The extent to which instances are good or poor examples of their categories (typicality) was varied in a concept identification (CI) task. Typicality was first established for the kind of artificial material traditionally used in CI tasks (experiment 1). This material was employed in a CI task (experiment 2) with a variety of dependent variables, including the galvanic skin response (GSR). To test the generality of the results, more realistic stimuli were employed (experiments 3 and 4). The results showed that typicality influenced performance on the CI task, that the GSR is primarily related to uncertainty reduction, that the findings with arbitrary materials are replicated with more meaningful materials, and that the multi-hypothesis sampling theory of Levine is supported by the findings.