Understanding Impression Updating Through Behavior Predictions

People largely preserve spontaneously drawn trait inferences from others' behaviors in their memory, even when contrary information is present. However, we suggest that updating could be manifested in future predictions about others. Experiment 1 replicated previous work, showing no updating in the memory of initially inferred traits upon learning trait-inconsistent new information. Experiments 2–4 tested whether people’s reliance on the initially implied traits in their future predictions changed as an indicator of updating. Participants aligned their predictions with the initially implied traits when the new information was neutral. However, after learning trait-inconsistent (vs. neutral) new information, they made weaker future predictions consistent with the initially implied traits and stronger predictions consistent with alternative (inconsistent) traits, revealing updating. The patterns were similar across continuous and categorical prediction measures (Experiment 3) and generalized to other valence-matching impressions (Experiment 4). The present work reveals nuances that extend our understanding of impression updating and suggests a utility in combining memory measures of impression formation with behavior predictions.