Computational and developmental specificity in face recognition: Testing the alternative explanations in a developmental prosopagnosic

The goal of this research is to use experimental investigations of prosopagnosia—an impairment in the ability to recognize faces—to test hypotheses about the nature of the computational machinery underlying normal face recognition. First, is face recognition the product of computational machinery specialized for face processing, or the byproduct of general-purpose recognition procedures in the visual system? Second, if it is a computationally specialized ability in adults, was this competence constructed during development by general-purpose expertise-building machinery, or is it the product of developmental machinery specialized for building a face-recognition competence? If face recognition is the result of more general abilities, then the loss of the ability to recognize faces should be accompanied by the loss of a broader class of abilities. Experiments conducted on a developmental prosopagnosic strongly support the hypothesis that face recognition is a computational specialization, ruling out three domain-general hypotheses—the individuation hypothesis, the two process hypothesis, and the configural processing hypothesis—as explanations for his impairment. If face recognition is computationally specialized in adults, but was built by general-purpose expertise-building machinery, then its absence in a developmental prosopagnosic—someone who never developed the ability to recognize faces—should be accompanied by the inability to develop other comparable levels of expertise in object recognition. Experiments on the same developmental prosopagnosic indicate that he has developed unimpaired abilities to recognize every non-face object class tested, and performed comparably to control participants in tests involving the development of expertise with new classes of artificial stimuli. Although further tests are needed, the evidence so far favors the developmental specificity of face recognition better than the expertise hypothesis.