¹ Bertram, Christoph, ed., Defense and Consensus: The Domestic Aspects of Western Security (New York: St. Martin's, 1983)CrossRefGoogle Scholar; Capitanchik, David and Eichenberg, Richard, Defence and Public Opinion (Boston: Routledge and Regan Paul, 1983)Google Scholar; Flynn, Gregory and Rattinger, Hans, eds, The Public and Atlantic Defense (New York: Rowman and Allanheld/London: Croom Helm, 1985).Google Scholar

² ‘Affectivity’ is used in the psychological sense to refer to subjective emotions and feelings. The other psychological components of attitudes are the ‘cognitive’ (objective evaluations) and the ‘behavioural’; see Hovland, Carl I. and Rosenberg, Milton J., Attitude Organization and Change (New Haven, Conn.: Yale University Press, 1960), pp. 1–14.Google Scholar

³ The list of ‘crises’ within the Western Alliance is long. For an early assessment of some of the disputes, see Kissinger, Henry A., The Troubled Partnership: A Re-appraisal of the Atlantic Alliance (Garden City, NY: Anchor Books, 1966)Google Scholar; a more recent analysis is Grosser, Alfred, The Atlantic Alliance (New York: Vintage Books, 1982).Google Scholar

⁴ Cohen, Eliot A., ‘The Long-Term Crisis of the Alliance’, Foreign Affairs, LXI (1982/1983), 325–43CrossRefGoogle Scholar; Hoffmann, Stanley, ‘The Western Alliance: Drift or Harmony’, International Security, VI (1981), 105–25CrossRefGoogle Scholar; Joffe, Josef, ‘European-American Relations: The Enduring Crisis’, Foreign Affairs, LIX (1981), 835–51.CrossRefGoogle Scholar

⁵ Stimson, James A., ‘Belief Systems: Constraint, Complexity, and the 1972 Election’, American Journal of Political Science, LXIX (1975), 393–418.CrossRefGoogle Scholar

⁶ Converse, Philip E., ‘The Nature of Belief Systems in Mass Publics’, in Apter, David, ed., Ideology and Discontent (New York: The Free Press of Glencoe, 1964), pp. 206–61.Google Scholar

⁷ Marcus, George E., Tabb, David and Sullivan, John L., ‘The Application of Individual Differences Scaling to the Measurement of Political Ideologies’, American Journal of Political Science, LXVIII (1974), 405–20, p. 407.CrossRefGoogle Scholar

⁸ There are several contending descriptions for this new American multi-dimensional belief system. A relatively complex multi-dimensional structure with five interpretable dimensions is presented in Bardes, Barbara and Oldendick, Robert, ‘Beyond Internationalism: A Case For Multiple Dimensions in the Structure of Foreign Policy Attitudes’, Social Science Quarterly, LIX (1978), 495–508Google Scholar. American respondents are categorized as internationalists, liberal internationalists and non-internationalists in Mandelbaum, Michael and Schneider, William, ‘The New Internationalism’, in Oye, Kenneth A., Rothchild, Donald and Lieber, Robert J., eds, Eagle Entangled: US Foreign Policy in a Complex World (New York: Longman, 1979), pp. 34–88Google Scholar. American attitudes are described with a four-cell typology comprising the categories of internationalists, accommodationists, hardliners and isolationists in Wittkopf, Eugene R., ‘The Structure of Foreign Policy Attitudes: An Alternative View’, Social Science Quarterly, LXII (1981), 108–23.Google Scholar

⁹ Feld, Werner J. and Wildgen, John R., NATO and the Atlantic Defense: Perception and Illusions (New York: Praeger, 1982).Google Scholar

¹⁰ Adler, Kenneth P. and Wertman, Douglas, ‘Is NATO in Trouble?’ Public Opinion, IV (08-09 1981), 8–13Google Scholar; Capitanchik, and Eichenberg, , Defence and Public Opinion.Google Scholar

¹¹ Flynn, and Rattinger, , The Public and Atlantic Defense.Google Scholar

¹² For a discussion of the advantages and techniques of various scaling methods, see McIver, John P. and Carmines, Edward G., Unidimensional Scaling (Beverly Hills, Calif.: Sage Publications, 1981).CrossRefGoogle Scholar

¹³ The data used in this study were made available by the Inter-University Consortium for Political and Social Research, Ann Arbor, Michigan. Neither the original collectors of the data, the Consortium, nor the sponsors of the study bear any responsibility for the analysis or interpretation presented here.

