OBJECTIVE Comparison of the suitability of two minor urinary metabolites of benzene, trans,trans-muconic acid (tt-MA) and S-phenylmercapturic acid (S-PMA), as biomarkers for low levels of benzene exposure. METHODS The sensitivity of analytical methods of measuring tt-MA and S-PMA were improved and applied to 434 urine samples collected from 188 workers in 12 studies in different petrochemical industries and from 52 control workers with no occupational exposure to benzene. In nine studies airborne benzene concentrations were assessed by personal air monitoring. RESULTS Strong correlations were found between tt-MA and S-PMA concentrations in samples from the end of the shift and between either of these variables and airborne benzene concentrations. It was calculated that exposure to 1 ppm (8 hour time weighted average (TWA)) benzene leads to an average concentration of 1.7 mg tt-MA and 47 micrograms S-PMA/g creatinine in samples from the end of the shift. It was estimated that, on average, 3.9% (range 1.9%-7.3%) of an inhaled dose of benzene was excreted as tt-MA with an apparent elimination half life of 5.0 (SD 2.3) hours and 0.11% (range 0.05%-0.26%) as S-PMA with a half life of 9.1 (SD 3.7) hours. The mean urinary S-PMA in 14 moderate smokers and 38 non-smokers was 3.61 and 1.99 micrograms/g creatinine, respectively and the mean urinary tt-MA was 0.058 and 0.037 mg/g creatinine, respectively. S-PMA proved to be more specific and more sensitive (P = 0.030, Fisher's exact test) than tt-MA. S-PMA, but not tt-MA, was always detectable in the urine of smokers who were not occupationally exposed. S-PMA was also detectable in 20 of the 38 non-smokers from the control group whereas tt-MA was detectable in only nine of these samples. The inferior specificity of tt-MA is due to relatively high background values (up to 0.71 mg/g creatinine in this study) that may be found in non-occupationally exposed people. CONCLUSIONS Although both tt-MA and S-PMA are sensitive biomarkers, only S-PMA allows reliable determination of benzene exposures down to 0.3 ppm (8 h TWA) due to its superior specificity. Because it has a longer elimination half life S-PMA is also a more reliable biomarker than tt-MA for benzene exposures during 12 hour shifts. For biological monitoring of exposure to benzene concentrations higher than 1 ppm (8 h TWA) tt-MA is also suitable and may even be preferred due to its greater ease of measurement.