Pseudocontingencies in stereotype formation : extending illusory correlations

Under the notion of illusory correlations, simple learning paradigms (e.g. Hamilton & Gifford, 1976) have been used to study the formation of stereotypes that discriminate between majorities and minorities. In the present paper, limitations of this approach in terms of theoretical explanations and empirical evidence are addressed. Theoretically, we propose pseudocontingencies (PCs, Fiedler, Freytag & Meiser, 2008) as a more robust mechanism behind illusory correlations. In contrast to previous explanations, PCs can explain illusory correlations when groups are never paired with valence. Empirically, we replicate earlier findings, i.e. that the more frequently observed group, the majority, is evaluated more in line with the more frequently observed valence. Crucially, we extend the empirical evidence in that illusory correlations prove robust over a very large number of observations (320) and under increasingly interactive task conditions, involving predictions of valence (Experiment 2) and reinforcement-learning conditions (Experiment 3). The latter provided evidence for illusory correlations on a new measure, participants’ predictions. These predictions reflect the expectations about the valence associated with majority and minority and might well affect real life behavior. The discussion focuses on possible reasons for why PCs are used in stereotypic judgments.