Ich bin mehr und mehr davon iiberzeugt, das man fiir diese Arbeit einen langen Atem braucht, das es sich urn eine endlose Abfolge dialektischer Konfrontationen von generativen Erklärungsversuchen mit den harten und unerbittlichen Tatsachen der Geschichte und der empirischen Soziologie handelt (Rokkan 1980, 118).

Rokkan's studies of cleavage structures and nation-building linked him, right from the start, to the American political sociology of the 1950s and the 19605: Parsonian structural functionalism, cybernetics and Eastonian systems theory, the main schools of political development and nationbuilding(Deutsch Yale; the Almond-Pye Committee at Princeton), as well as the Michigan School of electoral studies. Hence, taking Rokkan as a point of departure will enable us to throw light on the different approaches of an entire generation of scholars to which Rokkan himself belonged. Because of Rokkan's strong urge to synthesize and combine different approaches by making skilful use of other scholars' ideas and categories in a different context, Rokkan can be said to simultaneously incarnate the different schools of thought of the 'Golden Age of Political Sociology'. The Michigan school of empirical electoral studies was essentially ahistorical and strongly oriented towards the psychological aspects of partisanship. Rokkan was equally as data-fixed as the Michigan school, but he did not accept the ahistorical and psychological approach which downplayed

Resumé

This article presents the major characteristics of the tradition of comparative political sociology which Stein Rokkan incarnated. Special emphasis is put on Rokkan's scientific reasoning based on the organic relationship between the creation of infrastructures for research and research efforts. Rokkan's oeuvre is marked by an acute awareness of the historical sources of institutional variation and by a concern for constructing region-specific models reflecting culturally meaningful contexts. Rokkan's use of comparisons is treated at length before the conclusion states what the authors claim is worth preserving from the Rokkan tradition: the historical, institutional and comparative aspects of Rokkan's approach.

contexts. The concern for context brought Rokkan in contact with the contemporary schoolsof political development and 'nation-building'. These schools addressed themselves to problems of the contemporary context. They were ahistorical, and mainly interested in showing how aggregated individual political orientation related to political stability. Rokkan was also strongly concerned with general development processes; he, however, put a much stronger emphasis upon institutional variation, as a result of the historical strategies of elites far back in history. It became crucial for him to show how elite decisions became institutionalized and how variations in such early institutionalization influenced modern mass politics. This interest made him particularly concerned with developing strategies of systematic comparison which took into account the structural aspect. This is illustrated by his search for region-specific models, i.e. for a particular model capable of explaining mass political behaviour in Western Europe, and explains Rokkan's close contacts with the main historical sociologists of the epoch (e.g. Bendix, Moore, Eisenstadt).

Rokkan's research strategy, i.e. his attempts to explain structural variation retrospectively, raises problems of a general character. These problems are still highly relevant at a time when the micro-oriented rational choice paradigm is under attack and a return to political culture and institutional, analysis is in the making (March & Olsen 1989; Powell & DiMaggio 1991; Wildavsky 1987; Thompson et al. 1990). Rokkan's unique middle position as a structural functionalist and historical sociologist provides an opportunity highlight some of the main theoretical and methodological problems comparative research.

In order to arrive at a critical assessment of the strengths and weaknesses of the Rokkan tradition, we will describe the organic and intimately related links between the sum of Rokkan's efforts as an organizer and scholar: the creation of organizational networks for the exchange of ideas and explanatory attempts, the establishment of social science data archives in order to confront theories with 'the merciless facts of history', and the subjection of history and type of comparison to the choice of a constant explicanda, i.e. structural variation at a given point in time viewed in retrospect.

The Rokkan Legacy

Rokkan was for several reasons a 'tradition-builder' within the field of political sociology. Above all, this is related to the buildup of cross-national organizational networks among social scientists, and the establishment of social science data archives and databases (Torsvik 1981, 17). The purpose of his activities was always to create frameworks and infrastructures for research.

Within these infrastructures, however, Rokkan was primarily concerned with explaining macro variations of the political development within Western More precisely, some main themes and periods can be singled out:

(1) Studies of Norway in a comparative perspective where he stresses variation in electoral participation and the introduction of new groups into the political arena. These questions concern Rokkan in the late 1950s and in the early 1960s (Rokkan 1966, 1967, 1969, 1970a, 1970b).

(2) Studies of the small Western European countries where he brings into focus variations in cleavage formation and party systems. Rokkan's interests turned more in this direction from the late 1960s (Rokkan 1967, 1968, 1970c).

