An extensive multivariate analysis procedure for prediction of the blastfragmentation is presented. Several blasts performed in various mines and rock formations in the world are compiled and evaluated. Blast design parameters, modulus of elasticity, in situ block size are considered in the multivariate analysis. Two main approaches for prediction of blastfragmentation given in the literature are used for comparison. First, a meanfragmentsize prediction model is developed. Then fragmentationindices are determined for each blast and the fragmentationindex prediction model is created using discriminant analysis method. Discriminant analysis played a key role during design stage of both models. The structure and prediction ability of the two models are compared and discussed. Models can be used separately or to complement each other. The models are applied to a wide range of test blasts. The models are not complex and suitable for practical use in mines. (C) 2012 Elsevier Ltd. All rights reserved.