Zusammenfassung: An adjustment method for age-length keys is presented that incorporates yearly variability in cohort abundances. The method is based on two models (one for growth and the other for standard deviations of length versus age) and involves an iterative algorithm. The algorithm rapidly converges to stable results that yield, under some conditions, maximum likelihood estimates. It is also used to estimate parameters of the length standard deviation model. The method is applied to yellowfin tuna (Thunnus albacares) catches in the eastern Atlantic from 1975 to 1989.