Person memory : associative networks, categories, and schemas

Abstract

A model of impression memory was utilized which partitioned impression information into category-consistent (schema) and category-inconsistent information (peripheral). The memory for consistent and inconsistent information was investigated in a series of experiments. Two information-processing models, the Schema-Pointer-Plus-Tag (SP+T) and the Associative-Network-Plus-Elaborative-Processing (AN+EP) models were tested for their ability to account for impression memory data. Experiment 1 tested subjects' recognition memory for an impression. Subjects' hit rates for schema and peripheral impression traits were equal both at immediate and delayed test intervals. False alarm rates were greater for schema nonimpression distractors than for peripheral dis tractors at each retention interval. Reaction times paralleled recognition performance. Times to make schema and peripheral hits were equal, while correct rejections of schema nonimpression distractors took longer than peripheral distractors. Experiment 2 tested subjects' recall memory for an impression. Subjects recalled equal proportions of schema and peripheral impression traits at immediate test, but recalled a greater proportion of schema traits after a delay. Recall intrusions at immediate test were equal for schema and peripheral nonimpression traits, but there were more schema than peripheral intrusions at the delayed test. Experiment 3 also tested subjects recall memory for an impression, though in this experiment subjects were required to furnish impression traits themselves, rather than selecting them from a checklist as in Experiments 1 and 2. The pattern of results was similar to that of Experiment 2, except that there was a greater proportion of schema impression traits recalled at both retention intervals, rather than only at the delayed test. Experiment 4 found that recalled impression traits were no more or less associated or linked with other impression traits than nonrecalled impression traits. Additionally, recalled traits were no more likely to be linked with other recalled traits than nonrecalled traits, and no more likely to be the source of links to other impression traits. Experiment 5 found a similar pattern of results to Experiment 4 with respect to not only those links within the impression, but also between impression traits and other traits in the subjects' vocabularies. Recalled impression traits did not differ from nonrecalled traits on any of the measures of association or interlinking. Experiment 6 found a significant relationship between traits categorized together as measured in Experiments 1-3, and traits seen as associated when judged on a pairwise basis as measured in Experiments 4-5. The better recall of schema traits as found in Experiments 2 and 3, which contrasts with the failure of Experiments 4 and 5 to show that recalled traits were simply more interassociated than nonrecalled traits, cannot be attributed to measurement factors. Both the SP+T and the AN+EP models were found to be inappropriate for modelling impression memory data, as was a simple associative model. Memory for impression information was shown to be schematically driven, though not to the extent suggested by the SP+ T model.