A data set was generated aiming to predict the toxicity of PAHs to soil organisms. Toxicity data include the effects of 16 PAHs on the survival and reproduction of the soil-dwelling springtail Folsomia fimetaria. The results show that only PAHs with reported log Kow values < or = 5.2 (i.e., naphthalene, acenaphthene, acenaphthylene, anthracene, phenanthrene, fluorene, pyrene, and fluoranthene) significantly affected the survival or reproduction of the test organisms. Threshold values for the toxicity of the individual PAHs could be expressed as pore-water concentrations by the use of reported organic carbon-normalized soil-pore-water partitioning coefficients (Koc values). For the PAHs with a log Kow < or = 5.2, toxicity significantly increased with increasing lipophilicity of the substances (r2 = 0.67; p = 0.012; n = 8), suggesting a narcotic mode of toxic action for most substances. However, the position of anthracene in the regression plot indicated a more specific mode of toxic action than narcosis, and removing this data point yielded the following regression equation: log EC10 (micromol/L) = -0.97 log Kow + 4.0 (r2 = 0.80; p = 0.006; n = 7). Using this quantitative structure-activity relationship (QSAR) to calculate threshold values for the toxicity of the remaining nontoxic substances (benz[a]anthracene, chrysene, benzo[b]fluoranthene, benzo[k]fluoranthene, dibenz[a,h]anthracene, benzo[a]pyrene, perylene, and indeno[1,2,3-cd]pyrene), the absence of toxicity could, in most cases, be explained by a limited water solubility, indicating that these substances do act by narcosis as the mode of toxic action and that their toxicity is governed by concentrations in the pore-water. The results provide important input to future model predictions of the ecological risk posed by PAH contaminated sites.