Recently, obsessive-compulsive disorder (OCD)is often conceptualized as a heterogeneous disorder. To verify and more fully understand this OCD heterogeneity, more homogeneous and potentially valid phenotypic methods are needed. If OCD subtypes can be characterized using distinct features of psychobiology, then this would account for the variance in clinical and neurobiological studies on biological markers, and would potentially impact on treatment strategies for each patient. Most of the work on OCD subtypes has taken a categorical approach. For instance, it has been suggested that some forms of OCD are etiologically related to tic disorders, and OCD with comorbid tics has been characterized by specific phenomenological, genetic, and neuro-imaging features and a differential treatment response. There have been other attempts to subdivide OCD categorically based on predominant compulsions (e.g., cleaning or checking), age at onset (e.g., early-or late-onset), gender, impulsive features, comorbidity, or insight. The dimensional approach regards OCD as being composed of sets of obsessive-compulsive symptom dimensions. In recent studies, factor analyses have provided consistent evidence that distinct obsessive-compulsive symptom dimensions exist, including obsessions/ checking, contamination/washing, symmetry/ordering, and hoarding. It has been hypothesized that each symptom dimension may be underpinned by a distinctive set of bio-behavioral mechanisms. Indeed, neuroimaging studies have suggested particular neural correlates for different symptom dimensions, and some genetic and family studies are also consistent with such a hypothesis. Further, symptom dimensions may predict treatment responses ; for example, a higher hoarding dimension level has consistently been associated with a poorer treatment response to selective serotonin re-uptake inhibitors and cognitive-behavioral therapy. However, the dimensional structure of OCD symptoms is still not definitive, and the further standardization of methodological and analytic processes is also required. Thus, the relevance and limitations of each approach still remain to be discussed, and there is probably not one but several heuristic strategies that can be employed to identify more homogeneous OCD subtypes, which, when combined, may be, the most reasonable and useful method.