page 308 note 1. Robins, L. J., ‘The Labour Party and the European Economic Community 1961–1971’, (Unpublished Ph.D. thesis, Southampton University, 1974)Google Scholar establishes the context and analyses the composition and activities of Labour's internal factions from the standpoint of party management. The results presented here are a more extensive statistical analysis of a survey originally undertaken as part of this study.

page 308 note 2. The early Snyder-Bruck-Sapin decision-making model (Snyder, R. C., Bruck, H. W. and Sapin, B., Foreign Policy Decision-Making (New York, 1962)Google Scholar) emphasized the psychological and social processes operating in the internal and external setting which are involved in the making offoreign policy. Other basic texts and readers which embrace, not always explicitly, similar analytical frameworks include Goplin, W. D., Introduction to International (Chicago, 1971)Google Scholar; Edwards, D. V., International Political Analysis (New York, 1969)Google Scholar; Frankel, J., The Making of Foreign Policy (London, 1963)Google Scholar; Olson, W. G. and Sondermann, F. A., The Theory and Practice of International Relations (Englewood Cliffs, 1960)Google Scholar and Holsti, K. J., national Politics: A Frameworkfor Analysis (Englewood Cliffs, 1967)Google Scholar.

page 308 note 3. An influential paper examining the relation of decision-makers' perceptions of reality to state behaviour is Boulding's, K. E.‘National Images and International Systems’, Journal ofConflict Resolution, iii (1959), pp. 120–131CrossRefGoogle Scholar. Equally important empirical studies in this area are Holsti, O. R., ‘The Belief System and National Images: a case Study’, Journal of Conflict Resolution, vi (1962), pp. 244–52CrossRefGoogle Scholar, and Zinnes, D. A., ‘The Expression and Perception of Hostility in Pre-war Crisis: 1914’ in Singer, J. D. (ed.), Quantitative International Politics: Insight and Evidence (New York, 1968)Google Scholar. Other names associated with the study of “the human dimension of international relations” include Kelman, Singer, Jervis, Klineberg, Osgood, North and De Rivera. The approach is placed into a broader perspective byJames Rosenau's discussion of “pre-theoretical” research, and his own investigation of the relative potency of individual and role variables in the behaviour of U.S. Senators, reprinted in The Scientific Study of Foreign Policy (London, 1971) pp. 95–98Google Scholar, 108–9, 151–193.

page 309 note 1. Robert Putnam's pioneering study of “elite political culture” nevertheless relies upon a distributional analysis of answers obtained by a questionnaire of more-or-less manifest content, which seems likely to attract stereotyped responses. See Putnam, R. D., The Beliefs of Politicians (Yale, 1973)Google Scholar. The criticism of non-structural approaches is well expressed in a recent literature survey by Ashford, D. E., Ideology and Participation (London, 1972), p. 21Google Scholar, Singer, J. D., ‘Content Analysis of Elite Articulations’, Journal of Conflict Resolution, viii. (1964), pp. 424–85CrossRefGoogle Scholar provides an excellent example of the sophisticated comparative measurement of perceptions by content analysis; thirty-five dimensions in the images of United States and Soviet foreign policy elites are given quantitative expression, but the data do not facilitate the type of investigation we believe important.

page 310 note 1. Beer, S. H., Modern British Politics: a Study of Parties and Pressure Groups (London, 1965), p. 219Google Scholar.

page 310 note 2. McKenzie, R. T., British Political Parties: the Distribution of Power within the Conservative and Labour Parties (London, 1964), p. 628Google Scholar.

page 311 note 1. Gordon, M. R., Conflict and Consensus in Labour's Foreign Policy 1914–1965 (Stanford, 1969), p. 9Google Scholar, and Finer, S. E., Berrington, H. B. and Bartholomew, D. J., Backbench Opinion in the House of Commons 1955-59 (London, 1961)Google Scholar.

page 311 note 2. The theoretical background to the Semantic Differential is summarised in Osgood, C. E., Suci, G. J. and Tannenbaum, P. H., The Measurement of Meaning (Urbana, 1957)Google Scholar and Snider, J. G. and Osgood, G. E. (eds.), Semantic Differential Technique: a Sourcebook (Chicago, 1969)Google Scholar. For criticisms and limitations of the semantic differential technique see Weinreich, U., ‘Travels through Semantic Space’, Word, xiv, (1958), pp. 346–66CrossRefGoogle Scholar and Warr, P. B. and Knapper, C., The Perception of People and Events (London, 1968)Google Scholar.

