The prediction of particle size and minerals distribution in products from size reduction processes is one of the fundamental problems in mineral processing. In this paper, a 2D model based on cellular automata. is proposed to simulate particles generation by discriminatory size reduction. This model generates particles on images of ore textures, through the growing of nucleation seeds. The size reduction process is simulated as a growing process of seeds over representative images of the ore texture. Each particle grows according to rules related with the probability of agglomeration between different seeds. The final pattern represents the sections of particles corresponding to the mill product, for a specified milling time. Preliminary results of areal size distributions and mineral grade distribution shows values quite similar to experimental ones. The incorporation of an optimisation algorithm, will allow calibration of the model to improve these results. (C) 2006 Elsevier Ltd. All rights reserved.