Handling mortality occurs in mark–recapture experiments if animals handled and released in a given sample have a higher mortality rate than animals that were alive but not sampled. This violates the assumption of equal survival required for forming the Jolly–Seber estimates of population abundance, survival, and recruitment. We show that handling mortality can produce very large biases in these estimates, and we develop a test to detect it. We investigate the power of this test and find that quite large biases can be produced at handling mortality rates that are too low to be detected. We recommend methods to prevent handling mortality from occurring in fish sampling experiments and methods to reduce bias in the estimates. The test and the bias corrections are applied to mark–recapture data for a lake whitefish (Coregonus clupeaformis) population and to data from simulated mark–recapture experiments. Because of unavoidable inadequacies in the detection and reduction of bias due to handling effects, we strongly recommend that fisheries biologists take great care to prevent its occurrence.