¹ Key, V. O. Jr., Politics, Parties, and Pressure Groups (New York: Crowell, 1958) 4th ed., pp. 267–268Google Scholar; “Partisanship and County Office: The Case of Ohio,” this Review, 47 (1953), 525–532Google Scholar; Key, V. O. Jr., and Munger, Frank, “Social Determinism and Electoral Decisions …,” in Burdick, Eugene and Brodbeck, Arthur J. (eds.), American Voting Behavior (Glencoe: Free Press, 1959), pp. 281–299Google Scholar.

² See, inter alia, Harris, Louis, Is There a Republican Majority? (New York: Harper, 1954), pp. 118–138Google Scholar; Wood, Robert C., Suburbia, (Boston: Houghton Mifflin, 1958), pp. 134–153Google Scholar; Greenstein, Fred I. and Wolfinger, Raymond E., “The Suburbs and Shifting Party Loyalties,” Public Opinion Quarterly, 22 (1958), 473–482CrossRefGoogle Scholar.

³ Berelson, Bernard R.et al., Voting (Chicago: University of Chicago Press, 1954), pp. 98ffGoogle Scholar. et passim; Miller, Warren E., “One-Party Politics and the Voter,” this Review, 50 (1956), 707–725Google Scholar; Cutright, Phillips and Rossi, Peter, “Grass Root Politicians and the Vote,” American Sociological Review, 23 (1958), 171–179CrossRefGoogle Scholar; Katz, Daniel and Eldersfeld, Samuel J., “The Impact of Local Party Activity Upon the Electorate,” Public Opinion Quarterly, 25 (1961), 1–24CrossRefGoogle Scholar; Ennis, Philip H., “The Contextual Dimension in Voting,” in McPhee, William N. and Glaser, William A. (eds.), Public Opinion and Congressional Elections (Glencoe: Free Press, 1962), pp. 180–211Google Scholar; Campbell, Anguset al., The American Voter (New York: Wiley, 1960), pp. 286–289Google Scholar; Epstein, Leon, Politics in Wisconsin (Madison: University of Wisconsin Press, 1958), pp. 61–70Google Scholar. For a summary of further studies on the related topics of “clustering,” “enclave,” and similar effects, see Lane, Robert E., Political Life (Glencoe: Free Press, 1959), pp. 261–272Google Scholar. For a discussion of phenomena analogous to community influence in the context of trade union elections, see Lipset, Seymour Martinet al., Union Democracy (Glencoe: Free Press, 1956), pp. 338ff.Google Scholar

⁴ “The Political Implications of Community Identification,” in Young, Roland (ed.), Approaches to the Study of Politics (Evanston: Northwestern University Press, 1958), pp. 318–328Google Scholar.

⁵ See The American Voter, pp. 295–332.

⁶ For an introduction to the voluminous literature on personal influence, see Katz, Elihu and Lazarsfeld, Paul F., Personal Influence (Glencoe: Free Press, 1955)Google Scholar, Part One, and the items cited there. See also Berelson et al., op. cit., pp. 126–27.

⁷ See footnote 22 below.

⁸ Non-voters have been excluded from the entire analysis, as their inclusion would have complicated it immensely. Southern respondents have been excluded, since preliminary analysis of data from Southern respondents gave some reason to doubt that the nature of the community influence process is quite the same in that region as in the rest of the country. The entire study has, wherever possible, been replicated using the 1960 SRC survey data. All findings reported here were confirmed in that replication, except those involving primary and secondary group membership. It was impossible to replicate these latter findings, since the 1960 survey did not include the relevant questions.

⁹ Key, op. cit.; Miller, op. cit.; The American Voter, pp. 286–289.

¹⁰ Data from Scammon, Richard M. (ed.), America Votes (New York: Macmillan, 1956), Vol. IGoogle Scholar.

¹¹ Op. cit., p. 709, footnote 6.

¹² Kendall's tau c is a widely-used measure of the correlation between two ordinal-scale variables. It ranges between 0 (which indicates no correlation at all) and ±1.0 (which indicates perfect correlation), and thus is very roughly (not mathematically) equivalent to the more familiar Pearsonian r, which is used in connection with ratio-scale variables. See Blalock, Hubert M., Social Statistics (New York: McGraw-Hill, 1960), pp. 317–324Google Scholar; and Kendall, Maurice G., Rank Correlation Methods (London: C. Griffin, 1955)Google Scholar, chs. 1 and 3.

¹³ Note that a high positive tau for a given table does not mean that respondents represented in that table are concentrated in extreme oneparty areas; it means that the direction of their vote is highly dependent on their political environment. Hence, it is illegitimate to argue that this evidence of community influence is artifactual on the grounds that integrated respondents tend to be Republican and that most counties were Republican in 1952.

¹⁴ Sykes, Gresham, “The Differential Distribution of Community Knowledge,” in Hatt, Paul K. and Reiss, Albert J. Jr., Cities and Society (Glencoe: Free Press, 1957), pp. 711–721Google Scholar. Sykes also found location of work in the local community to be an indicator of “localism,” but we have no information on this for our respondents.

¹⁵ Even clear positive results would be ambiguous since length of residence is also related to membership in community organizations, and, as we shall find later, such membership is clearly linked to community influence.

