Individual differences in association value scales and paired-associate learning

Abstract

An attempt was made to investigate the theoretical and methodological problems involved in the study of individual differences in verbal learning. Particular emphasis was placed upon association value scales and paired-associate learning. An association value scale of the test items used was obtained for each subject collecting data on their behaviour in a number of similar learning tasks. The scales so obtained were compared with association value scales which were obtained by subjects' rating the test items on a five-point scale. The comparisons between individual and group results indicated that a group association value scale could not sufficiently control the item homogeneity for individual subjects. However, the rated association value scales obtained for individual subjects were highly correlated with their own association value scales obtained in actual learning situations, indicating that subjects' own ratings of the items can be used to yield more accurate predictions of their performance in verbal learning tasks. In studying individual differences in paired-associate learning the emphasis was placed upon various mathematical models with different theoretical assumptions. Thus, individual differences variables were investigated in terms of parameter values of paired-associate learning models. For each subject repeated measurements were taken by running them on a number of parallel tasks which were designed according to their own association value scales. A set of parameter values was estimated for each subject separately. Changes in the values of the parameters representing various theoretical constructs of the models used were investigated with regard to changes in subjects' overall performance. Results indicated that when fitting models to data from individual subjects, one particular model may best fit the behaviour of one group of subjects whereas the behaviour of other subjects may be better represented by other models.