¹ See Sullivan, John L., Piereson, James and Marcus, George E., ‘An Alternative Conceptualization of Political Tolerance: Illusory Increases, 1950s–1970s’, American Political Science Review, 73 (1979), 781–94CrossRefGoogle Scholar; and Sullivan, John L., Piereson, James and Marcus, George E., Political Tolerance and American Democracy (Chicago, Ill.: University of Chicago Press, 1982).Google Scholar

² See Nunn, Clyde A., Crockett, Harry J. Jr, and Williams, J. Allen Jr, Tolerance for Nonconformity (San Francisco: Jossey Bass, 1978), p. 43.Google Scholar

³ See Sullivan, et al. , ‘An Alternative Conceptualization of Political Tolerance’.Google Scholar

⁴ See, for example, Lipset, Seymour and Raab, Earl, The Politics of Unreason: Right-Winy Extremism in America, 1790–1970 (New York: Harper & Row Publishers, 1970).Google Scholar

⁵ For example, Nunn, et al. , Tolerance for Nonconformity.Google Scholar

⁶ Sullivan, et al. , Political Tolerance and American Democracy, p. 186 (italics in original).Google Scholar

⁷ See, for example, Stouffer, Samuel, Communism, Conformity and Civil Liberties (New York: Doubleday, 1955)Google Scholar; also Nunn, et al. , Tolerance for Nonconformity.Google Scholar

⁸ Sullivan, et al. , Political Tolerance and American Democracy, pp. 114ff.Google Scholar

⁹ See, for example, Prothro, James W. and Grigg, Charles W., ‘Fundamental Principles of Democracy: Bases of Agreement and Disagreement’, Journal of Politics, 22 (1960), 276–94CrossRefGoogle Scholar; and McClosky, Herbert, ‘Consensus and Ideology in American Politics’, American Political Science Review, 58 (1964), 361–82.CrossRefGoogle Scholar

¹⁰ See Bachrach, Peter, The Theory of Democratic Elitism (Boston: Little, Brown, 1967).Google Scholar

¹¹ See Sullivan, , ‘An Alternative Conceptualization of Political Tolerance’.Google Scholar

¹² The General Social Survey provides exemplary opportunities for cross-validation. We have taken advantage of them, examining two other years, 1980 and 1984, and focusing on selected comparisons, the results of which always sustain the findings reported from 1977. In particular, the 1977 GSS does not include a Stouffer item series on socialists, but the 1980 and 1984 surveys, among others, do. Analyses of these data, not reported here, confirm principled consistency holds for attitudes to socialists, too.

¹³ Because the tolerance measures are dichotomous, the standard errors and chi-square associated with the maximum likelihood estimates are unreliable. The estimates themselves are still reliable, as the data are not highly skewed (see Jöreskog, Karl G. and Sorborn, Dag, LISREL VI: Analysis of Linear Structural Relationships by the Method of Maximum Likelihood (Moorville, Indiana: Scientific Software, 1984), Chap. IV).Google Scholar

¹⁴ Herson, Lawrence J. R. and Hofstetter, C. R., ‘Tolerance, Consensus and the Democratic Creed: A Contextual Exploration’, Journal of Politics, 37 (1975), 1007–32.Google Scholar

¹⁵ McCutcheon, Allan L., ‘A Latent Class Analysis of Tolerance for Nonconformity in the American Public’, Public Opinion Quarterly, 49 (1985), 474–88.CrossRefGoogle Scholar

¹⁶ It is worth remarking that the ratios for liberals are higher than for conservatives for several group pairs (e.g., atheists and communists and militarists and homosexuals), a point we shall return to and build on in our analysis of principled consistency.

¹⁷ Duncan, Otis Dudley, ‘Indicators of Sex Typing’, American Journal of Sociology, 85 (1979), 251–60.CrossRefGoogle Scholar

¹⁸ It is worth considering the logical fitness of latent class analysis employed by McCutcheon, in ‘A Latent Class Analysis’Google Scholar, the alternative response model to Duncan's. The details of McCutcheon's latent class analysis are puzzling – Table 2, for instance, fails to sum as it should – but the more vital point to grasp is that latent class analysis is a methodologically inappropriate technique for the assessment of consistency. It is inappropriate because it presumes that principled consistency accounts for all of the correlations. In contrast, the Duncan technique discounts the zero-order correlations on behalf of principled consistency, taking as principled only that consistency that cannot be accounted for by the pairwise correlations among items. In short, we have chosen the technique that makes it hardest, not easiest, to confirm our principal hypothesis.

¹⁹ See Sniderman, Paul M., Personality and Democratic Politics (Berkeley: University of California Press, 1975)Google Scholar; and McCloskey, Herbert and Brill, Alida, Dimensions of Tolerance (New York: Russell Sage, 1983).Google Scholar

²⁰ See, for example, Rosch, Eleanor, ‘Principles of Categorization’, in Rosch, Eleanor and Lloyd, Barbara B., eds, Cognition and Categorization (Hillside, NJ: Lawrence Erlbaum, 1978).Google Scholar

²¹ See Quine, W. V., ‘Natural Kinds’, in his Ontological Relativity and Other Essays (New York: Columbia University Press, 1969), p. 122.Google Scholar

²² To be specific: 70 per cent of those who strongly disagree with the proposition that white people have a right to keep blacks out of their neighbourhoods if they want to, and that blacks should respect that right, agree that a person who believes that blacks are genetically inferior should be allowed to speak; by comparison, only 50 per cent of those who strongly agree that whites should, as a matter of principle, be entitled to segregate their neighbourhoods believe that a spokesman for the inherent inferiority of blacks should be allowed to present his point of view. This finding that racists are less likely than non-racists to support the rights of a racist group certainly runs counter to the suggestion of ‘opportunistic tolerance’ implicit in Sullivan et al.'s argument. On the other hand, the data on atheists at first appears to run counter to the finding on racists. In brief: 95 per cent of persons with no religious preference endorse free speech for atheists compared with 61 per cent of persons stating a religious preference. Similar differences, moreover, hold for questions on teaching and books. On fuller examination, however, these results fail to support the emotivist thesis; for atheists are more likely to endorse the civil liberties of all groups by margins that are only slightly smaller than the margin of their greater support for atheists' civil liberties.

²³ See Sullivan, et al. , ‘An Alternative Conceptualization of Political Tolerance’.Google Scholar

²⁴ See Goodman, Leo A., ‘The Analysis of Cross-Classified Data’, Journal of the American Statistical Association, 63 (1968), 1091–131Google Scholar; and Fienberg, Stephen E., The Analysis of Cross-Classified Data, 2nd edn (Cambridge, Mass.: MIT Press, 1980).Google Scholar

²⁵ Fienberg, , The Analysis of Cross-Classified Data, pp. 148–52.Google Scholar

²⁶ See Goodman, Leo A., ‘Multivariate Analysis of Qualitative Data’, Journal of the American Statistical Association, 65 (1970), 226–86CrossRefGoogle Scholar and Goodman, Leo A., ‘A General Model for the Analysis of Surveys’, American Journal of Sociology, 77 (1972), 1035–86.CrossRefGoogle Scholar

²⁷ See Raftery, Adrian, ‘Choosing Models for Cross-Classifications’, American Sociological Review, 51 (1986), 145–6.CrossRefGoogle Scholar