A critical look at some widely used estimators in mark-resighting experiments.

1. For many species and circumstances, mark-resighting procedures constitute valid alternatives to capture-recapture methods. Indeed, resightings are generally cheaper to acquire than physically recapturing and rehandling the animals, especially when radiotelemetry or other tracking devices are available. 2. In order to estimate population abundance, the joint hypergeometric maximum likelihood estimator, the Minta-Mangel estimator and the Bowden estimator are implemented in noremark, software which has become very popular with biologists in the past decade. 3. In this paper, the basic assumptions regarding these widely applied procedures are delineated and discussed. A simulation study is performed in order to investigate the robustness of the estimators under failure of the assumptions. 4. Theoretical considerations and simulation results motivate the use of the Bowden estimator which, when marks are distributed quite evenly among groups, constitutes the sole reliable method, offering computational simplicity and robustness. On the other hand, if the marks are distributed unevenly, no mark-resighting procedure seems reliable. An application to a case study is considered.