Generalization Versus Contextualization in Automatic Evaluation Revisited: A Meta-Analysis of Successful and Failed Replications

To account for disparate findings in the literature on automatic evaluation, Gawronski, Rydell, Vervliet, and De Houwer (2010) proposed a representational theory that specifies the contextual conditions under which automatic evaluations reflect initially acquired attitudinal information or subsequently acquired counterattitudinal information. The theory predicts that automatic evaluations should reflect the valence of expectancy-violating counterattitudinal information only in the context in which this information had been learned. In contrast, automatic evaluations should reflect the valence of initial attitudinal information in any other context, be it the context in which the initial attitudinal information had been acquired (ABA renewal) or a novel context in which the target object had not been encountered before (ABC renewal). The current article presents a meta-analysis of all published and unpublished studies from the authors' research groups regardless of whether they produced the predicted pattern of results. Results revealed average effect sizes of d = 0.249 for ABA renewal (30 studies, N = 3,142) and d = 0.174 for ABC renewal (27 studies, N = 2,930), both of which were significantly different from zero. Effect sizes were moderated by attention to context during learning, order of positive and negative information, context-valence contingencies during learning, and sample country. Although some of the obtained moderator effects are consistent with the representational theory, others require theoretical refinements and future research to gain deeper insights into the mechanisms underlying contextual renewal.