(3) Studies of variations in centre formation and nation-building in Western Europe, which culminate in his 'Conceptual Map of Europe'. These concerns are at the centre of Rokkan's interests from the early 1970s (Rokkan 1973a, 1975a).

(4) Studies of variations in territorial identity and boundary formation (Exit, Voice) within the context of the so-called ETI-project (Economy, Territory, and Identity), which occupied Rokkan in the late 1970s (Rokkan & Urwin 1983).

As is apparent from this brief overview, one of the most salient aspects of Rokkan's models is the detailed coverage of territorial and geographical factors. Rokkan was influenced by the seminal book The Bias of Communication written by the Canadian economic geographer Harold A. Innis. In fact, among the leading scholars of nation-building, Stein Rokkan is the only one who develops a conceptual system for concrete analyses of centre-periphery relations. The works of other leading scholars - e.g. Barrington Moore Jr.'s studies (1966) about the rise of different types of political systems - have been truly path-breaking, but their importance lies in the establishment of certain prototypes of development rather than in providing a methodology for analyses of territorial processes and concrete centre-periphery relations. There is in Barrington Moore Jr.'s analysis a strong large nation bias, and a lack of interest in small units and variations in the attributes of peripheries. Likewise, in the studies of Immanuel Wallerstein (1974), the distinction between centre and periphery is of utmost importance, but Wallerstein speaks in terms of a world system and is not much concerned with single states and nations.

At the centre of Rokkan's own work is the so-called 'Basic Model'. This systematization of the most important elements of his various models of structural variation represents an attempt at identifying 'the crucial variables in the long and complex process that led up to the current constellations

of territories, economies and political alignment systems' (Rokkan 1976, 2). The essential message of the model is simple enough: one cannot explain variations in the structuration of mass politics in Western Europe without going/ar back in history, without analysing differences in the initial conditionsand early process of territorial organization and state-building and combination of resources. In reality this meant going back to the High Middle Ages, 'to an analysis of the decisive differences in the conditions of centre formation and territorial control' (Rokkan 1976, 2).

In the 'Basic Model' there is a simple classification of sources of variation at this 'take-off stage. It identifies the strength of the urban network and the accompanying flow in long-distance trade as a basic economic variable. It also identifies the military-administrative strength of the dynastic centres as a territorial variable par excellence, and suggests an equally important cultural variable - the ethnic-linguistic homogeneity of the populations controlled by certain centres. The 'Basic Model' then continues with a corresponding specification of variables for the next stage of development, i.e., the consolidation of territorial states during the troubled period from 1500 to 1648. The subsequent period of consolidation from the Peace of Westphalia to the French Revolution (1648-1789), by comparison, contains only one source of variation: the strength of representative institutions during the age of absolutism. Together, this complex set of Precondition Variables constitutes a launching pad for an analysis of a set of Intervening Process Variables. These in turn constitute the point of departure for the analysis of the Explicanda - namely variations in the structures of political response.

In his well-known 'Conceptual Map of Europe' (Rokkan 1973a), Rokkan sets out to explain 'the greatest paradox of European development', i.e., the fact that the strongest and most enduring state formations emerged on the outskirts of the Roman Empire, whereas the core areas in Italy and Germany remained politically fragmented well into the 19th century. This is an attempt at systematizing relations between Precondition variables and Intervening Process Variables in the 'Basic Model'. The purpose of the conceptual map is to explain variations in the state formations of Europe, which again are intended to explain variations in the development of political and economic citizen rights.

After having briefly presented Rokkan's most enduring and original contributions, it may be useful to examine his scientific reasoning step-bystep order to understand the close relationship between the buildup of cross-national social science data archives and organizational networks and corresponding research efforts. It is to such a close examination of Rokkan's intellectual development that we now turn.

Rokkan's Approach

Rokkan's first studies were centred around political philosophy. He spent the year 1948-49 at UNESCO in Paris together with the philosopher Arne Næss in order to organize an international enquete on the meaning of the word 'democracy'. This enterprise aroused Rokkan's increasing interest in problems related to mass political participation, and he became involved in a team conducting international survey analyses. Rokkan was one of the first to carry out a thorough analysis of comparative survey data - data concerning teachers in seven Western European countries (Rokkan 1955). It was precisely this analysis that 'forced' Rokkan to become interested in the theoretical problems of comparison. How could one account for differences between countries?

