Patient classification in neuropsychological research

In a number of papers we have been concerned with the type of inferences that are legitimate in "experiments of nature" where the experimenter does not and cannot control the modifications to the cognitive system that are introduced by brain damage. We have argued that in such cases very restrictive conditions must be met in order to be able to draw valid inferences about the structure of normal cognitive mechanisms. Two consequences of these conditions are (1) patient classification into syndrome types (e.g., phonological dysgraphia, agrammatism, and so forth) can play no useful role in research concerned with issues about the structure of normal cognitive functioning or its dissolution under conditions of brain damage; and (2) only single-patient studies allow valid inferences about the structure of cognitive mechanisms from the analysis of impaired performance. Zurif, Gardner, and Brownell (1989, Brain and Cognition, 10, 237-255) have taken exception to our conclusions and propose to show the limitations of our arguments. In this paper we respond to their criticisms.