The restrictive assumptions required by most available methods for estimating survival probabilities render them unsuitable for analyzing real banding data. A model is proposed which allows survival rates and recovery rates to vary with the calendar year, and also allows juveniles to have rates different from adults. In addition to survival rates and recovery rates, the differential vulnerability factors of juveniles relative to adults are estimated. Minimum values of the variances of the estimators are also given. The new procedure is applied to sets of duck and goose data in which reasonably large numbers of adult and juvenile birds were banded. The results are shown to be generally comparable to those procured by other methods, but, in addition, insight into the extent of annual variation is gained. Combining data from adults and juveniles also increases the effective sample size, since the juveniles are assumed to enter the adult age class after surviving their initial year.