THE SELECTION INDEX - THEN, NOW, AND FOR THE FUTURE

When improvement is desired for several traits that may differ in vari ability, heritability, economic importance, and in the correlation amo ng their phenotypes and genotypes, simultaneous multiple-trait index s election was more effective than independent culling levels or sequent ial selection. Such comparisons required definition of aggregate breed ing value determined jointly by breeding values and the economic impor tance of the component traits. The economic weight should approximate the partial regression of cost per unit of enterprise output value on breeding value for each trait. These can vary with production and mark eting system, with performance of traits, and with breed role (i.e., p aternal, maternal, or general) in crossbreeding systems. Genetic gains desired to maintain competitive ranking also may define the relative importance of traits. Because information available to estimate breedi ng values varies among the ages and categories of individuals under se lection and because means are unknown, regressed (BLUP) predictions of trait breeding values are useful. They allow appropriate economic wei ghts to be applied as the last step for predicting aggregate breeding values for individuals of different age classes, and they simplify cho osing the proportions of selected breeders from each age class that ma ximize rate of change in aggregate breeding values. Inappropriate econ omic weights or errors in the parameters used to predict trait breedin g values overestimate realized response in true aggregate breeding val ue.