Coalition phenomena are at the heart of politics everywhere, but they acquire a special significance in fragmented multiparty systems. In these systems, the distribution of political forces is such that parties are compelled to seek out at least one and often two or more partners, depending on the level of the system's fragmentation. The existence of a structural constraint that forces co-operative interaction is a major distinguishing feature of these systems and, accordingly, coalition-related phenomena occupy a central position in the political process. Coalition-maintenance activities are prominent, debates over coalition alternatives are frequent, procedures for the resolution of cabinet crises become institutionalized, and appropriate general orientations develop both at the level of the specialized actors in the political system and of the citizenry.

Some political theorists maintain that Niccolò Machiavelli was a rather immoral sort. His exhortations to guile, perfidy, deception and opportunism were numerous, his scruples few. Others, in a more revisionist vein, suggest that his preferred tactics were only meant for the common good. Yet, whether Machiavelli was a scientist, a descriptivist, a technician, a moralist or an immoralist is immaterial from one standpoint: he taught us something about the nature of human interaction in the State. Machiavellian interpretations of human events underlie many of our personal impressions of political life. We speak of strategy, tactics, morality, honesty as if the locations of political leaders along those dimensions determine what governments do for us or to us.

This article develops a number of themes first raised in an earlier paper where we attempted to publicize the existence of Census data based, for the first time, on British parliamentary constituencies, and where we briefly described the potential and limits of a variety of available statistical techniques of analysis. Until the earlier paper was published, studies of British electoral behaviour using aggregate data were largely historical, generally used only the simplest statistical techniques such as cross-tabulations, and usually proceeded blithely unaware of the snares of ecological inference. A small number of more advanced analyses had appeared but none focused on Britain or even on England as a whole. Since our earlier article appeared, there have been two attempts to construct predictive models of Labour support by applying multivariate statistical analysis to aggregate-level data. As we show in this paper, both Barnett and Rasmussen produce models that are statistically less powerful than our own and are subject to various weaknesses, of which the most important is the failure to tackle the problem of ecological inference.

The ‘primacy principle’ comprises three assumptions about political orientations. The first is that they are learned during childhood. The second is that this childhood learning further shapes any subsequent modifications of them. The third is that the scale of any such subsequent modifications is small: fundamental political orientations tend to endure through life. We propose, using cohort analysis, to examine the extent to which three political orientations – party identification, political efficacy and political trust – do, as a matter of fact, endure through adulthood.

Ever since the technique widely known as the semantic differential was pioneered by Charles Osgood, it has been widely applied in most of the social sciences, as well as in a variety of other fields. While the broadest applications of the technique, including much of Osgood's own work, are in the areas of experimental and social psychology, variants of the semantic differential are widely scattered through research in such diverse fields as linguistics, anthropology, and education, not to mention such non-academic fields as journalism and marketing. The attempt by a number of political scientists, particularly in Britain and in Canada, to adapt this technique to the study of political party images represents one of these variants.

In an important contribution to the improvement of data analytical techniques in political science, Budge and Farlie examine the predictive success of various background characteristics in determining political activism [Ian Budge and Dennis Farlie, ‘Political Recruitment and Dropout’, this Journal, v (1975), 33–68]. The authors use the framework of Bayesian statistics, in which the subjective probability that a given individual will be a political activist is revised in the light of sample information about the background characteristics of activists to give a posterior (i.e. after the information or event) probability that the individual is an activist. Unfortunately, as the authors admit, they do not utilize fully all the components of the Bayesian approach.

Whiteley's comment presents a ‘full’ Bayesian argument for the identification of councillors. A Bayesian analysis is a procedure by which an individual can make coherent judgements between hypotheses in terms of conceptual gambles. It is important to recognize that the coherence is between the individual's prior and posterior beliefs when presented with evidence and is not to be interpreted as the agreement between the decision and the correct decision. There is in the argument no appeal to the probability of a correct decision as a relative frequency concept. In many circumstances the correct decision may be undecidable so that a frequentist approach would be inapplicable anyway. Two Bayesians who disagree on their prior beliefs concerning a finite number of hypotheses will disagree on their posterior beliefs when presented with the same data provided that no hypothesis is declared impossible by the data.