As a neuropsychologist, Tim Shallice considers the general question of what can be learned about the operation of the normal cognitive system--including perception, memory, and language--from the study of the cognitive difficulties arising from neurological damage and disease. He distinguishes two type of theories of normal function--primarily modular and primarily non-modular--and argues that the problems of making valid inferences about normal function from studies of brain-damaged subjects are more severe in the latter. He first analyzes five areas in which modularity can be assumed. He then examines these inferences, from group studies, from individual case studies, and from group studies, and from non-modular systems, more theoretically. Finally he considers five areas where theories of normal function are relatively undeveloped and neuropsychologists provide counter-intuitive phenomena and guides to theory-building.