A method is proposed to calibrate the well known population balance model to batch grinding test results using a data reconciliation approach. The method uses empirical representations for the breakage and selection functions. The main difference between the proposed method and conventional calibration approaches is related to the fact that the feed size distribution is estimated as well as the product size distributions observed for various grinding times. The method is illustrated with data from a batch grinding test. (c) 2005 Elsevier Ltd. All rights reserved.