Memory for faces: encoding and retrieval operations.

The reliable finding that trait judgments of faces yield better recognition memory than do feature judgments of faces is conceptualized as an encoding-specificity effect. Specifically, both trait-judgment encodings of faces and face-recognition tests are argued to be holistic, involving topographical information with between-feature processing. Consistent with the concept that encoding and retrieval operations interact to produce retrieval success, it was expected that a memory-for-face test using the Identi-kit (which requires reconstructions of the face at a feature level of analysis) would show trait-encoding tasks to be inferior to feature-encoding tasks. Eighty subjects were assigned randomly to judge a face on 10 trait dimensions (e.g., honesty-dishonesty) or on 10 feature dimensions (e.g., narrow nose-wide nose) and subsequently attempted to recognize the target among five distractors or to reconstruct the face from an Identi-kit. The significant interaction between encoding and retrieval operations indicated that the face was best identified under trait-encoding conditions but best reconstructed under featureencoding conditions. The match between trait encoding and recognition in yielding high memory performance suggests strongly that trait judgments foster holistic processing of faces (i.e., interfeature topographical information is part of the context) and that the recognition of faces also is holistic. Finally, the utility of the featurevs, holistic-processing distinction is questioned, and an alternative is proposed.