An in situ experiment was designed (1) to determine the relative importance of Cd in the sediment versus Cd in the overlying water for its accumulation in benthic animals and (2) to test in situ the acid volatile sulfide (AVS) model. Sediments of a low-Cd shield lake were artificially contaminated with various amounts of Cd and placed at a littoral site in open plastic containers in the lake bottom for 11 months to allow colonization by indigenous lake bcnthos. Gradients in sedimentary and interstitial water Cd concentrations were, thus, created in situ, although Cd concentrations in the overlying water remained low. Accumulation of Cd by benthic invertebrates was significantly and linearly correlated with sediment Cd concentrations across treatment levels for most taxa. The magnitude of the response to the sediment Cd gradient was taxon specific and consistent with differences in animal ecology. Despite the high Cd contamination levels attained in the experiment, the population densities of most taxa appear to have been unaffected by the presence of Cd. We used the results from the experimental containers to predict the sources of Cd for benthic animals in the environs. For almost all of the taxa studied, individuals living in the lowCd sediments outside the experimental containers accumulated their Cd almost exclusively from the overlying water; only the burrowing, sediment-feeding taxa Chironomus stuegeri and members of the Tubificidae obtained substantial Cd from the sediment. Cadmium activities in the pore waters of the containers were in agreement with the AVS model, but the Cd concentrations found in the benlhic organisms did not. Our results from this shield lake suggest that predictions of Cd concentrations in most benthic animals would be more accurate if they were based on water column rather than on sedimentary Cd concentrations.Metals associated with sediments pose a potential threat to benthic organisms and from there to upper trophic levels. Before expensive remedial actions are undertaken, the extent to which benthic animals accumulate sedimentary metals, the biological effects of these contaminants, and the effects of various remedial actions on both water quality and organisms need to be determined. To accomplish these tasks, environmental agencies need reliable predictive tools; however, the majority of the tools currently available are unsatisfactory (Luoma 1995). Our ability to predict the biological consequences of metal-contaminated sediments would be improved by the development of reliable models relating metal bioaccumulation or toxicity to metal concentrations in the environment of benthic organisms. For general application, such models should be based on solid scientific grounds, i.e., they should have a theoretical rather than a I Present address: