Helmut Leder and Claus-Christian Carbon

Some theories of holistic face processing propose that parts in faces (eyes, nose, mouth, etc.) are not explicitly represented. So far, the empirical evidence has shown that whole-to-part superi-ority is found when wholes are learned. We substantiated this using photographic faces. More importantly, we investigated whether learning parts also reveals holistic effects. This has not been attempted before. Four experiments showed that after learning facial parts, recognition of these parts was disrupted when the part was shown in the full face. This distraction effect was strongest when perceivers were not directed to focus on a particular facial feature. Thus, it is very difficult to ignore irrelevant parts in faces. In fact, this might be the essence of holistic face processing. In face recognition, holistic processing has been defined as a representation in which parts are not explicitly represented (Tanaka & Farah, 1993). This holistic processing proposal has been supported by findings that recognition of facial parts from facial context (full faces) is superior to recognition of facial parts presented alone. Tanaka and Farah had participants learn schematic faces composed of shared features. In a recognition test, participants had to choose which face corresponded to a learned face from either two facial parts (e.g., two dif