A globulin fraction rich in bound testosterone was separated from serum by ammonium sulfate precipitation. Testosterone and related steroids were quantitatively extracted from this fraction with a benzene petroleum ether mixture and measured by radioimmunoassay using testosterone-3-tyrosine methyl ester-125I and a specific antiserum elicited against testosterone-3-oxime bovine serum albumin. This assay procedure, which avoids chromatography and uses only one organic solvent extraction had blank values which approached zero. The within and between assay error, measured as the coefficient of variation was 7.8% (n = 64) and 8.2% (n = 15), respectively. The minimum amount of hormone per sample which can be measured with good reliability is 0.17 ng. This provides sufficient hormone to account for losses during the procedure, and to allow quadruplicate samples for radioimmunoassay and an aliquot for determining recovery. Although introduction of an additional thin layer chromatographic step was associated with results which were approximately 10% lower than those obtained without chromatography, these differences were not statistically significant. An excellent agreement was found between the amount of unlabeled hormone added to male, female, or pregnancy serum and the amount measured by the present technique. The mean testosterone concentration (±SD) in nine normal men, six normal cycling women, and four pregnant women (third trimester) were 5.96 ± 2.02; 0.46 ± 0.21 and 0.97 ± 0.24 ng/ml respectively. The method is simple and practicable; one technician can easily complete 300 assays in five working days.