Résumé
The great flexibility of capture-recapture models to estimate survival and recruitment parameters is described. Recruitment models appear as analogous to survival models applied to capture histories reversed in time. The available flexibility is used at its best by basing model selection on the minimization of Akaike's information criterion within a set of biologically plausible models. However, to proceed to model selection, one needs to determine the number of identifiable parameters. We explain how SURGE version 5.0 (available on the Internet by anonymous ftp from ftp.cefe.cnrs-mop.fr) makes it possible to fit survival and recruitment models and determines numerically how many and which parameters are concerned with identifiability problems. Model selection for survival and recruitment models is illustrated using European Dipper Cinclus cinclus data. The advantages and shortcomings are discussed.