¹ Mill, J. S., On Liberty, cited in Group Dynamics, ed. Cartwright, Dorwin and Zander, Alvin (New York: Harper & Row, 3rd ed., 1968), p. 141 Google Scholar.

² Verba, Sidney, Small Groups and Political Behavior (Princeton: Princeton University Press, 1961), pp. 21ffGoogle Scholar. The classic statements of this position in the voting behavior literature are Lazarsfeld, Paul, Berelson, Bernard, and Gaudet, Hazel, The People's Choice (New York: Duell, Sloan, and Pearce, 1944)Google Scholar and Berelson, Bernard R., Lazarsfeld, Paul F., and McPhee, William N., Voting (Chicago: The University of Chicago Press, 1954)Google Scholar.

³ Kornhauser, William, The Politics of Mass Society (New York: The Free Press, 1959)Google Scholar.

⁴ Putnam, Robert D., “Political Attitudes and the Local Community,” American Political Science Review, 60 (09, 1966), 640–654 CrossRefGoogle Scholar.

⁵ Langton, Kenneth P. and Jennings, M. Kent, “Mothers Versus Fathers in the Formation of Political Orientations,” in Political Socialization, ed. Langton, Kenneth P. (New York: Oxford University Press, 1969), pp. 67–69 Google Scholar.

⁶ Middleton, Russell and Putney, Snell, “Political Expression of Adolescent Rebellion,” American Journal of Sociology, 68 (03 1963), 527–535 CrossRefGoogle Scholar. McClosky, Herbert and Dahlgren, Harold report similar findings in their “Primary Group Influence on Party Loyalty,” American Political Science Review, 53 (09 1959), 757–776 CrossRefGoogle Scholar.

⁷ The sample was designed by and the data collected under the direction of Donald E. Stokes and Warren E. Miller. I served as Assistant Study Director. The data were collected through the Detroit Area Study of the University of Michigan.

⁸ The universe of work groups was stratified by company, shift, political activity level of local union leaders, and ability of the work group to talk informally during the working day. The work groups were then chosen randomly to balance the stratification variables.

⁹ This is similar to Landecker's definition of group cohesion as “communicative integration.” Landecker, Werner S., “Types of Integration and Their Measurement,” in The Language of Social Research, ed. Lazarsfeld, Paul F. and Rosenberg, Morris, (Glencoe: The Free Press, 1955), pp. 19–27 Google Scholar.

¹⁰ Some measurement error results from the fact that response rates were not 100 per cent for any of the work groups. This meant that some, respondents gave one or more of their choices to individuals with whom we did not secure interviews, so that reciprocity was precluded. For these respondents, the number of reciprocated choices may underestimate the number of friends they actually had. An additional possible source of difficulty is that the groups varied in size. Because respondents were asked to choose five friends from the list of persons in their work group, the measure appears to favor individuals in smaller-sized groups, where the likelihood of reciprocation is greater. A number of corrections to the raw score were attempted to adjust for this possibility, but none of these measures had as good validity as the raw score itself. And, as reported in Table ID, the anticipated difficulty of a negative correlation between the raw number of reciprocated choices and group size did not materialize. This fortunate result appears to have been caused by an interesting “natural” correction mechanism in the sociometric question, whereby people in smaller groups tended to give fewer choices than we requested, while those in larger groups insisted on giving more (all choices were coded). Thus, the probability of a reciprocated choice in the different sized groups was substantially equalized by this phenomenon.

¹¹ Criswell, Joan, “The Measurement of Group Integration,” Sociometry, 10 (08 1947), 259–267 CrossRefGoogle Scholar. See footnote 10 for further discussion of the relationship between group size and number of reciprocated choices.

¹² For a general discussion of the conditions under which primary groups support or oppose the larger organization of which they are a part, see Verba, , Small Groups and Political Behavior, 54–60 Google Scholar.

¹³ Converse, Philip E., “The Nature of Belief Systems in Mass Publics,” in Ideology and Discontent, ed. Apter, David, (New York: The Free Press, 1964), pp. 238–241 Google Scholar.