¹⁶ Another possible indicator of community identification was respondents' agreement or disagreement with the statement that “local elections are often unimportant.” In the 1952 survey the tau for those respondents who disagreed (hypothetioally, the “locals”) was .18 (N = 833) and for those who agreed was .13 (N = 127). In the 1960 survey the comparable figures were .21 (N = 932) and .36 (N = 72). In other words, this item, too, fails to support the hypothesis linking community identification and community influence.

¹⁷ It is not surprising, in terms of the social interaction theory as explicated here, that members of labor unions are less, not more, susceptible to community influence than non-members. We would not expect union membership to involve social interaction with anything like a crosssection of community sentiment. For the 1952 survey tau for union members was .11 (N = 299) compared to a tau of .14 (N = 659) for nonmembers. The comparable figures for 1960 were .11 (N = 302) and .24 (N = 701). If community influence is to a considerable extent mediated through social interaction, then we might expect that minority groups would be less sensitive to the overall community environment, insofar as they are excluded from extensive interaction with much of the population. In 1952 the tau for Protestants was .14 (N = 619) and the tau for non-Protestants was .06 (N = 344). In 1952 the tau for whites was .16 (N = 930), while the tau for non-whites was −.21 (sic) (N=36). These comparisons were confirmed in the replication using the 1960 survey.

¹⁸ Respondents with “mixed” or unknown friendship groups constitute a constant 34%–38% of the sample in all four partisan environments and are excluded from the present analysis. Their inclusion would complicate the analysis, but not alter the conclusions.

¹⁹ Loc. cit.

²⁰ Note that if “homophily” were complete the correlations in this table would be identical with the correlation between respondent's vote and county environment, viz., tau =.18. The fact that the obtained tau for non-members is .24 indicates that even non-members are not able to insulate themselves from the general community environment completely.

²¹ As in any cross-sectional analysis, the direction of causality is not definitely established by the above findings. That is, there may be a tendency for members of the political majority in an area to become more involved in secondary associations, rather than (or in addition to) a tendency for members of secondary associations to become more like the political majority. While this alternative cannot be excluded, its plausibility is reduced by the generally low salience of partisan politics for the average American. It is somewhat difficult to imagine that one's position in the political majority or minority in a locality has much to do with one's decision to become involved in nonpolitical associations in that locality. Moreover, the plausibility of the other alternative (that associations mediate community influence) is enhanced by the pattern of findings linking the secondary and primary group phenomena.

²² Broadly speaking, we have examined three independent variables in this study: party contact, community identification, and social interaction. It may be of interest to note the intercorrelations among these three variables. There is virtually no relationship between party contact and any of the other variables. (tau≤03) There is virtually no relationship between social interaction and intent to remain in the local community (tau = .04). There are moderate positive relationships between social interaction and length of residence, homeownership, and parental status (tau = .09, .12, and .21, respectively). Thus, social interaction is far from perfectly correlated with our indicators of community identification. The overall effect of these findings is to strengthen our principal findings, for if we were to control for the spurious effect of organizational membership, the correlations between community influence and our indicators of community identification would be even weaker than those shown in Table 4.

²³ As examples of this growing literature, see Almond, Gabriel A. and Verba, Sidney, The Civic Culture (Princeton: Princeton University Press, 1963)CrossRefGoogle Scholar; Pye, Lucian and Verba, Sidney, Political Culture and Political Development (Princeton: Princeton University Press, 1965)CrossRefGoogle Scholar; and Lane, Robert E., Political Ideology (New York: Free Press, 1962)Google Scholar. Students of local political systems are paying increasing attention to the importance of local political culture. See Dahl, Robert A., Who Governs (New Haven: Yale University Press, 1961), pp. 311–325Google Scholar, and Agger, Robert E.et al., The Rulers and the Ruled (New York: Wiley, 1964)Google Scholar, passim.

²⁴ Dahl, op. cit., pp. 298 and 172.

²⁵ See Stouffer, Samuel A., Communism, Conformity, and Civil Liberties (Garden City, New York: Doubleday, 1955), pp. 26–57Google Scholar, and McClosky, Herbert, “Consensus and Ideology in American Politics,” this Review, 58 (1964), 361–382Google Scholar. Of course, I do not mean that the processes discussed in the present paper are the sole explanation for these findings about activists' attitudes.

²⁶ Here, as at a number of other points, our findings converge with those of mathematicallyinclined sociologists who have developed and tested formal models of the diffusion of social attitudes and behavior. See Coleman, James S., Introduction to Mathematical Sociology (New York: Free Press, 1964), pp. 336–353, 492–514Google Scholar, and the works cited there.

²⁷ For evidence corroborating these hypotheses, see Greer, Scott and Orleans, Peter, “Political Sociology,” in Faris, Robert E. L. (ed.), Handbook of Modern Sociology (Chicago: Rand McNally, 1964), pp. 815–818Google Scholar, and the works cited there.

²⁸ To be sure, there are probably many other effects of group membership besides this sort of “integration” into community social relations, and some of these other effects might be supportive of democracy even in a community where antidemocratic attitudes were rampant. See Kornhauser, William, The Politics of Mass Society (Glencoe: Free Press, 1959)Google Scholar, passim.