At the end of his entertaining and thoughtful review [this Journal, iv (1974), 345–69, p. 362], Professor Berrington writes ‘if it is lonely at the top it is because it is the lonely who seek to climb’. But this is to miss a point that undermines the significance of Mrs Iremonger's thesis. It is indeed lonely at the top, and men who have already coped with loneliness are peculiarly fitted to bear the burdens of the Prince. Nor is there a contradiction between the public aloofness of prime ministers and their domestic felicity: those who are surrounded by a close and affectionate family and supported by a devoted wife can afford to do without the gratification of friendship in public life. They make good butchers. This suggests that the ideal characteristics required of prime ministers are not those put forward by Mrs Iremonger and apparently accepted by Professor Berrington. I recall, in loose translation, the words of a chronicler on King Stephen: ‘He was a mild man and good and did no justice’. The world has need of its bastards.

In a recent article on the comparative method in the study of politics, 1 Arend Lijphart discusses the place of the case study and suggests that there are six types of case study, four of which may be characterized as theoretical, two as atheoretical. The first of the theoretical types is the hypothesis-generating case study. Of studies of this type, Lijphart says: ‘They start out with a more or less vague notion of possible hypotheses, and attempt to formulate definite hypotheses to be tested subsequently among a large number of cases.2 The study upon which this paper is based satisfies entirely the first part of this description in that it began with a ‘more or less vague notion of possible hypotheses’ and it is hoped that the more definite hypotheses which were generated will be worthy of the large-scale testing that will eventually lead to the construction of a general theory of non-partisan civic government in Canada.

In his review article on policy analysis [Hugh Heclo, ‘Policy Analysis’, this Journal, II (1972), 83–108], Heclo describes public policy as ‘necessarily at the heart of the political scientist’s concern’. But because public policy ‘is concerned with metachoices’ – ‘choices as to how others shall make choices in whatever sphere public authority is intervening’ – ‘the study of policy therefore necessarily straddles a number of previously distinct academic disciplines’.‘However, I wonder to what extent political scientists are willing to take part in a ‘coherent “interdiscipline” of policy studies’. Heclo quotes Vernon Van Dyke’s warning that the prospect of this for political scientists is ‘appalling’.2 If we examine firstly what political science has to offer in policy analysis and secondly what policy analysis itself has to offer, I feel that Van Dyke will suddenly find himself with massive support.

Ivor Crewe begins his paper on the affluent worker research by giving clear expression to a number of objections which may be brought against that work. He then goes on to make two tests of the embourgeoisement thesis which Goldthorpe, Lockwood and their colleagues called into question. This Note concerns itself with only one of the two tests, namely that based on correlations of residuals.

Kramer has shown how singularly restrictive are all ‘similarity’ conditions which guarantee transitivity of majority rule by limiting the family of admissible preference orderings (so-called exclusion restrictions). For suppose, plausibly, that social alternatives are points in an open convex policy space S ⊂ Rn, n ≥ 2, and that voters' preferences, {Rl)1=1…‥ l, are representable by continuously differentiable semi-strictly quasi-concave utility functions ul,. Suppose further that at a single point x ε S, any three voters' utility functions have gradients ∇ul(x), ∇uf(X), ∇uk(x), no one of which can be expressed as a positive linear combination of the other two, and no two of which are linearly dependent. Then all exclusion conditions must fail on S.

Hope's comments provide me with the opportunity to correct one statistical error which unfortunately found its way into the original article. He points out that, as the correlations rac and rbc were each about +0.2, deviations from the expected level of car- and house-ownership given a constituency's proportion of manual workers can only account for about 4 per cent of the variance in deviations from the expected Conservative share of the two-party vote. This, it is gently implied, hardly amounts to convincing evidence on behalf of the embourgeoisement thesis. The correlation coefficients given in the article were serious underestimates, however, and are in fact +0.571 and +0.328 respectively. Thus, deviations from the expected level of houseownership explain almost a third (32.6 per cent) of the variance in deviations from the expected Conservative share of the two-party vote, and deviations from the expected level of car-ownership explain over one-tenth (10.8 per cent) – which is more substantial

Political socialization research has been characterized by a number of poorly documented but widely accepted generalizations. In particular, it has been assumed that indetgenarational consistency in political attitudes is the usual, if not the inevitable, outcome of the political socialization process in Western democracies.

L. J. Sharpe's two-part article ('American Democracy Reconsidered’ in this Journal, III [1973], 1–28,129–68) is an important and controversial one. He questions many of the orthodox interpretations of American urban politics, suggests that some of the supposedly well-established observations about British and American contrasts may boil down to bits of conventional wisdom, and raises many points about social justice and democracy, which are at the centre of everyday practical politics but which sometimes slide from view in modern ‘value free‘ social science. The article will hopefully attract the critical attention of American writers, but meanwhile here are some reflections from a slightly different point of view on some of the themes of the article.

Cross-cutting cleavages do seem to help moderate social conflict.1 This can be explained in either of two ways. One argument focuses on the logic of electoral competition. Where parties must appeal to an electorate with diverse tastes along many dimensions, politicians must take moderate positions (defined as near the median voter) in most dimensions of cleavage if they are to win. A socialist party which draws its support from both Protestants and Catholics cannot take extreme positions on the religious question without alienating potential supporters and jeopardizing its electoral chances.2

I report with regret that several of the tables in ‘The Queen and the Prime Minister [this Journal, IV (1974), 257–87], contained small mechanical errors, largely affecting percentage distribution by one or two per cent. Should anyone need a ‘laundered’ reprint of the article, please let me know. There is one error of substantive importance: in Table 6 (p, 282) the correct percentage of French and U.S. white children perceiving that a head of state stopped by a traffic policeman would be described as ‘above the law’ is 43 and 31 per cent respectively. English nationalists will be pleased to know that this revised statistic eliminates the ‘finding’that U.S. whites are less likely to evince this sentiment than are English children. The correct statistics will appear in my December 1975 American Political Science Review article ‘The Benevolent Leader Revisited’.