This retrospective study analyzes variation in prostate-specific antigen (PSA) levels in 129 males who were not diagnosed with prostate cancer or other known malignancies. The extent to which the assay and the biologic variation contributed to the variation in PSA concentration was evaluated from analysis of slopes characterizing PSA concentration as a function of time. The mean coefficient of variation on observations was 58.0 percent. The estimated mean biologic coefficient of variation was 55.3 percent versus a mean assay coefficient of variation of 13.2 percent, indicating that the assay variation contributed negligibly to variation compared with the biologic variation. The concept that a PSA level which rises more than that attributable to assay variation indicates the need for invasive testing for prostate cancer is questionable. A decreasing PSA level was seen as often and of the same magnitude as an increasing level within a relatively narrow window of approximately one year. We are aware of no reason why this variability in PSA values would not be observed in patients with occult prostate cancer.