Explains that from a prediction standpoint the confluence model is not very efficient. Very modest increments in accuracy are associated with family configuration variables once chronological age is covaried. Suggests that the major postulates of the theory be tested directly, within individuals and with longitudinal data. (Author/AS)