasks preceded and followed by the preference test; (5) two consecutive concept identification tasks preceded and followed by the preference test; and (6) a control group receiving only a concept identification task. In addition, groups 1, 2 and 5 were asked to express confidence ratings of all cues on each trial of the concept identification problem. With the preference data of the posttest in groups 1 to 4 as dependent variable, no effects of pretesting, concept identification experience or their interaction were found. Moreover, the concept identification groups did not differ from each other in the number of errors to criterion. However, the preference data failed to predict the learning rate. This could not be attributed to lack of validity of the cue preference data. The predicted relations between cue weight and first trial confidence ratings, and between first trial confidence ratings and learning rate were not confirmed. Furthermore, learning rate was not predictable from the performances in a comparable paired associates learning experiment. However, some evidence for an individual differences interpretation was obtained.