Earthworms, Lumbricus rubellus, were collected in three successive years, 1997, 1998 and 1999 from 5, 6 and 4 sites, respectively, along a metal gradient near a smelter at Avonmouth, England. In total 271 worms were collected, 47 in 1997, 97 in 1998, and 127 in 1999. The worms were analysed for the enzyme esterase by means of agarose electrophoresis and for internal metal concentration. At one site collected in three successive years, 64 individuals of 66 were heterozygous for the esterase enzyme. The metal concentrations (Cu and Zn) in worms were significantly correlated with the total concentration and the CaCl2 extractable fraction of the metal in the soil. One of the homozygous esterase genotypes was significantly correlated with distance from the smelter, the extractable fraction of metals from soil, and the metal concentration in the worms. The other homozygous genotype was correlated with the CaCl2 extractable Cu and the Cu concentration in the worms. The internal levels of Cd, Cu, and Zn showed a higher classification potential (Discriminant Analysis) for 50 of 127 individuals, collected in 1999, when based on the original habitat of the earthworms than on the esterase genotype. The remaining 77 specimens, representing three sites, were exposed for 30 days in the laboratory to soils from four sites. For the individuals in this transplantation experiment, parameters measured were the esterase phenotype, internal metal concentration (Cd, Cu, and Zn), mortality, wet weight change, cocoon production, and cocoon hatching success. Cocoon production was highest among the worms from the most contaminated area, while worms from the least contaminated area had the highest hatching success. Neither the metal concentrations, in soil or internal, nor the life-history parameters could fully explain the genotypic distribution of esterase along the gradient. The internal concentrations of Cd, Cu and Zn in the worms used for the transplantation experiment showed the highest classification potential (Discriminant Analysis) when the prediction was based on the original habitat of the earthworms. The classification potential was less strong when based on esterase genotype and least on the transplantation conditions. Including life history traits did not enhance the prediction potential.