Patients meeting the social phobia criteria of the revised third edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-III-R) on the DSM-III-R Structured Clinical Interview (n = 101) entered a long-term moclobemide treatment study. These patients were treated for 2 years with moclobemide (phase I) followed by drug withdrawal, in most cases abruptly (phase II). Those who relapsed entered phase III for a further period of 2 years of treatment. During phase I 40 patients (39.6%) withdrew due to inefficacy or relapse. Two patients were removed from the study because of other diagnoses (borderline or schizophreniform). At the end of phase I the remaining patients (58.4%) were rated as not ill (45.5%) or minimally ill (11.9%). Effort was taken to achieve the maximum dose of moclobemide (750 mg/day) and the mean (+/-SD) dose was 723.3 +/- 67.7 mg/day (month 21). A marked decrease in symptoms in the patients who responded was recorded on the Liebowitz Scale for Social Phobia, Clinical Global Impressions. Hamilton Anxiety Scale and Hamilton Depression Scale. Non-response was mainly associated with co-morbidity, especially alcohol abuse, axis II disorders, and a history of major depression or secondary dysthymia. The drug was well tolerated; the more frequent side effects were mild and occurred mainly in the first 2 months of phase I, including nausea, headaches or insomnia. In phase II there was a relapse rate of 88% and 51 patients entered phase III; these patients are still being treated.