Estimation of contaminant concentrations in earthworms is a critical component in many ecological risk assessments. Without site-specific data, literature-derived uptake factors or models are frequently used. Although considerable research has been conducted on contaminant transfer from soil to earthworms, most studies focus on only a single location. External validation of transfer models has not been performed. We developed a database of soil and tissue concentrations for nine inorganic and two organic chemicals. Only studies that presented total concentrations in depurated earthworms were included. Uptake factors and simple and multiple regression models of natural-log-transformed concentrations of each analyte in soil and earthworms were developed using data from 26 studies. These models were then applied to data from six additional studies. Estimated and observed earthworm concentrations were compared using nonparametricWilcoxon signed-rank tests. Relative accuracy and quality of different estimation methods were evaluated by calculating the proportional deviation ([measured - estimate]/measured) of the estimate from the measured value. With the exception of Cr, significant, single-variable (e.g., soil concentration) regression models were fit for each analyte. Inclusion of soil Ca improved model fits for Cd and Pb. Soil pH only marginally improved model fits. The best general estimates of chemical concentrations in earthworms were generated by simple ln-ln regression models for As, Cd, Cu, Hg, Mn, Pb, Zn, and polychlorinated biphenyls. No method accurately estimated Cr or Ni in earthworms. Although multiple regression models including pH generated better estimates for a few analytes, in general, the predictive utility gained by incorporating environmental variables was marginal.