Eight polycyclic aromatic compounds (PACs) were tested for their toxic effect on the soil nitrification process, bacterial genetic diversity, and the total number of protozoans (naked amoebae and heterotrophic flagellates). After four weeks of exposure in a well-characterized agricultural soil, toxic effects were evaluated by comparison to uncontaminated control soils. All PACs affected the nitrification process, and the calculated no-observed-effect concentrations (NOECs) for nitrification were 79 mg/kg for pyrene, 24 mg/kg for fluoranthene, 26 mg/kg for phenanthrene, 72 mg/kg for fluorene, 23 mg/kg for carbazole, 22 mg/kg for dibenzothiophene, 75 mg/kg for dibenzofuran, and 1,100 mg/kg for acridine. For all substances but acridine, nitrification was the most sensitive of the three toxicity indicators evaluated. No effect of the tested substances on bacterial diversity was found, as measured by denaturant gradient gel electrophoresis. In general, only weak effects at very high concentrations were found for the protozoans. However, for acridine, protozoan numbers were reduced at lower concentrations than those that affected the nitrification process, that is, with a 5% reduction at 380 mg/kg. For effects on nitrification, toxicity (NOEC values) expressed as soil pore-water concentrations (log10(micromol/L)) showed a significant inverse relationship with lipophilicity (log octanol-water partition coefficient) of the substances (r2 = 0.69, p = 0.011, n = 8). This finding could indicate that the toxicity of substances similar to those tested might be predicted by a quantitative structure-activity relationship with lipophilicity as the predictor variable.