The use of the prostate-specific antigen (PSA) test has been increasing rapidly in Canada since itsintroduction in 1988. The reasons for using the PSA test in patients without known prostate cancer are unclear. This paper reports on the first study in Canada to use physician records to assess the use of PSA testing. A questionnaire was mailed to physicians attending 475 patients without diagnosed prostate cancer. The patients were randomly selected from 2 laboratory databases of PSA test records in the Greater Toronto Area during 1995. The physicians were asked to consult their patient records to avoid recall bias. Information obtained included physician's specialty, patient's age at the time of the PSA test and reason(s) for the test. There were 264 responses (56%), of which 240 (91%) were usable. Of these 240, 63% indicated that the test was conducted to screen for prostate cancer, 40% said it was to investigate urinary symptoms, and 33% for follow-up to a medical procedure or drug therapy. More than one reason was permitted. Of 151 responses indicating screening as one reason for testing, 64% stated that it was initiated by the patient, and 73% stated that it was part of a routine examination. For 19%, both investigation of symptoms and screening asymptomatic patients were given as reasons for testing, and for another 19% both follow-up of a medical procedure and screening were provided as reasons. Screening was recorded as a reason for testing far more commonly for patients seen by family physicians and general practitioners than for patients seen by urologists (67% v. 29%, p < 0.001). In contrast, the use of PSA testing to diagnose urinary symptoms was more common for patients seen by urologists than for those seen by family physicians and general practitioners (52% v. 37%, p = 0.044). No significant difference was found between physician groups in the use of PSA testing as a follow-up of a medical procedure (42% for urologists and 31% for family physicians and general practitioners). About 24% of the PSA test records were for patients younger than 50 and older than 70 years. PSA testing initiated by patients was more common in the practices of family physicians and general practitioners than in the practices of urologists (44% v. 13%, p < 0.001). Screening for prostate cancer was the most common reason for PSA testing in our study group; it occurred most commonly in the family and general practice setting and was usually initiated by the patient. Differences in reasons for testing were identified by practice specialty. Although PSA screening for prostate cancer is sometimes recommended for men between 50 and 70 years of age, it is being conducted in men outside this age group.