A leading account of high-level visual recognition proposes that the recognition of faces, objects, and words is mediated by two processing capacities. Words are assumed to require the capacity to represent numerous parts, whereas faces are processed wholistically, and hence require the representation of complex units. Object recognition requires the capacity to represent both numerous and complex parts. As object recognition depends upon the same processing capacities underlying face and word recognition, this account predicts that patients with severe alexia and prosopagnosia should be deficient in tests of object recognition. We report a patient who is unable to recognize words or faces, yet performs relatively well on tests of object recognition. The two-capacity theory cannot accommodate this pattern of performance without additional assumptions.