Arguments which suggest that improved prediction of multiple criteria can be achieved employing pattern scoring of responses, as opposed to conventional methods, are examined. Models for improving prediction of single and multiple criteria were examined. The findings are: (1) simple linear combinations of predictor variables perform as well predicting single criterion as do nonlinear combinations of the variables; and (2) for predicting multiple criteria, the most appropriate model also appears to be simple linear combinations of the predictor variables. It is suggested, however, that when a number of measures, multifactor in nature, are to be predicted, the model used in the present study may be appropriate for improving predictability. (CK)