OBJECTIVE: This study examined the 1-year temporal stability of a National Institute of Mental Health Diagnostic Interview Schedule (DIS) lifetime diagnosis of obsessive-compulsive disorder in the Epidemiologic Catchment Area (ECA) study. METHOD: In that study, 20,862 individuals, aged 18 years and over, at five sites were evaluated by lay interviewers using the DIS (wave 1). All of those who were available 12 months later were reinterviewed (wave 2). In the present study, the temporal stability of wave 1 obsessive-compulsive disorder diagnoses at wave 2 was examined, as well as relationships with comorbid diagnoses. The consistency of reports of "new-onset" illness was also examined. Factors contributing to these measures were evaluated. RESULTS: The temporal stability of the diagnosis of obsessive-compulsive disorder was very low. Subjects with a stable diagnosis of obsessive-compulsive disorder had a higher rate of both obsessions and compulsions, an earlier age at onset, and more comorbid anxiety, affective, and alcohol abuse/dependence disorders at initial assessment. The originally reported 1-year incidence estimates for obsessive-compulsive disorder primarily reflect data from subjects at wave 2 who reported the onset of symptoms as preceding the wave 1 interview. Older and less-educated subjects had significantly higher error rates in reporting onset. CONCLUSIONS: The DIS diagnosis of obsessive-compulsive disorder has poor validity, leaving the true incidence and prevalence of the disorder unknown. Older and less- educated subjects require special attention in the design of instruments for use with community samples.