¹ Dahl, Robert, A Preface to Democratic Theory (Chicago: Univ. of Chicago Press, 1956), chapters 3 and 5Google Scholar.

² Downs, Anthony, An Economic Theory of Democracy (New York: Harper, 1957), pp. 23–24Google Scholar.

³ Mayo, Henry B., An Introduction to Democratic Theory (New York: Oxford University Press, 1960), chapter 4Google Scholar.

⁴ Berns, Walter, “A Critique of Berelson, Lazarsfeld, and McPhee's Voting,” in Public Opinion and Public Policy: Models of Political Linkage, ed. Luttbeg, Norman R. (Homewood, Illinois: Dorsey, 1968), p. 32Google Scholar.

⁵ Miller, Warren E. and Stokes, Donald E., “Constituency Influence in Congress,” American Political Science Review, 57 (03, 1963), 49CrossRefGoogle Scholar.

⁶ See, for example, Converse, Philip E., “The Nature of Belief Systems in Mass Publics,” Ideology and Discontent, ed. Apter, David (New York: Free Press, 1964), pp. 206–261Google Scholar, and Luttbeg, Public Opinion and Public Policy.

⁷ An exception to this statement is found in Fishel, Jeff, “Party, Ideology, and the Congressional Challenger,” American Political Science Review, 63 (12, 1969), 1213–1232CrossRefGoogle Scholar.

⁸ John L. Sullivan, “Linkage Models of the Political System,” Public Opinion and Political Attitudes: A Reader, ed. Alan R. Wilcox, forthcoming.

⁹ See, among others, Truman, David B., The Congressional Party (New York: Wiley, 1959)Google Scholar; Turner, Julius, Party and Constituency: Pressures on Congress (Baltimore: The Johns Hopkins University Press, 1951)Google Scholar; and MacRae, Duncan, Dimensions of Congressional Voting (Berkeley: University of California Press, 1958)Google Scholar.

¹⁰ Fishel, p. 1221.

¹¹ Sullivan, Part V.

¹² The individual voter cannot arrive at this conclusion unless he has first familiarized himself with the issue positions of the candidates and the later roll-call behavior of the winners. We know that the possession of this information is the exception rather than the rule. If, however, a district (or an entire political system) has in the recent past offered little choice, it may become part of the collective attitude of the citizenry that all attempts to gather this type of information are wasted.

¹³ Rose, Richard and Mossawir, Harve, “Voting and Elections: A Functional Analysis,” Political Studies, 15 (06, 1967), 173–201CrossRefGoogle Scholar.

¹⁴ See Appendix A for the questionnaire, roll calls, and scaling. Congressional Quarterly was able to obtain a 100 per cent response rate by making ample use of telephone followups. In several cases, aides answered the questions, but presumably they gave the same answers the candidate would have given. If the incumbent was absent for a roll call, he was asked how he would have voted had he been present.

¹⁵ The roll calls for the 90th Congress are presented in Appendix B. If the representative was absent but announced for or against, we accepted this announcement as indicating how he would have voted had he been present. If there was no information available, we coded his response as a moderate one, except where otherwise noted (Table 7).

¹⁶ One complicating factor is the amount of choice offered in the primary elections. The available evidence, however, suggests that issues only weakly define primary elections. See Key, V. O. Jr., Southern Politics (New York: Knopf, 1949), pp. 105, 117, 142Google Scholar. The evidence further suggests that the amount of choice offered in primary elections in safe districts is not great. For example, Turner, Julius, “Primary Elections as the Alternative to Party Competition in ‘Safe’ Districts,” Journal of Politics, 15 (05, 1953), 197–210CrossRefGoogle Scholar, found that, between the years 1944–1950, only half the primaries in safe districts were even contested. No doubt the candidates differed only marginally on the issues in at least some of the contested primaries.

¹⁷ In each case, the most liberal and most conservative Congresses are based on the scores within the issue domain considered, rather than on the overall scale scores. Again, because of the scalability of the items, the most liberal and most conservative Congresses do not differ appreciably even when the overall scale scores are used to determine which candidate is more liberal. In no case is the mean or median different from those presented in Tables 3–5 because there are very few cases where one candidate is more liberal on the overall liberalism-conservatism scale but more conservative on one of the within-issue area scales, or vice versa. It was done this way to ensure that Tables 3–5 present the most conservative and most liberal Congresses possible in each issue cluster. The same procedure is followed in the data presented in Table 12.

