I compared results by three different routine methods for analysis for urinary cortisol with those by a highly specific reference method based on isotope dilution--mass spectrometry (i). A "high-performance" liquid-chromatographic method (II) gave the most comparable results (regression coefficient 0.86, intercept 9 nmol/L). For some urines much lower values were obtained by I than by II. Two radioimmunoassay (III) methods, one involving direct assay and one involving extraction, gave less-accurate results (regression coefficients of 1.87 and 1.52 and intercepts of 86 and 12 nmol/L, respectively, although values obtained by III and by I correlated well (r = 0.95-0.99), indicating a relation between the free cortisol and the compounds interfering in III. The apparent accuracy for the extraction method was previously assayed by I (regression coefficient 0.90, intercept 6 nmol/L). All four methods investigated showed a statistically significant sex-related difference in 24-h urinary cortisol excretion; evidently such a finding should be a prerequisite in any such method proposed for routine use.