¹⁴ Smith, M. Brewster, Bruner, Jerome S., and White, Robert W., Opinions and Personality (New York: John Wiley and Sons, Inc., 1956), pp. 41–42 Google Scholar.

¹⁵ Ibid., p. 43.

¹⁶ Rosenberg, Morris, “The Meaning of Politics in Mass Society,” Public Opinion Quarterly, 15 (Spring, 1951) 6–7 and 8 CrossRefGoogle Scholar. Rosenberg credits Frotnm, Erich, in his Escape from Freedom (New York: Rinehart, 1941)Google Scholar, with identification of this last source of gratification.

¹⁷ Putnam, , “Political Attitudes and the Local Community,” 648–649 Google Scholar.

¹⁸ Pepitone, A. and Wilpizeski, C., “Some Consequences of Experimental Rejection,” Journal of Abnormal and Social Psychology, 60 (05 1960), 359–364 CrossRefGoogle ScholarPubMed. See also Schachter, Stanley, The Psychology of Affiliation (Stanford: Stanford University Press, 1959)Google Scholar. Schachter stimulated anxiety by threat of electric shock. Further experimental research supports the positive relationship between threat and desire for affiliation using both economic and social threats. See Mulder, Mauk and Stemerding, Ad, “Threat, Attraction to Group, and Need for Strong Leadership,” Human Relations, 16 (11 1963), 317–334 CrossRefGoogle Scholar. Notable examples of the manipulative use of this relationship are reported by Franz Schurman and Edgar Schein. Schurman points out that the criticism program in the Chinese Communist study groups, designed to identify individual deviance from group norms, evokes great anxiety and insecurity, and sets up a need for reintegration with the group (“Organization and Response in Communist China,” Annals, 321 [01 1959], 51–66 Google Scholar). From his studies of prisoners in the Korean War, Schein (“Interpersonal Communication, Group Solidarity, and Social Influence,” Sociometry, 23 [06 1960], 148–160 Google Scholar) argues that lack of interpersonal communication induces strong motives for reintegration and search behavior designed to re-establish communication through new social relationships. The Chinese Communists used this tendency to encourage prisoners, who were denied communication with each other, to develop relationships with the Communists or with cooperative prisoners. The type of interpersonal isolation implied by the present findings obviously is hardly comparable to the extreme, extended, and enforced isolation endured by the prisoners of war. Nevertheless, in relative scale, the motive for reintegration may also be present among political deviants who find few people in their social environments with whom to share political communication. In a less extreme situation, Alexander and Campbell found that lack of parental support for an adolescent's attitudes is associated with a stronger network of friendships among children. They suggest that the lack of support from parents leads to an imbalance which is resolved through greater support of friends. Alexander, C. Norman Jr. and Campbell, Ernest Q., “Balance Forces and Environmental Effects: Factors Influencing the Cohesiveness of Adolescent Drinking Groups,” Social Forces, 46 (03 1968), 367–374 CrossRefGoogle Scholar. In the search for social support, the friendship group of two persons may be especially important, for the dyad is the only size group in which supportive interactions exceed aggressive interactions. Zimet, Carl N. and Schneider, Carol, “Effects of Group Size on Interaction in Small Groups,” The Journal of Social Psychology, 77 (1969), 177–187 CrossRefGoogle Scholar.

¹⁹ Festinger, Leon, “A Theory of Social Comparison Processes,” Human Relations, 7 (1954), 117–140 CrossRefGoogle Scholar.

²⁰ Festinger, Leon and Aronson, Elliot, “Arousal and Reduction of Dissonance in Social Contexts,” in Cartwright, and Zander, , Group Dynamics, p. 130 Google Scholar.

²¹ Some research suggests that deviants also reject conformers. Norrison, W. F. and Corment, D. W., “Participation and Opinion Change in Two Person Groups as Related to Amount of Peer Group Support,” British Journal of Social and Clinical Psychology, 7 (09 1968), 176–183 CrossRefGoogle ScholarPubMed.

