Categorical perception of facial expressions: evidence for a "category adjustment" model

Four experiments probed the nature of categorical perception (CP) for facial expressions. A model based on naming alone failed to accurately predict performance onthese tasks. The data are instead consistentwith an extension of the category adjustment model (Huttenlocher et al., 2000), in which the generation of a verbal code (e.g., "happy") activated knowledge ofthe expression category's range andcentral tendency (prototype) in memory, which was retained as veridical perceptual memory faded.Further support for amemory bias toward the category center came from a consistently asymmetric pattern of within-category errors. Verbal interference in the retention interval selectively removed CP for facial expressions, under blocked, but not under randomized presentation conditions. However, verbal interference at encoding removed CPeven under randomized conditions and these effects were shown to extend even to caricatured expressions, which lie outside the normal range of expression categories.