Although the recruitment of breeders is as important to population dynamics as mortality, it has received far less attention from statisticians working at modelling capture histories of marked animals. However, two main approaches to studying local recruitment have become available in the last three years. Both deal with birds marked as chicks and then possibly resighted or recaptured only as breeders (the same approaches may be applied more generally whenever breeders and non-breeders can be distinguished in a population by working with the sole observations of breeding animals). This situation is natural to species of colonial birds sampled on the breeding site where the young remain away from the breeding site previous to their first breeding attempt. The first method published (Clobert et al., 1994) contrasts, in the context of a survival analysis, the capture rates of known breeders and the capture rates of animals up to their first observed breeding attempt. Young animals as a group are less likely to be 'captured' if some have not yet started to reproduce. It is then possible to estimate age-specific breeding probabilities, i.e. the proportions of breeding animals in each age-class. The second approach (Pradel, 1996, Pradel et al., 1997) addresses recruitment directly. Capture histories are read backwards starting from the last observed breeding event and asking the question: when in the past did an animal start breeding? In this way, one can estimate the probability that a breeding animal is a first-time breeder as a function of age, individual characteristics or environmental conditions. A third approach based on the history of individual animals is presented here for the first time. It involves estimating the probability that an as yet inexperienced animal starts to breed and can be implemented as a particular case of a twostate model with a non-observable state (Lebreton et al., 1999, this volume). Under general assumptions, there are simple relationships between the different parameters that measure accession to reproduction but not all seem equally flexible nor relevant to the same questions. Although the recruitment of breeders is as important to population dynamics as mortality, it has received far less attention from statisticians working at modelling capture histories of marked animals. However, two main approaches to studying local recruitment have become available in the last three years. Both deal with birds marked as chicks and then possibly resighted or recaptured only as breeders (the same approaches may be applied more generally whenever breeders and non-breeders can be distinguished in a population by working with the sole observations of breeding animals). This situation is natural to species of colonial birds sampled on the breeding site where the young remain away from the breeding site previous to their first breeding attempt. The first method published (Clobert et al., 1994) contrasts, in the context of a survival analysis, the capture rates of known breeders and the capture rates of animals up to their first observed breeding attempt. Young animals as a group are less likely to be 'captured' if some have not yet started to reproduce. It is then possible to estimate age-specific breeding probabilities, i.e. the proportions of breeding animals in each age-class. The second approach (Pradel, 1996, Pradel et al., 1997) addresses recruitment directly. Capture histories are read backwards starting from the last observed breeding event and asking the question: when in the past did an animal start breeding? In this way, one can estimate the probability that a breeding animal is a first-time breeder as a function of age, individual characteristics or environmental conditions. A third approach based on the history of individual animals is presented here for the first time. It involves estimating the probability that an as yet inexperienced animal starts to breed and can be implemented as a particular case of a twostate model with a non-observable state (Lebreton et al., 1999, this volume). Under general assumptions, there are simple relationships between the different parameters that measure accession to reproduction but not all seem equally flexible nor relevant to the same questions.