page 311 note 3. See Osgood et al., op. cil. (1957), pp. 31–75.

page 312 note 1. The questionnaire consisted of seven-point Likert scales in diagrammatic form, from which responses were subsequently coded numerically as integers in the range —3 to + 3. These scores formed the basis for the calculation of matrices of distances between concept-profiles, and for the separate analysis of evaluations to be described later. The term “new Commonwealth” was qualified in the questionnaire as “Nigeria, Zambia, Singapore, etc.” and the “old Commonwealth” as “Australia, New Zealand, Canada”. A total of two M.P.s and six activists did not receive a sheet relating to the General and Municipal Workers Union, and were given neutral profiles for this concept, but there was no apparent effect on the overall results so no correction is mentioned in what follows. The instructions on the questionnaire were based on those given by Osgood et al., op. cit. (1957) with their term ‘test’ replaced by the word ‘survey’. The questionnaire was reproduced on foolscap pages, and the instructions and sample page are reproduced in Fig. 3 in a reduced form.

page 314 note 1. See, for example, Protho, E. T., and Keehn, J. D., ‘Stereotypes and Semantic Space’. Journal ofSocial Psychology, xiv (1957), 197–209CrossRefGoogle Scholar, and Morrison, A. and McIntyre, D., ‘Attitudes of Students towards International Affairs’, British Journal of Social and Clinical Psychology, v (1966), pp. 17–23.CrossRefGoogle Scholar

page 314 note 2. All questionnaires were completed in a ten week period in the Spring of 1971. Of the M.P.s who completed questionnaires, thirteen were signatories of the anti-Common Market Early Day Motion circulated in January 1971, and nineteen signed a pro-Market statement in The Guardian newspaper on 11 May 1971. This basis of classification clearly fails to reflect the extent of tactical signature by those who joined the dominant party view in opposing British membership, while some of those who supported The Guardian statement at the same time opposed membership on the terms which seemed likely to emerge from the negotiations. With the exception of nine members of identifiable pressure groups, the activists could not be classified in this way, although sixteen appeared to oppose and thirteen support British membership of the EE.

Kitzinger has correlated Labour behaviour in the historic Parliamentary vote on EEC membership of 28 October, 1971 with personal attributes in an investigation of European attitudes. (Kitzinger, U., ‘An Anatomy of Rebellion’, New Society, 4 May, 1972)Google Scholar The Labour rebels were compared with those who conformed to the party whip but, of course, amongst the conformers were both those with hostile attitudes towards Europe and those who had never voluntarily revealed their feelings, a considerable number of whom were probably uncommitted on the issue. Also a number of M,P.s of all opinions came under increasing pressure from constituency, trade union and regional organisations as the crucial vote approached.

1. From the raw questionnaire responses, Euclidean distances were computed between all pairs of concepts to produce, for each individual respondent, a perceptual map consisting of seventy-eight distances.

for i = 1, 13; j = 1, 13; i ↗ j ; where the terms xik and xjk are the scored responses for each pair of concepts on the k-th semantic scale of those listed in Fig. 1. This is used as an index of similarity between response profiles, and does not appear in itself to have imposed an unwarranted degree of homogeneity on the data. While not the only possibility, the metric which is thus imposed seems to be consistent with the conceptualisation originally discussed. Given the number of respondents available, it was not possible to test rigorously the assumptions necessary for its application, but the contributions of the various semantic scales to each concept profile were separately examined as a further check against the use of redundant scales. No scale contributed less than 3 per cent on average to the calculations of distances between concept pairs for individual respondents, i.e.:

for k = 1, … 20.

Again, the scales introduced in the present study contributed as much as those previously tested by Osgood.

