When predictors or independent variables are correlated, it is difficult to interpret their relationships to the dependent variable. If the variables are orthogonalized, then independent tests of significance are possible and the situation is greatly simplified. The oorrelated independent variables can be orthogonalized in a variety of ways, one of which gives the classical beta weights. Three general cases of orthogonalizing predictors are presented; each accounts for theme amount of variance as does the usual multiple correlation but may lead to different conclusions. Several of these methods underlie the ANOVA procedures used when cell frequencies are propontional rather than equal. Replication of the results across both subjects and variables is discussed. Selection of a particular procedure for dividing the overlapping variances is dictated by the theoretical perspective of the study.