Estimation of age composition was considered in the context of non-parametric mixture distribution problems. The problem was formulated as a Fredholm first-kind equation with respect to the distribution function of age. The solution of the equation was obtained using methods for ‘ill-posed problems’. The approach is able utilise single and multi-dimensional empirical distributions of age-dependant variables. School whiting (Sillago flindersi) fish length and otolith weight measurements were used in a numerical example. Different combinations and sample sizes of input data were used for numerical error analysis.