Relative Effects of Implicit and Explicit Attitudes on Behavior: Meta-Analysis and Test of Key Moderators

Based on dual process models of attitude, we investigated the unique effects of implicit and explicit forms of attitude on behavior and, critically, effects of conditions and contextual moderators expected to determine the relative strength of each form on behavior in a meta-analytic synthesis of studies. Specifically, we tested effects of salient methodological (e.g., implicit attitudemeasure characteristics, behavior measure characteristics, presentation order), conceptual (e.g., conceptualization of explicit attitude, degree of conscious control, social desirability concerns, target behavior), and sample-related (e.g., age) moderators on the relative effects of each attitude form on behavior. A systematic literature search identified 720 studies and 1108 unique samplesreporting useable effect size data on relations between attitude forms and behavior. A proposed model specifying unique effects of attitude forms on behavior was estimated using multi-level meta-analytic structural equation modeling. Moderator effects were tested by estimating the modelacross groups of studies at each level of the moderator. Consistent with prior research, we found unique averaged effects of both forms of attitude on prospectively-measured behavior controlling for past behavior, with a larger effect size for explicit attitudes. Both attitude forms partially mediated the past behavior-behavior effect. Moderator analyses revealed few variations in effectsof each form of attitude on behavior in most cases. Findings indicate generally consistent small unique effects of implicit attitudes on behavior across contexts and samples. However, high levels of residual heterogeneity were observed in the effect sizes after accounting for moderators signal the need for high-quality primary research that systematically tests moderator effects.