In this study we evaluated the physiological variation of free and total prostatespeci®c antigen (PSA) levels to determine how the percent freeatotal PSA was affected. Twenty four patients had blood drawn for ten consecutive weekdays. The percent coef®cient of variation (%CV) of biological variation was calculated. The results were log-normally distributed with geometric means of 12.0% CV, 7.3% CV, and 8.8% CV for free, total, and percent freeatotal PSA, respectively. When applied, the percent freeatotal, PSA would need to¯uctuate by 31% to indicate that a signi®cant change (critical difference, P`0.05) between two measurements had occurred. Biological variation of PSA measurements is substantial.