¹⁴ Original multi-country data sets other than the Eurobarometers are rarely available for the kind of analysis performed here. This is why most studies in this field have relied on a secondary analysis of single items from published opinion polls such as Gallup and Harris. Another difficulty is that the number of foreign policy and security items included in any one survey is usually very limited, thus preventing a meaningful analysis of underlying dimensions. No Eurobarometer survey includes as many foreign policy and security items as Eurobarometer 14. Other potential sources for original machine-readable data sets are those produced by the United States Information Agency (USIA). These are available from the National Archives in Washington, DC, and the Roper Center in New Haven, Connecticut. However, the documentation and code books provided are often incomplete and inaccurate and the data are not ‘cleaned’ adequately. Extensive efforts by the author and others to use these data in their machine-readable form have been largely unsuccessful. In comparison with this rather meagre supply of data on European foreign policy attitudes, analysts of American foreign policy attitudes have available the surveys conducted by the Chicago Council on Foreign Relations. These surveys, conducted periodically since 1974, each contain over 200 questionnaire items specifically on foreign policy issues. Currently, there are no data sets on European attitudes remotely equivalent to this rich source.

¹⁵ The number of response categories varied between items. A range of 1 to 7 was used to recode the numerical values for the responses so that the range of values would be equivalent for all items. For example, items with three responses – (1), (2), and (3) – were receded (1), (4) and (7); and items with four responses – (1), (2), (3) and (4) – were recoded (1), (3), (5) and (7). In each case the higher numbers reflect greater support for Atlantic co-operation. For the factor analysis this recoding is not necessary; however, for the summary scales discussed in the next section it is essential.

¹⁶ Kim, Jao-On and Mueller, Charles W., Introduction to Factor Analysis (Beverly Hills, Calif.: Sage Publications, 1978).CrossRefGoogle Scholar

¹⁷ This article represents one part of a broader research project focusing on these four countries. It is clearly recognized that other interesting cases are not included here. For example, the Netherlands and Belgium should be obvious candidates for inclusion in any future studies of this sort. However, the Eurobarometers are administered only in European Community countries, so Norway and some other Alliance members are automatically excluded from this data source.

¹⁸ Missing data on any questions were automatically recorded to the mean for that item. In this way no respondents were deleted from the analysis. Additional analyses using pairwise and listwise deletion of missing data yielded essentially identical factor solutions.

¹⁹ Eigenvalues are produced by the principal components method and are useful in determining the number of relevant dimensions. This statistic is computed for each of the factors generated by the analysis up to the number of variables entered, and it can be interpreted as the total variance explained by each factor. In a principal components analysis, factors are standardized such that total variance equals the number of variables, and each variable has a variance of 1. For this reason, factors with eigenvalues less than 1 are no better than a single variable, and the normal criterion is to accept as valid conceptual dimensions only those factors with eigenvalues greater than 1. See Norusis, Marija I., SPSSx Advanced Statistics Guide (New York: McGraw-Hill, 1985).Google Scholar

²⁰ Communalities can be interpreted as the proportion of a variable's total variance that is accounted for by all of the factors, and it is derived from the sum of the squared loadings for a variable. These values do not change with rotation. The percentage of variance in a variable accounted for by all of the factors is h ² × 100. Variables with extremely low communalities would be deleted because of their limited contribution to the model. The sum of h ² values × 100 equals the percentage of total variance in all of the variables accounted for by the factors. See Rummel, R. J., Applied Factor Analysis (Evanston, Ill.: Northwestern University Press, 1970).Google Scholar

²¹ Rummel, , Applied Factor Analysis, p. 386.Google Scholar

²² The correlations between the rotated factors in the oblique solution are listed below. These low coefficients support the conclusion for orthogonality.

²³ The figures entered in the factor columns are the ‘loadings’. These measure the degree to which variables are related to the factors. Specifically, they are correlation coefficients between variables and factors. The square of the loading multiplied by 100 equals the percentage of a variable's variance accounted for by the factor. Normally, according to Rummel, in Applied Factor Analysis, p. 139Google Scholar, factors are interpreted as consisting only of those variables with loadings equal to or greater than an absolute value of 0.40; that is, those variables with 16 per cent or more of their variation explained by the factor. Factor loadings meeting this criterion may have been underscored, and they can be used to identify which variables belong to which factors. The communalities and loadings together show at a glance how the data are ordered and explained by the factors.

²⁴ In separate analyses for each national sample, a two-factor solution produced the loadings listed in Appendix II. The international expectations item was not included because of its independent effects indicated in Table 2.

