The recent epidemiologic studies report extremely varied rates for social phobia (SP). One of the reasons for this may be the difficulty in diagnosing SP, the boundaries of which are uncertain. A community survey was carried out using doctors with experience in clinical psychiatry as interviewers, and a clinical diagnostic instrument. Two thousand three hundred and fifty-five people (out of the 2,500 randomly selected from the population) living in Sesto Fiorentino, a suburb of Florence, Italy, were interviewed by their own general practitioner, using the MINI plus six additional questions. Six hundred and ten of the 623 subjects that were found positive for any form of psychopathology at the screening interview, and 57 negative subjects, were re-interviewed by residents in psychiatry using the Florence Psychiatric Interview (FPI). The FPI is a validated composite instrument that has the format of a structured clinical research record. It was found that 6.58% of subjects showed social anxiety not attributable to other psychiatric or medical conditions during their life. Social or occupational impairments meeting DSM-IV diagnostic requirements for SP was detected in 76 subjects (lifetime prevalence = 3.27%). Correction for age raises the lifetime expected prevalence to 4%. Sex ratio was approximately (F:M) 2:1. The most common fear was speaking in public (89.4%), followed by entering a room occupied by others (63.1%) and meeting with strangers (47.3%). Eighty-six point nine percent of subjects with SP complained of more than one fear. The mean age of onset (when the subjects first fully met DSM-IV criteria for SP) was 28.8 years, but the first symptoms of SP usually occurred much earlier, with a mean age of onset at 15.5 years. Ninety-two percent of cases with SP also showed at least one other co-morbid psychiatric disorder during their life. Lifetime prevalence of avoidant personality disorder (APD) was 3.6%. Forty-two point nine percent of cases with SP also had APD, whereas 37.9% of cases with APD developed SP.