²² For a discussion of Durkheim's methods of subgroup comparison as a type of replication, see Selvin, Hanan C., “Durkheim's Suicide: Further Thoughts on a Methodological Classic,” in Emile Durkheim, ed. Nisbet, Robert A. (New York: Prentice-Hall, 1965), 113–136 Google Scholar. For a discussion of a variety of statistical procedures for “pseudoreplication” of findings on the same data set, see Finifter, Bernard M., “The Generation of Confidence: Evaluating Research Findings by Random Subsample Replication,” in Sociological Methodology 1972, ed. Costner, Herbert L., (San Francisco: Jossey-Bass, 1972), 112–175 Google Scholar. Since the basic purpose of significance tests is to suggest the degree of reliability of a finding, Finifter argues that replications serve this purpose without making distributional and other assumptions. Following this logic and because of the nature of the present sample, I have refrained from using significance tests except in the section on the composition of the friendship groups, below.

²³ For a discussion of “systematic” and other types of replication, see Sidman, Murray, Tactics of Scientific Research: Evaluating Experimental Data in Psychology (New York: Basic Books, 1960)Google Scholar.

²⁴ Even after controls for ten life-situation variables were introduced (representing region, size of place of residence, and social class), nonsouthern blacks were almost 12 per cent more Democratic in their voting behavior than were comparable whites. Campbell, Angus, Converse, Phillip E., Miller, Warren E., and Stokes, Donald E., The American Voter (New York: John Wiley and Sons, Inc., 1960), pp. 300–306 Google Scholar.

²⁵ Kelley, Stanley Jr., “The 1960 Presidential Election,” in American Government Annual, 1961–1962, ed. Hinderaker, Ivan (New York: Holt, Rinehart and Winston, Inc.), pp. 57–75 Google Scholar.

²⁸ Campbell, et al., The American Voter, pp. 298–306 Google Scholar.

²⁷ Converse, Philip E., Campbell, Angus, Miller, Warren E., and Stokes, Donald E., “Stability and Change in 1960: A Reinstating Election,” The American Political Science Review, 55 (06 1961), 273 CrossRefGoogle Scholar.

²⁸ Technically, this is a “specification” analysis, in which the focus is on identifying the conditions under which the initial relationship is weakened or strengthened. The reader should note, however, that in controlling for the specification variables we are also in effect checking on the possibility of a spurious relationship resulting from any association between the control variables and the independent variables. That the resulting relationships become both stronger and weaker in a systematic pattern suggests that the spurious relationship type of interpretation is inappropriate. An analysis which presented only one partial correlation for each control variable (interpretable as an average relationship in the subcategories of the control variable) would obscure this pattern.

²⁹ Berelson, et al., Voting, p. 126 Google Scholar. Emphasis added.

³⁰ Briefly, normative reference groups set standards which the individual is motivated to follow; comparative reference groups are used by the individual to evaluate himself and others. The same group may also perform both functions. See Kelley, Harold H., “Two Functions of Reference Groups,” in Readings in Reference Group Theory and Research, ed. Hyman, Herbert H. and Singer, Eleanor, (New York: The Free Press, 1968), pp. 77–83 Google Scholar.

³¹ Festinger's hypothesis, which has been supported by experimental evidence, is that dissonance as a function of disagreement with a group increases with the importance of the group to the individual. Festinger, and Aronson, , “Arousal and Reduction of Dissonance,” p. 130 Google Scholar.

³² The subgroup size difficulties are amply illustrated in the impossibility of doing a within-group analysis of black Republicans. There were only three of these and all were highly identified with their racial group and perceived it to have a Democratic political norm.

³³ Previous analyses have shown that group identification increases compliance with group norms. See, for example, Campbell, et al., The American Voter, pp. 295–332 Google Scholar. The present analysis in a sense focuses on one aspect of the unexplained variance in this relationship, i.e., what is the effect on high identifiers of not conforming?