²⁵ Another method would be to use factor scores to construct ‘factor scales’. Factor scores are estimates for respondents on a factor computed by various types of differential weighting of variables based on the factor loadings. However, if the factor analysis is used primarily as a heuristic device for sorting out major clusters of variables, the simpler method of summary scales is preferred. For a comparison of ‘factor scales’ with ‘factor-based scales’, see Kim, Jao-On and Mueller, Charles W., Factor Analysis: Statistical Methods and Practical Issues (Beverly Hills, Calif.: Sage Publications, 1985), pp. 60–73.Google Scholar

²⁶ No effort was made to vary the weight assigned to the items because it has been found that little is to be gained: unweighted and weighted summary scores regularly correlate quite highly, according to McIver, and Carmines, , Unidimensional Scaling, p. 27.Google Scholar

²⁷ Cronbach's alpha coefficient is the preferred estimate of scale reliability according to McIver, and Carmines, , Unidimensional Scaling, p. 29Google Scholar. This statistic provides an estimate of reliability based upon the inter-item correlation matrix. Computer routines that test for reliability with alpha also indicate the effects upon the value of alpha if individual items are deleted from the scale. This item analysis provides another source for checking reliability.

²⁸ For similar decision rules, see Wittkopf, Eugene R. and Maggiotto, Michael A., ‘American Public Attitudes Toward Foreign Policy’, International Studies Quarterly, XXV (1981), 601–31Google Scholar; and Wittkopf, Eugene R. and Maggiotto, Michael A., ‘Elites and Masses: A Comparative Analysis Toward America's World Role’, Journal of Politics, XLV (1983), 303–34.CrossRefGoogle Scholar

²⁹ McIver, and Carmines, , Unidimensional Scaling, pp. 27, 28.Google Scholar

³⁰ McIver, and Carmines, , Unidimensional Scaling, p. 28.Google Scholar

³¹ This is the technique used by Wittkopf, , ‘The Structure of Foreign Policy Attitudes: An Alternative View’Google Scholar, and Wittkopf, and Maggiotto, , ‘American Public Attitudes Toward Foreign Policy’.Google Scholar

³² This again borrows from the techniques used by Wittkopf, , ‘The Structure of Foreign Policy Attitudes: An Alternative View’Google Scholar, and Wittkopf, and Maggiotto, , ‘American Public Attitudes Toward Foreign Policy’.Google Scholar

³³ Capitanchik, and Eichenberg, , Defence and Public Opinion, p. 57.Google Scholar

³⁴ Capitanchik, and Eichenberg, , Defence and Public Opinion, p. 57.Google Scholar

³⁵ Joffe, Josef, ‘The Foreign Policy of the Federal Republic of Germany’, in Macridis, Roy C., ed., Foreign Policy in World Politics, 6th edn (Englewood Cliffs, NJ: Prentice-Hall, 1985), pp. 72–113 at p. 79.Google Scholar

³⁶ Joffe, , ‘The Foreign Policy of the Federal Republic of Germany’, p. 88.Google Scholar

³⁷ Capitanchik, and Eichenberg, , Defence and Public Opinion, p. 56.Google Scholar

³⁸ Putnam, Robert D., ‘Italian Foreign Policy: The Emergent Consensus’, in Penniman, Howard R., ed., Italy at the Polls: The Parliamentary Elections of 1976 (Washington, DC: American Enterprise Institute, 1977), p. 296.Google Scholar

³⁹ Hassner, Pierre, ‘The Political Evolution of Italy and the International Context: A Personal View’, paper prepared for a joint meeting of the Istituto Affari Internazionale and Chatham House, Manziani, 11 1975Google Scholar, cited in Putnam, , ‘Italian Foreign Policy: The Emergent Consensus’, p. 296.Google Scholar

⁴⁰ A note of caution should be raised here. The structure derived from exploratory factor analysis is obviously a function of the items entered. There are many other foreign policy issues related to Atlantic co-operation not available in the data set analysed. As such, the conclusion in regard to the apparent simplicity of the structure of European foreign policy attitudes at the mass level could be exaggerated.

⁴¹ Flynn, and Rattinger, , The Public and Atlantic Defense, pp. 382, 383.Google Scholar

⁴² For a brief comparison of the origins, activities and achievements of the peace movements in several countries, see Kaltefleiter, Werner and Pfaltzgraff, Robert L., eds, The Peace Movements in Europe and the United States (New York: St. Martin's, 1985).Google Scholar

⁴³ Dumont, Joel-François, ‘The Peace Movement in France’Google Scholar, in Kaltefleiter, and Pfaltzraff, , eds, The Peace Movements in Europe and the United States, pp. 132–9.Google Scholar

⁴⁴ Rossi, Sergio A. and Ilari, Virgilio, ‘The Peace Movement in Italy’Google Scholar, in Kaltefleiter, and Pfaltzgraff, , eds, The Peace Movements in Europe and the United States, pp. 140–61.Google Scholar

⁴⁵ Boyd, Peter, ‘The Development of the Peace Movement in Britain’Google Scholar, in Kaltefleiter, and Pfaltzgraff, , eds, The Peace Movements in Europe and the United States, pp. 63–103.Google Scholar

⁴⁶ The final two items, ECPOL and SCTYPOL, each contained two responses that received equal values during recoding. This was necessary because in each case the pair of responses were interpreted as not being amenable to ordinal ranking.