While controlled trials with SRIs have demonstrated a selective efficacy in obsessive-compulsive disorder (OCD), up to 40-60% of patients do not have a satisfactory outcome. Non-response to treatment in OCD is associated with serious social disability. There are a large number of non-responsive patients, and they are difficult to cluster due to ambiguities in the diagnostic criteria, possibility of subtypes, and a high rate of comorbidity. Moreover, the findings of current studies of so-called 'non-responsive' cases are currently non-generalizable because of the lack of an operational definition of non-response. The result has been that a cumulative body of data on a reasonably homogeneous sample of non-responders has not been developed. The aims of this paper are to clarify some of the obstacles in defining stages of response and levels of non-response and, through a comprehensive analysis, to propose a systematic nosology for this rather common condition. Better characterization of which patients respond and do not respond to various treatments will enable more accurate clustering of patients, and help facilitate multi-site data collection for future research trials.