It soon became evident that it was impossible to explain all differences through variables at the survey level, i.e. at the micro or individual level. It was necessary to move on to an analysis of macro-characteristics, to context variables. This raised a series of difficult questions: Why are the contexts so different? How have they developed, and how can these differences be explained by a general model? However, Rokkan was not the only one to encounter these problems. S.M. Lipset and Juan Linz faced the same micro-macro problems in their analyses of survey data concerning the share of left- and right-wing parties comprised of different groups of workers. They soon realized that it was not possible to explain differences in the electoral behaviour of the workers without analysing the alternatives that the party systems of each country presented the workers with.

This conviction was strengthened by experiences from the Norwegian electoral research project, started in 1956. Angus Campbell was convinced that the Norwegian case would reveal the same differences in political participation according to profession and level of education as was found in the United States. This did not turn out to be the case and the conclusion was that the difference between the two countries had to be explained at the structural level (Campbell & Rokkan 1960). The Norwegian five-party system corresponded more directly to the socio-cultural cleavages than did the US two-party system.

This is how Rokkan became concerned with the construction of a model to explain the relation between cleavage structures and party systems. The first sketches towards such a model stem from 1960 (Rokkan 1962). The first elaborated version was published in the introductory chapter to the Lipset and Rokkan volume in 1967. It was a model consisting of four fundamental cleavages: two cultural - state vs. church and centre vs. periphery - and two economic - town vs. countryside and labour vs. capital. Why did these cleavages not become politicized to the same extent in all countries? Why did some cleavages lead to the formation of a political party in some countries and ie others not? Rokkan's answer:

Ich håbe die Antwort in der Geschichte der Allianz - und Oppositionsbildung innerhalb
jeder territorialen Elite gefunden (Rokkan 1980).

In his effort to systematize such alliance formations, Rokkan reduced the
number of possible variations to eight types, and specified the dominant
cleavages for each type and the corresponding party systems.

But Rokkan was not satisfied with this solution:

Ich war seinerzeit mit dieser Typologie von acht Fallen wenig zufrieden: in der Arbeit von 1967 habe ich auf eine Reihe von Schwierigkeiten hingewiesen, und in einem Kapitel, dass ich ein Jahr darauf veröffentlichte, bin ich zu einer viel detaillierten Differenzierung gelangt (Rokkan 1980).

Rokkan became increasingly aware of the importance of longue durée, the long-term effects of decisions and alliance formations occurring throughout the histories of each country. This is evident in his systematic analyses of variations in the conditions for party mobilization, i.e. the suffrage extensions 1968, 1970a). Rokkan realized, however, that his cleavage model treated each case as too isolated from each other, without considering neighbouring contexts and geopolitical positions. This is how he began to study the spatial ties between the different cases, and he became convinced of the decisive importance of the interterritorial connections for the process of nation-building as well as for the later structuring of mass politics.

This interest triggered Rokkan's fascination with Hirschman's Exit, Voice, Loyalty (1970) because these concepts vested Rokkan's models with a more dynamic character, i.e. they helped specify processes of boundary construction (Rokkan 1973a, 1975a, 1975b, 1975c, 1977). Hirschman's concepts challenge Rokkan's somewhat static and closed systems theory from the 19605. In Rokkan's early versions the relations between the territorial and functional dimensions are not clarified, and we get to know very little about what actually determines the boundary construction itself (the unit of analysis) (Rokkan 1974a, 1974b, 1976). Hirschman's concepts made clear how unstable boundaries can be, which implies that one cannot define them a priori. In other words, Rokkan, like other system functionalists, from closed to more open models.

For Rokkan it now became imperative to demonstrate how and to what extent variations in the possibilities for exit decisively influence the preconditions for voice (Rokkan 1973b, 53). It was by way of a thoughtful consideration of problems related to these spatial connections that Rokkan developed 'The Conceptual Map of Europe', a topological description of a combination of fundamental variables (Rokkan 1973a, 1975a).

In the search for variables to explain structural variations, Rokkan focused on certain historical junctures. A certain kind of determinism but also of voluntarism lies underneath: certain structures create different spaces of possibility; and when the space of possibility changes, it provides

variation. Processes are institutionalized through crucial junctures, which manifest cleavages, and freeze into structures. For each period there thus exists a certain space for possible action. When decisions are made, they impose new limits to what is possible or impossible, probable or improbable. History engenders institutions and institutions impose constraints.

