Abstract
Faces are complex visual objects that can be distinguished from other objects that occur in our visual environment using first-order information. The term “first-order information” refers to the basic spatial layout of features that is repeated in all faces (two eyes, above a nose, above a mouth). An outstanding question is how the detection of the first-order, or canonical, configuration interacts with the processes that underlie exemplar discrimination. Here, we begin to address this question by examining how the first-order configuration of a face contributes to an exemplar-based discrimination task in two nonhuman primate species. Twelve subjects (six chimpanzees Pan troglodytes and six rhesus monkeys Macaca mulatta) were trained to discriminate scrambled faces. Subjects were then able to generalize from the learned configuration to both the canonical configuration and a novel configuration. In an alternative condition, the subjects were trained with whole faces (in the canonical configuration) and tested with scrambled faces. A comparison between these two conditions demonstrates that the presence of the canonical configuration changed the perception of local features. These results are, thus, consistent with the concept of holistic processing for whole faces. We also present new data showing that both species tended to match the configuration of the face over-and-above second-order information, the unique variation among faces assumed to be the basis for exemplar discrimination. These data make a valuable contribution by clarifying the definition of terms that have become a source of confusion in face recognition research. We propose that the first-order configuration of a face serves a crucial social function by enabling the involuntary integration of features during the early stages of face perception.