Previous studies of toxicant effects on fish foraging behavior have taken a strictly empirical approach which had little basis in ecological theory and did not allow development of a testable hypothesis a priori to field verification of laboratory results. A mechanistic approach was taken in this study to examine the sublethal effects of four copper concentrations (0, 25, 150, 1800 ug/l) on the foraging behavior of bluegill Lepomis macrochirus. Twelve separate laboratory experiments were conducted. Two experiments assessed copper effects on the reaction distance of bluegill to untreated zooplankton. Ten experiments were done to assess copper effects on the functional response of bluegill to untreated (5 tests) and treated (5 tests) invertebrate prey of different types. Copper had no effect on the reaction distance of fish to zooplankton. Copper reduced bluegill consumption rates of all untreated prey but not treated prey groups. Prey handling time for bluegill capturing treated and untreated prey increased significantly with copper concentration and was the most consistently sensitive parameter measured. Capture efficiency by bluegill was not as consistent a measure of toxicant stress. These experiments suggest that mechanistic measures are consistent indicators of toxicant effects on fish feeding behavior and that copper concentrations as low as the USEPA water quality criteria may alter food consumption and reduce growth of fish in the wild;Laboratory results were incorporated into an optimal foraging model to predict the diet breadth of treated and untreated fish in a lake. Bluegill in replicated, littoral-zone mesocosms were exposed to 1 of 3 copper concentrations (0, 36, 130 ug/l) for four days. The stomach contents of the fish were compared with ambient prey availability and the diet predicted by the optimal foraging model. The model did not provide a good estimate of diet breadth for fish in the mesocosms. Small differences in the energetic return rate between optimal and suboptimal diets, and errors in determining estimates of model parameters may account for discrepancies between the observed and predicted diet.