Abstract
How do observers distinguish previously viewed faces from new faces containing old parts? Meltzer et al. (see VSS, 2016; VSS, 2017) proposed that this type of configural recognition is based on familiarity differences – old faces evoking stronger familiarity than new faces with old parts – in conditions supportive of holistic or "unitized" encoding. By contrast, configural recognition is based on conjunctive representations supporting an analytical process of match-mismatch detection in conditions disruptive to holistic processing. In tests of this hypothesis, participants studied lists of well-formed, upright faces or faces that were inverted and/or misaligned to disrupt holistic processing. In the subsequent test, participants viewed a sequence of faces, responding to each with an old-new judgment for (a) the upper half of the face, (b) the lower half of the face, and (c) the face as a whole. The accuracy of whole-face judgments was controlled by adjusting study-presentation frequency to match configural recognition across stimulus conditions. With such matching in place, the key finding was that high-confidence recognition of facial parts was consistently more accurate with inverted and misaligned faces than with well-formed upright faces. One interpretation of our findings is that holistic processing impairs the formation of conjunctive representations that support match-mismatch detection. An alternative view is that the formation of conjunctive representations occurs independently of holistic processing as a function of study time. To test these ideas, we employed the paradigm described above with upright and inverted faces, each studied two or eight times. The results suggest that conjunctive representations supporting match-mismatch detection are formed at least as efficiently with upright faces (allowing for holistic processing) as with inverted faces (disrupting such processing) with repeated stimulus exposures. These findings have important implications for the role of holistic processing in recognition memory for newly learned faces.
Meeting abstract presented at VSS 2018