We report a method, based on isotope dilution--mass spectrometry, for determining cortisol in a pooled specimen of human serum. Isotopically labeled cortisol is added to 5.0 mL of serum so that the molar concentrations of labeled cortisol and unlabeled cortisol are approximately equal. The specimen and two calibration standards are extracted with dichloromethane, and the extracted cortisol is converted to the methoxime-trimethylsilyl ether derivative. Samples and standards are analyzed by gas chromatography--mass spectrometry by monitoring the peak areas for m/z 605 and 608. The cortisol concentration is calculated by linear interpolation between the two bracketing standards. Variances of data collected during six weeks showed that the overall coefficient of variation (CV) was 0.69% (n = 32); the within-vial CV, 0.63%; the among-vial CV, 0.22%; and the among-day CV, 0.15% (means = 3.973 nmol/vial). Method specificity was demonstrated by liquid chromatographic as well as C8 mini-column cleanup of samples before derivation, by alternative ion monitoring at m/z 636 and 639, and by negative-ion chemical ionization at m/z 459 and 462. Derivatives of all observed degradation products of cortisol under basic, neutral, and acidic conditions did not interfere.