Traditional estimation of age composition of catch requires that age/length data and random length frequency have been sampled from the same population. If such sampling is not possible the estimates of the age distribution could be biased. By using an iteration algorithm it is possible to overcome this limitation. The age composition of catch for Dissostichus eleginoides was obtained with the application of a single age/length key to the monthly length distribution using the non-linear model (Hoenig and Heisey, 1987). We found that age distribution derived from the model is not significantly different from the one obtained by using traditional methods