On this measure ofprofile similarity cf Bijnen, E. J., Cluster Analysis (Tilburg, 1973), pp. 10–22CrossRefGoogle Scholar. If the distance measure is re-scaled by division by the number of items of measurement n, so that

when n is reasonably large the expression and 1 – (2/n) are respectively the approximate standard error and expected value of the coefficient djk. Sokal, R. R. and Sneath, P. H. A., Principles of Numerical Taxomony (San Francisco, 1963), pp. 125–8Google Scholar, 147–50.

page 319 note 1. Likelihood ratio tests were made of homogeneity of group multivariate dispersions. Hope, Cf K., Methods of Multivariate Analysis (London, 1968), pp. 23–32Google Scholar. Singularity of some of the dispersion matrices (usually involving small numbers of individual cases) prevented complete analysis in every case. Given the large number of variables and a relatively small number of respondents available, only differences with a random probability of less than o01, after application of appropriate small sample corrections, were considered significant. In fact, in none of the results discussed were group dispersions found to differ significantly. Unless otherwise stated, all chi-squared values in this paper include a small sample correction. Departures from grand variable means indicated in multivariate analysis of variance were also considerably smaller between a priori groupings than between groups obtained by cluster analysis. The results of these comparisons are summarized in Table 1.

page 319 note 2. The computer algorithm used seeks to minimize the total sum of squares of distances between observations and the centres of gravity of the clusters to which they are assigned, i.e. to minimize the expression.

where j indexes the number of variables, i indexes the number of individual observations (vectors a) and k indexes the number of cluster centroids (vectors c). The procedure used allows reallocation of cases for each partition to find a local minimum over k clusters. The method begins from an initial allocation, here based on the thirty-two M.P.s, and is not fully enumerative, and therefore possibly sub-optional when compared with a fully enumerative equivalent, but is computationally much quicker and seems appropriate to the present problem. The point was not to identify any one ‘best’ solution, but to examine the relative homogeneity of perceptions among different possible partitions, subject to the above criterion. Press, Cf S. J., Applied Multivariate Analysis (New York, 1972), pp. 408–17Google Scholar, and Everitt, B., Cluster Analysis (London, 1974), pp. 24–6, 62–4.Google Scholar

page 321 note 1. In view of the relatively small number of respondents available, it was not feasible to examine exhaustively the sources of variation in cluster solutions, but separate analyses were made as follows:

A: using the full set of twenty semantic scales listed in Fig. 1, including all respondents;

B: using the full set of twenty semantic scales, M.P.s only;

C: using only those scales developed by Osgood et al., op. cit. including all respondents; and

D: using only Osgood's scales, including M.P.s only.

page 323 note 1. The smallest inter-correlations involved the small groups giving unusually extreme answers which are isolated when solutions in three rather than two groups are examined. Given the similarity in the allocation of individuals to groups using either set of scales considered here, it seems likely that the use of different scales contribute less to the heterogeneity of clusters than the relative size of the clusters concerned.

page 324 note 1. Attitudes refer to an individual's evaluative beliefs connected with an object or concept, that is, whether the object is attributed with favourable or unfavourable qualities. In tests concerning the validity of the Semantic Differential as a measure of attitude, Osgood et al., op. cit. (1957) found that the evaluative scales “good-bad”, “fair-unfair” and “valuable-worthless” compared so well with results from a Gutmann scale that both were “measuring the same thing to a considerable degree”, (p. 194) In a study of attitudes towards Arab Unity (Diab, L. N., ‘Studies in Social Attitudes: Attitude Assessment through the Semantic Differential Technique’. The Journal of Social Psychology, lxvii (1965), pp. 303–14)CrossRefGoogle Scholar the “good-bad” scale was used as representative of the evaluative factor. In this study we followed Osgood et al., op. cit. in utilizing the “fair-unfair” scale as identifiable with the evaluative factor in the judgement of political objects. It is a purely evaluative scale that records differing degrees of liking for objects without revealing differences in meaning between objects.

page 325 note 1. The two cluster solution was again the first to eliminate single member clusters.

page 326 note 1. Haseler, S., The Gaitskellites: Revisionism in the British Labour Party 1951–64 (London 1969), p. 227CrossRefGoogle Scholar.

page 327 note 1. For a multivariate analysis of variance for the concept distance scales, L = .0096, X² = 92.83 (d.f. = 78).

page 329 note 1. The United Kingdom and the European Communities (Labour Party, 1971).Google Scholar