Abstract Estimates of population size are needed in populations targeted by recreational fisheries, to determine their conservational status and to formulate fishing regulations. Mark–recapture monitoring is a promising tool for this because of the increased popularity of catch-and-release angling. Northern pike, Esox lucius L., population sizes were estimated in four lakes based on experimental mark–recapture fishing using a hierarchical Bayesian model that also estimated natural mortality. The approach was compared with output from traditional Petersen’s population size estimates. Despite intensive fishing with multiple gears, the population size estimates suggested that a substantial proportion of individuals remained unobserved by fishing. Petersen’s estimates depended on the assumed rate of natural mortality, and no mortality rate scenario matched with the Bayesian estimates. Consequently, the number of individuals caught by recreational anglers is a poor proxy of population size for which estimations should be carried out using a modelling approach that also estimates natural mortality.