This study tested the hypothesis that the effect of in-group-serving expectancy biases on memory and judgement is moderated by processing goal. In-group-serving biases in both recall and judgements (estimated frequencies of desirable and undesirable in-group and out-group behaviour, evaluative ratings) were hypothesized to be stronger when processing goal interfered with online impression formation compared with when attention is focused on the formation of group impressions. This hypothesis was tested in an illusory correlation experiment in which in-group-serving expectancy bias and infrequency bias (Hamilton & Gifford, 1976) could be studied simultaneously. Participants read statements describing desirable and undesirable behaviour of members of the in-group and out-group, while the in-group held either minority or majority status, while participants were instructed either to form group impressions or to memorize the information. As predicted, memorizing compared with impression formation resulted in in-group-serving illusory correlations in recall and frequency estimates and in in-group bias in group evaluations. Infrequency of occurrence affected recall and judgments irrespective of processing goal. The implications of these results for social stereotyping are discussed.