Abstract
The subordinate level hypothesis states that prosopagnosics have impairments to mechanisms required for recognition of specific exemplars, and so it predicts that individuals with face recognition impairments should also be impaired with non-face recognition. Edward, a 51-year-old developmental prosopagnosic, is severely impaired with recognition of identity, emotions, and gender from the face. He was able to name only 3 of 25 famous faces despite reporting significant exposure to 19 of these celebrities and performed far out of the normal range on a test of unfamiliar face recognition. In order to test the subordinate level hypothesis, we tested his face and non-face discrimination using the same method. On two versions of the task that used faces, Edward's accuracy was severely impaired, and his response times were also far longer than the controls. His accuracy was normal with horses, cars, houses, and tools, borderline with guns, and impaired with landscapes. These results demonstrate that face discrimination can be impaired while discrimination with non-face categories is normal, and so indicate that the subordinate level hypothesis cannot account for some cases of prosopagnosia. Finally, because Edward's difficulties with faces appear to be the result of developmental problems, the results suggest that different developmental processes are responsible for building the separate mechanisms involved with face and non-face recognition.