The subject and level of Rokkan's explicanda in this fashion changed over time, but the aim of explaining structural variations remains constant. Rokkan was not, in other words, primarily concerned with explaining the suffrage extensions, the party systems or the state- and nation-building processes in Western Europe, but rather with how these vary within a given geographical context. In his retrospective analyses Rokkan ransacks history for factors or variables that can explain variations across cases, and always in accordance with the principle of 'parsimony', i.e. he consciously rejects all variables that are not capable of explaining at least one important difference between at least two cases (Rokkan 1980, 125, 1974c). Rokkan thus always had an open mind towards introducing new variables into his models in keeping with the rules of variance reduction. For a given amount of reduced variance, he always preferred the smallest number of variables.

Time and again Rokkan stressed the openness of his models and his open attitude towards other social sciences and towards the cultural sciences (Rokkan 1972, 1974a, 1974b). Rokkan's relationship to other sciences was thus highly eclectic. They only interested him as sources of inspiration for new ideas which could take his own project further. The models are open in the sense that Rokkan continuously included new variables and variable constellations in an incessant series of explanatory attempts. This is related to Rokkan's concept of research and science as a process located in a field of tensions between accumulation and innovation (Rokkan 1972). He strongly stresses that the models are heuristic and that the concepts are to be considered as opening words. Hence, the central concern for Rokkan is the research process itself.

Rokkan's strong belief in a long-term cumulative research process, in which the aim is to reveal more and more of the empirical reality by incessantly introducing new variables, suggests a belief that an 'objective' picture of reality can be found. Such a view comes close to the positivistic scientific ideal. If one can find all the constituent parts, it is eventually possible to explain why things turned out the way they did. Rokkan's concern with linguistics - i.e. with the idea that all languages can ultimately be reduced to small elements called 'phonemes' - also points in this direction (Rokkan 1972, 1974b). Thus Rokkan's scientific reasoning fits in nicely with his efforts as an organizer of social science infrastructures: the stepwise creation of the building blocks of the great scientific edifice.

Rokkan Compares: Comparison Inseparable from Theory?

Rokkan's focus on structural variation certainly requires a comparative strategy of explanation. But what type of comparison? In the literature several typologies of comparisons have been presented (Bonnell 1980; Skocpol 1984; Tilly 1984). However, Tilly (1984) is so far the only one who has ventured to typologize Rokkan's use of comparison. Tilly's typology combines two dimensions of comparison: the degree of sharing of all instances and multiplicity of forms. In sharing, the statement resulting from a comparison can range from a single instance (getting the characteristics of the case right) to all instances of the phenomenon (getting the characteristics all cases right). In multiplicity, the statement emerging from a comparison can range from single (all instances of a phenomenon have common properties) to multiple (many forms of the phenomenon exist). Tilly's typology of strategies of comparison thus consists of four pure types: individualizing, universalizing, variation-finding and encompassing.

A purely individualizing comparison treats each case as unique, taking up one instance at a time, and minimizing its common properties with other instances. Reinhard Bendix is a case in point (1964, 1974, 1978). Bendix explicitly avoids drawing conclusions about causal connections on the basis of comparative material (1984). It is Bendix's conviction that the number of cases is too limited and the number of variables too large in order to draw causal conclusions in macrosociological studies. Bendix's analytical use of comparison enables him to identify general developmental patterns at the same time as the sense of historical particularities is preserved.

A pure universalizing comparison, by contrast, identifies common properties all instances of a phenomenon. Skocpol's study of States and Social Revolutions (1979) is a case in point. In this study the phenomenon of social revolution has common properties in all three instances, namely France, Russia and China.

Variation-finding comparisons are, in turn, supposed to establish a principleof in the character or intensity of a phenomenon by examiningsystematic among instances. Variation-finding historical sociologists are concerned with variations in history with the aim of proving causal patterns. The difference between individualizing and variationfindingcomparisons be illustrated by comparing Reinhard Bendix's and Barrington Moore Jr.'s different views regarding the purpose of comparisons.According Bendix, comparisons should only be used for contrastingsocio-historical with each other. Comparative studies cannot replace causal analysis. According to Barrington Moore Jr., on the other hand, comparisons may serve as crude negative controls of established historical explanations. A comparative approach may form the platform

for new historical generalizations. Rather than contrasting entire historical trajectories in relation to given concepts or themes, variation-finding historicalsociologists more in terms of alternative hypotheses and comparisons across relevant aspects of the historical units being compared.

