A candidate reference measurement procedure for total testosterone in human serum involving isotope dilution (ID) coupled with liquid chromatography/tandem mass spectrometry (LC/MS/MS) has been developed and critically evaluated. The endogenous testosterone and its internal standard (testosterone-d (3)) were extracted from the serum matrix using a combination of solid-phase extraction and liquid-liquid extraction prior to reversed-phase LC/MS/MS. Accuracy of the measurements was evaluated by a recovery study using testosterone-spiked serum. The recovery of the added testosterone ranged from 100.0 to 100.3%. This method was applied to the determination of testosterone in frozen serum samples from three individual donors (one female and two males) with the testosterone concentrations ranging from 0.3 to 8.5 ng g(-1). Repeatability with within-set coefficients of variation (CVs) from 0.1 to 1.0% and intermediate precision with between-set CVs from 0.1 to 0.5% for both female and male serum materials were demonstrated. Excellent linearity was obtained for all linear regression lines. The detection limit at a signal-to-noise ratio of approximately 3 was 2 pg of testosterone in serum. Structural analogs as well as testosterone metabolites were tested and found to not interfere with the measurement of testosterone. This well-characterized LC/MS/MS method for serum testosterone, which demonstrates good accuracy and precision, and low susceptibility to interferences, qualifies as a reference measurement procedure that can be used to provide an accuracy base to which routine methods for testosterone can be compared and that will serve as a standard of higher order for measurement traceability.