¹⁸ Riker, William, The Theory of Political Coalitions (New Haven: Yale University Press, 1962)Google Scholar.

¹⁹ Miller, and Stokes, , “Constituency Influence in Congress,” p. 48Google Scholar, report correlations between the true opinion of Congressmen (as measured by a questionnaire) and their roll-call behavior of 0.75 among those who felt their attitudes were expressed in their roll calls, but only 0.04 among those who felt that they were not.

²⁰ Wicker, Allan W., “Attitudes versus Actions: The Relationship of Verbal and Overt Behavioral Responses to Attitude Objects,” Journal of Social Issues, 25 (Autumn, 1969), 65CrossRefGoogle Scholar.

²¹ Wicker, pp. 67–74.

²² There were no roll calls comparable to the foreign affairs, labor, or voting rights items of the pre-election scale. Therefore, in the tables that follow, these items have been dropped from the initial measures. The included roll calls are listed in Appendix B. It should also be noted that the foreign affairs data are questionnaire responses for both incumbents and nonincumbents (see Appendix A). Therefore, although we cannot examine the linkage condition in foreign affairs, the methodological question of comparability does not arise.

²³ Gamma is stronger than Sommer's d_xy because it ignores all ties. It considers only the concordant and discordant pairs, while Somer's d_xy is penalized for every case in which the dependent variable varies but the independent variable does not vary. The actual theoretical relationships of Tables 7–10 are probably overestimated by the strength of gamma because if two cases are tied on one variable but not on the other, they ought not to be ignored unless the theory predicts that this should happen. This is not the case here. Conversely, Somer's d_xy probably underestimates the relationship because some of the cases which lower Somer's d_xy, cases which differ on the dependent variable but not on the independent variable, do so because of measurement error. Two nonincumbents whose preelection scores are 8 on domestic issues might be expected to differ in postelection behavior because the intermediate option is not so easily taken on roll calls as it is on questionnaires, and their reasons for hedging on one issue in the pre-election questionnaire may be quite different, leading to a different postelection roll-call decision. Somer's d_xy is included, however, to illustrate the range of values obtained, depending on the method of handling ties. Somer's d_xy was selected over d_yx, which is reduced by cases which vary on the independent but not the dependent variable, because of the nature of the two different kinds of measurement. The questionnaire allowed more options than a roll call, in that abstention on a roll call is more difficult than selecting the intermediate option on a questionnaire. Therefore, the situation of variation on the independent variable coupled with ties on the dependent variable is slightly more likely to occur than is variation on the dependent variable coupled with ties on the independent variable.

²⁴ Shapiro, Michael J., “Rational Political Man: A Synthesis of Economic and Social-Psychological Perspectives,” American Political Science Review, 63 (12, 1969), 1106–1119CrossRefGoogle Scholar.

²⁵ Brown, Steven R. and Taylor, Richard W., “Objectivity and Subjectivity in Concept Formation: Problems of Perspective, Partition, and Frames of Reference,” paper delivered at the sixty-sixth annual meeting of the American Political Science Association, Los Angeles, California, 09, 1970Google Scholar.

²⁶ Miller and Stokes, p. 50, report correlations between constituency majority and congressional roll-call votes of 0.4, and between district majority and nonincumbent candidate attitudes of −0.4, in the social welfare policy domain. This suggests that within the confines of the amount and kind of choice presented, the masses are indeed selecting those elites whose attitudes are most congruent with their own. The fact that the congruency is not higher may be due primarily to the differing attitudinal structures of elites and nonelites.

²⁷ If the electorate votes primarily on the basis of party identification and if these identifications are unevenly distributed within a large percentage of districts, the result is a lack of intradistrict competition. If this uneven distribution of identifications reflects an equally uneven distribution of opinions and preferences on issues, then linkage could occur without a choice of candidates who differ in their issue positions. Therefore the optimal amount of choice in each district would depend on its internal distribution of opinion. Since data of this nature are not available, this analysis has focused on the systemic, rather than district, consequences of choice. From this framework, the more choice, the greater the opportunity for the electorate as a whole to exercise control over policy.

²⁸ Turner, , “Primary Elections,” p. 198Google Scholar.

²⁹ In all three issue clusters, the actual Congress resembles the most conservative Congress probable, given the existence of a large number of safe districts. To the extent that it makes sense to label an election a liberal or a conservative victory, the 1966 House election was a conservative victory.