The role of category variability in belief revision

Eight experiments examined the role of category variability in revision of beliefs about category central tendency following exposure to disconfirming information. Experiments 1-3 focused on the effects of variability in categories drawn from stereotyped or stereotype-neutral domains. Participants learned the characteristics of various groups via exposure to numerical distributions of category attributes. These distributions had the same mean values, but either high or low variability around the mean. Participants were then exposed to discrepant test exemplars and asked to re-estimate the central tendency of the category and rate the likelihood that the category had changed. In both category domains beliefs about categories with low variability were more likely to be revised than beliefs about highly variable categories. Experiments 4-6 examined the extent to which other factors including causal beliefs and temporal framing might influence the effects of category variability. A key finding was that variability only affected belief revision when training and discrepant test exemplars came from samples taken at different times. Experiments 7-8 manipulated perceived rather than actual category variability. It was found that when training exemplars were presented in an ordered fashion such that the values of successive exemplars could be predicted from previous observations, discrepant test exemplars were discounted and little belief revision took place. Most of these results are consistent with a Bayesian model of belief revision. Extensions of existing Bayesian models are discussed, along with more general implications for existing models of categorization