A series of single-subject studies of brain-damaged patients who each make semantic errors in naming is presented. A cognitive analysis of performance in each case provides evidence for proposing relatively selective impairment to one or more components of lexical processing. Studies of the effectiveness of treatment for these patients indicate that some patients with the same putative locus of impairment require different treatments, and some patients with different loci of impairment respond to the same treatment. Three forms of the “diagnosis7treatment relationship are illustrated: (1) two very different treatment approaches are equally appropriate with respect to a given locus of damage, but individual patients respond differentially to the two approaches; (2) a given treatment strategy is equally appropriate for multiple levels of disruption, although different components of the treatment may affect separate levels of processing; and (3) one strategy is appropriate for one level of damage, and a different strategy is required for a different level of damage. The first two cases are illustrated with multiple baseline (across behaviours) studies to measure the effects of intervention on treated and untreated naming tasks. Each therapy was sequentially applied to various tasks or to various sets of stimuli, for two patients using the same stimuli and procedures. We use the contrasting results of therapy for patients who presumably had the same loci of damage as a basis for addressing recent claims about the role of cognitive neuropsychological theories in developing and evaluating treatment strategies. Specifically, we examine models of lexical tasks with 450respect to predictions regarding the effects of particular interventions that follow from postulating a specific locus of disruption, and we consider the extent to which results of treatment for a given patient bear out predictions concerning whether or not the intervention will facilitate performance of other individuals with the same hypothesised impairment. We conclude by discussing the potential relevance of cognitive neuropsychological theories for the development and/or evaluation of a theory of rehabilitation—i.e. how cognitive processes are altered in response to specific interventions.