Objectives-To investigate the specificity of biological monitoring variables (excretion of phenanthrene and pyrene metabolites in urine) and the usefulness of some biomarkers of effect (alkaline filter elution, 32p postlabelling assay, measurement of sister chromatid exchange) in workers exposed to polycyclic aromatic hydrocarbons (PAHs). Methods-29 coke oven workers and a standardised control group were investigated for frequencies of DNA single strand breakage, DNA protein cross links (alkaline filter elution assay), sister chromatid exchange, and DNA adducts (32p postlabelling assay) in lymphocytes. Phenanthrene and pyrene metabolites were measured in 24 hour urine samples. 19 different PAHs (including benzo-(a)pyrene, pyrene, and phenanthrene) were measured at the workplace by personal air monitoring. The GSTT1 activity in erythrocytes and lymphocyte subpopulations in blood was also measured. Results-Concentrations of phenanthrene, pyrene, and benzo(a)pyrene in air correlated well with the concentration of total PAHs in air; they could be used for comparisons of different workplaces if the emission compositions were known. The measurement of phenanthrene metabolites in urine proved to be a better biological monitoring variable than the measurement of 1-hydroxypyrene. Significantly more DNA strand breaks in lymphocytes of coke oven workers were found (alkaline filter elution assay); the DNA adduct rate was not significantly increased in workers, but correlated with exposure to PAHs in a semiquantitative manner. The number of sister chromatid exchanges was lower in coke oven workers but this was not significant; thus counting sister chromatid exchanges was not a good variable for biomonitoring of coke oven workers. Also, indications for immunotoxic influences (changes in lymphocyte subpopulations) were found. Conclusions-The measurement of phenanthrene metabolites in urine seems to be a better biological monitoring variable for exposure to PAHs than measurement of hydroxypyrene. The alkaline filter elution assay proved to be the most sensitive biomarker for genotoxic damage, whereas the postlabelling assay was the only one with some specificity for DNA alterations caused by known compounds.