The measurement and analysis of change remain persistent dilemmas in aging research. The most frequently suggested technique for the analysis of change is residualized change score analysis, which is methodologically superior to the use of raw change scores. The use of residualized change score analysis, however, addresses a very specific substantive question: net of initial level what factors best predict the dependent variable at a later point in time? There are other questions that might be posed to longitudinal data. One such question is: what factors best predict differential patterns of change within a group? The primary purpose of this paper is to present a canonical correlation technique that can be used to predict patterns of change. In order to illustrate the implications of examining patterns of change as compared to aggregate change, the canonical correlation procedure is compared to residualized change score analysis. Discussion focuses upon the importance of matching substantive questions about the nature and antecedents of change to appropriate analytic techniques. © 1982 Taylor & Francis Group, LLC.