The magnitude of biases introduced into population estimates by the varying catchability of fish was observed by employing several strategies for estimating population sizes of northern pikeEsox luciusin six north-central Minnesota lakes. The strategies relied on commonly used sampling techniques for northern pike and differed in the degree to which they addressed differences in catchability among fish. Biases were measured as percent differences among population estimates. Estimates of known populations of tagged fish showed that the Petersen technique, which incorporated independent gear types for marking and recapture, was relatively unbiased. Biases from differences in catchability due to fish size were small. In contrast, large biases from differing fish behaviors and habitat uses were found in removal or multiple mark–recapture methods. Both method types relied on sampling nearshore movements of northern pike during early spring, and both severely underestimated population sizes. Removal estimates were lower than the number of fish actually caught, and multiple mark–recapture estimates averaged 39% lower than Petersen estimates. These results suggest that population estimates derived by sampling a fish behavior pattern (such as nearshore spring movements of northern pike) may not be valid. The assumption of equal catchability during such movements was very sensitive. Violations of this assumption caused critically large differences in population estimates for northern pike.