This paper presents a general method leading to the determination of the breakage and selection matrices from particle size analysis data obtained in batch grinding tests. It is based on a mathematical transformation of the batch grinding problem which allows sequential differentiation to determine the selection matrix independently of the breakage matrix. This latter is then obtained by simple unconstrained linear optimization. When tested in a simple procedure, the proposed technique is shown to give good results with respect to a direct search method based on the Levenberg-Marquardt algorithm.