Current methods for estimating age-specific survival rates from the ring recoveries of dead birds depend critically on a number of assumptions. Even if these assumptions hold, current estimating procedures require the imposition of a constraint. Because survival probabilities of birds are believed to become constant with increasing age, a commonly used constraint is that the survival probabilities in successive age-groups of older birds are equal. Although this constraint appears biologically reasonable, its use can be highly misleading. Even slight departure from exact equality in the survival of older age-groups can produce survival estimates which are grossly in error. The constraint achieves mathematical tractability but does not necessarily give reliable survival estimates, no matter how many birds are ringed. The need for the constraint arises, in the 1st place, because the generally used design of field observations is inadequate to provide sufficient biological information to estimate the parameters of interest. If the usual program of observations can be supplemented by additional relevant information, this could make the problem of survival estimation mathematically feasible and at the same time reduce the degree of uncertainty in the estimates. Estimates of age-specific survival are likely to be untrustworthy, if they are based solely on recoveries of dead birds.