Abstract
Much evidence supports holistic processing of faces, but the role of such processing in long-term memory for faces remains unclear. A prior study (Meltzer et al., VSS, 2015) reported evidence that holistic processing supports familiarity-based memory for relations among facial parts or regions. The authors found that with upright faces, thought to be processed holistically, faces containing parts in a previously seen configuration (intact faces) were more familiar to participants than faces containing parts in a novel configuration (conjunction faces). By contrast, they found that with faces that were both inverted and misaligned, participants discriminated intact items from conjunction items partly by recollecting the configurations in which they had previously seen individual face parts. Here, we sought to understand how holistic processing leads to familiarity-based associative recognition for faces. Specifically, we tested the possibility that holistic processing impairs independent recognition of face parts, which could prevent participants from recollecting the context in which they previously saw those parts. Participants studied a list of faces and then took a recognition test containing intact, conjunction, and new faces. For each test face, they judged whether they had seen its top part before, whether they had seen its bottom half before, and whether they had seen the whole face before. Consistent with the holistic-familiarity hypothesis, an ROC analysis suggested that associative recognition judgments made by participants in the upright condition were more familiarity-based than those made by participants in the inverted, misaligned condition. By contrast, participants in the inverted-misaligned condition were better at recognizing the halves of conjunctions they had rejected as wholes than participants in the upright condition. These results, along with those of a follow-up investigation, suggest that holistic processing causes new faces composed of old parts to be perceived as globally unfamiliar by impairing independent recognition of face parts.
Meeting abstract presented at VSS 2016