OBJECTIVE The aim of this article is to critically summarize the most promising attempts to split obsessive-compulsive disorder into subgroups based on clinical characteristics (i.e. age of onset, presence of comorbid tics, positive family history) and symptom theme, with particular emphasis on the latter. RESULTS Attempts to split obsessive-compulsive disorder into mutually exclusive sub-groups based on clinical characteristics have been useful but not exempt of problems. The complex clinical presentation of the condition can be reduced to a few consistent, temporally stable symptom dimensions that can coexist in any given individual. Researchers have begun to investigate the genetics and neural mechanisms of these symptom dimensions and to develop specific assessment and treatment protocols for each particular problem. CONCLUSIONS The multidimensional model of obsessive-compulsive disorder proposes a middle ground between the 'lumping' and 'splitting' perspectives. The disorder can be better understood as a spectrum of multiple potentially overlapping syndromes. The most fruitful research strategy will be to examine the common and specific etiological factors implicated in each symptom dimension.