We offer a theory of the direction and nature of representational linkages between constituents and their elected representatives based on two attributes of issues: their complexity and their relationship to the lines of partisan cleavage. We show that the theory is compatible with the existing evidence on representation and then offer results of tests of new predictions from the theory for both simple and complex party-defining issues. For additional evidence of the dyadic basis of these findings, we also show that the strength of the observed linkages varies in accordance with theoretical expectations about the seniority of members of Congress and, for senators, recency of election. We also explain how the theory can account for a number of seemingly contradictory empirical findings in the large literature on policy representation and how it allows scholars to make precise predictions about the characteristics of representational linkages.The study of policy representation enjoys a long research history in political science. The bulk of this research rests on a theory of representation that posits constituency opinion as an independent variable and representatives' attitudes and actions as dependent variables. We have learned a great deal about the relationships between masses and policy-making elites from research in this tradition (see, among many others, Cnudde and McCrone 1966;Kuklinski and Elling 1977;McCrone and Kuklinski 1979;Miller and Stokes 1963;Page et al. 1984). Yet, newer work on the subject raises doubts about the universal applicability of the static demand-input model of the representational process underlying such research by demonstrating that public opinion is not necessarily an exogenous and fixed force in the political system. Rather, public opinion itself can at times be influenced by events, elite political actors, or by public policy itself (see, among others, Cunningham andWe owe thanks to Scott Robinson for insightful comments that contributed to the development of our theoretical argument presented here, to Stephen Hanna and Carl Klarner for substantial assistance with the data collection and analysis for this paper, and to Gerald Wright, Jr., for a portion of the data used in the analysis. Other data were made available by the ICPSR, which bears no responsibility for the analysis or interpretations.