Super face-inversion effects for isolated internal or external features, and for fractured faces.

Two experiments were conducted to determine the contributions of the face and object systems to the recognition of upright and inverted faces. In Experiment 1, CK, a person with object agnosia and normal recognition of upright faces, and 12 controls attempted to identify faces when presented with upright or inverted versions of the whole face, or with only their internal or external features. CK recognised as many upright whole faces as controls and the performance of both dropped slightly in the upright, internal feature condition. CK's recognition, however, was impaired in the upright, external condition, and severely impaired in the inverted whole condition, whereas control performance was equivalent in the two, and only somewhat worse than in the upright whole condition. Recognition in the inverted internal and external condition was extremely poor for all participants, leading to a super-inversion effect. This super-inversion effect suggested that recognition depends on more than just piecemeal identification of individual features. Experiment 2 was conducted to determine whether relational information is needed even for the identification of inverted faces. Twelve controls were required to identify whole and fractured faces in the upright and inverted orientation. The fractured faces had all the parts in the canonical order (eyes above nose above mouth) but they were separated by gaps, thereby altering the spatial relation among them. Recognition of inverted fractured faces was much worse than recognition of upright fractured faces and inverted whole faces, producing yet another super-inversion effect. The deficit in the inverted fractured condition was equal to the combined drop in performance in the other two conditions, indicating that the effects of inversion and fracturing are additive. On the basis of these results, we proposed that the face system forms holistic representations of faces based on orientation-specific global configurations primarily of internal features. When this information is unavailable, as when viewing inverted or fractured faces, the object system is needed to integrate information about individual features, which themselves may be orientation-specific, with information about the local or categorical relations among them into an object-system counterpart of the face-system representation. The creation of the facial counterpart by the object system and the consequent identification by the face system involves an exchange of information between the two systems according to an interactive activation model.