Five hundred and nine workers at a manganese (Mn) smelting works comprising eight production facilities and 67 external controls were studied cross-sectionally. Exposure measures from personal sampling included inhalable dust, cumulative exposure indices (CEI) and average intensity (INT = CEI/years exposed) calculated for the current job at the smelter and also across all jobs held by subjects. Biological exposure was measured by Mn in the blood (MnB) and urine (MnU) and biological effect was measured by serum prolactin. Average lifetime exposure intensity across all jobs ranged from near 0 (0.06 microg/m3) for unexposed external referents to 5 mg/m3. Atmospheric exposures and MnB and MnU distributions were consistent with published data for both unexposed and smelter workers. Associations between biological exposures and groups defined by atmospheric exposures in the current job were substantial for MnB, less so for MnU and absent for serum prolactin. Random sampling of MnB measurements representative of a group of workers with more than 1-2 years of service in the same job and notionally homogenous exposure conditions could serve as a cross-sectional predictor of atmospheric Mn exposure in the current job, as well as for surveillance of Mn exposure trends over time. Correlations at the individual level were only modest for MnB (33% of the variance in log atmospheric Mn intensity in the current job was explained by log MnB), much worse for MnU (only 7%). However, a receiver operating characteristic (ROC) analysis was performed which showed that it is possible to use a MnB cut-off of 10 microg/l (the 95th percentile in the unexposed) to good effect as a screening tool to discriminate between individual exposures exceeding and falling below a relatively strict atmospheric Mn exposure threshold at the ACGIH threshold limit value (TLV) of 0.2 mg/m3. MnU has no utility as a measure of biological exposure nor does serum prolactin as a measure of biological effect.