³⁴ Because the same reference group questions are used for both of these social categories, they are combined in the analysis. For the total sample, the gamma between income and education is .60.

³⁵ I am grateful to Leo Katz, Director of the Statistics Laboratory at Michigan State University, for advice on this analysis.

³⁶ The pooled chi² value is simply the sum of the separate chi² tests and the sum of the degrees of freedom, which is then evaluated in the usual way. “In effect, we are saying that if a relationship comes out roughly the same each time but the probabilities of the separate results are each greater than .05, we still may ask ourselves how likely such a combination of outcomes would be if there were no relationship in any of the … tables.” Blalock, H. M. Jr., Social Statistics, (New York: McGraw-Hill, 2nd ed. 1972), pp. 309–310 Google Scholar. In the present case, 8 of the 18 individual relationships were also separately significant. Chi square is, of course, highly sensitive to N, and the Ns for these analyses are rather small. In fact, the use of chi square in the individual work groups is technically inappropriate because the expected values for dyad types involving Republicans and Independents frequently fall below the generally accepted minimum. Nor does any combination of categories that is theoretically meaningful solve the problem. The results should therefore be interpreted in a tentative and exploratory spirit.

³⁷ Some work group tables had fewer than four degrees of freedom because they had no or only one Independent, thus eliminating the II and/or ID, IR lines.

³⁸ Lone Republicans in the remaining groups had slightly fewer friends than Republicans in these groups where like-minded partisans were available (although still more than Democrats); they demonstrated their partisan selectivity by disproportionate friendships with Independents over Democrats.

³⁹ Cartwright and Zander, Group Dynamics, argue that “public behavior is more likely to be conforming than private behavior” because of the greater likelihood of sanctions being invoked when deviance becomes known (p. 145). Some Wallace campaign activists, for example, reported in the media that during the 1972 presidential primaries there were numerous cases of individuals who covertly expressed support for Wallace but who could not be induced to participate in his campaign until they were persuaded that their “friends and neighbors” were for him too.

⁴⁰ Yinger, J. Milton, “Contraculture and Subculture,” American Sociological Review, 25 (10 1960), 625–635 CrossRefGoogle Scholar.

⁴¹ Golembiewski, Robert T., Welsh, William A., and Crotty, William J., A Methodological Primer for Political Scientists (Chicago: Rand McNally and Co., 1969), p. 78, p. 81 Google Scholar.

⁴² See, for example, Kornhauser, The Politics of Mass Society; also Verba, , Small Groups and Political Behavior, pp. 52–54 Google Scholar.

⁴³ See, for example, Pollis, Nicholas P. and Cammalleri, Joseph A., “Social Conditions and Differential Resistance to Majority Pressure,” Journal of Psychology, 70, 1 (09 1968), 69–76 CrossRefGoogle ScholarPubMed; and Allen, Vernon L. and Levine, John M., “Social Support, Dissent and Conformity,” Sociometry, 31, 2 (06 1968), 138–149 CrossRefGoogle ScholarPubMed.

⁴⁴ While the experimenter's ability to control and manipulate variables is enviable, one must not forget his concern that the artificiality of the laboratory situation may severely limit the generalizability of findings. See, for example, Blake, Robert R. and Mouton, Jane S., “Conformity, Resistance and Conversion,” in Conformity and Deviation, ed. Berg, Irwin A. and Bass, Bernard M. (New York: Harper and Brothers, 1961), esp. pp. 5–7 and 24–25 Google Scholar. Replication of experimental findings in real-life situations is thus especially necessary.

⁴⁵ Milgram, Stanley, “Liberating Effects of Group Pressure,” Journal of Personality and Social Psychology, 1, 2 (02 1965), 127–134 CrossRefGoogle ScholarPubMed.

⁴⁶ Blau, Peter M., “Patterns of Deviation in Work Groups,” Sociometry, 23, 3 (09 1960), 245–261 CrossRefGoogle Scholar.