An iterative method is presented, based on the method of Moav and Hill (1966) to derive the optimum linear selection index for any number of traits with linear or non-linear profit functions. For non-linear profit functions the index weights will be functions of the trait means prior to selection and the selection intensity. Using the equations developed, the optimum selection index for three dairy cattle milk production traits was computed. Convergence was obtained after three to four iterations, and was robust to the starting values used for iteration. The ratio of expected genetic gains were only marginally different for selection intensities of 1 and 4 standard deviation units. Differences were greater for the index coefficients. All alternative indices tested gave lower gains in profit than the optimum index. For linear profit functions this index reduces to the standard linear index, and for two uncorrelated traits this index reduces to the index of Moav and Hill (1966).