Finally, comparisons of the encompassing type place different instances at various locations within the same system, on the way to explaining their characteristics as a function of their varying relationships to the system as a whole. Here Wallerstein (1974) is a case in point. His unit of analysis is one single transnational system consisting of many parts. Wallerstein uses comparison primarily to throw light on the position and function of the constituent elements within the world system. The different units are not compared with each other but with the theoretical model. The use of comparison in this situation has the purpose of illustrating an a priori chosen model. Wallerstein refuses to use comparisons analytically because he doesn't consider the comparative logic applicable to partial and geographically units, such as national states within a capitalist world economy.

Wallerstein, in other words, primarily uses comparison to demonstrate the position and function of the constituent elements within his world system. By referring to 'The Conceptual Map of Europe', Tilly maintains that Rokkan originally constructed the conceptual map of Europe for the purpose of studying links and relations between the different units. Geopolitical position and spatial location of an area (country) within larger interregional landscape contexts (Europe) became increasingly important for Rokkan (Rokkan 1976, 9). In this sense Europe is Rokkan's equivalent of Wallerstein's world system. The point made by Tilly, however, is that Rokkan actually keeps harking back to 'variation-finding' comparisons epitomized by Barrington Moore Jr. (Tilly 1984, 116-124). Despite single hints at 'interdependence', Rokkan treats the different national experiences of development as individual 'cases' which illustrate the consequences of having been exposed to different combinations of 'variables'.

Although we agree that the distinction between variation-finding and encompassing comparisons is somewhat diffuse in Rokkan's work, it is our view that there was a clear development from variation-finding towards more encompassing comparisons. This is intimately related to Rokkan's increasing realization that he had tended to treat his cases too much as isolated cases. We have in mind Rokkan's application of Hirschman's concepts of boundary constructions. In stark contrast to Wallerstein, Rokkan to be encompassing in his comparisons only within the framework region-specific contexts. It is these contexts that Rokkan considers as systems and within which he intends to find and explain variations.

Between Durkheim and Weber

The blurred boundary between variation-finding and encompassing comparisons in our view due to a tension between history and generality in Rokkan's political sociology (Berntzen & Selle 1990). Rokkan did not use comparisons to throw light on special characteristics of each single historical case. The goal is not primarily to specify particularities through contrasts in order to underline the unique and the particular. If that were the goal Rokkan wouldn't need so many cases within one single cultural region, but would instead tend to use pairwise comparisons, as Bendix (1984) does. But it remains obscure, however, whether Rokkan considered 'regional models' purely as a preliminary stage, a step on the long and winding road towards a distant goal - a general model for political development in modern industrial societies. Or was Rokkan rather concerned with regionspecific because regions satisfy the methodological requirement of 'most similar design' in comparative analysis (Przeworski & Teune 1970), where the aim is to develop models of the middle range type (Merton 1968)? Rokkan always presupposed a kind of contextual control. He was unwilling to accept any interpretation which lacked the necessary basic contextual understanding. For Rokkan it was always necessary for interpreters of data to be area or country specialists in order to avoid 'numerological nonsense' (Rokkan 1970d, 288).

Although Rokkan showed a constant concern for outliers, the cases that did not behave according to the predictions of the model, we tend to believe that Rokkan goes a long way towards considering Western Europe as a system of relatively fixed boundaries vis-å-vis the outside world, and that he is seeking to explain functionally internal variance within this very system. But this is a form of comparative analysis of covariation which separates data from history, a separation Rokkan is able to make because Western Europe has so much in common. It is precisely the historical community that makes it acceptable to separate history and data. Rokkan's emphasis on the establishment of historical data archives for purposes of ecological analysis appears consistent with this perspective. Statistical ecological analysis is based on such a separation of history and data.

Because of Rokkan's choice of measuring variance within the system of Western Europe, he must be said to belong to the Durkheimian tradition with respect to the method of measuring variation, which is largely in accordance with modern use of statistical analysis of social data (Ragin & Zaret 1983). The Durkheimian tradition in Rokkan's approach is further underlined by Rokkan's strong belief in the long-term cumulative character of the research process (1972, 1975b). Rokkan's strategy of measuring variance, like that of Durkheim, is highly variable-based. The units of analysis are supposed to be autonomous and constitute independent observations. units are considered as systems.

The supposition that the units constitute discrete systems entails an interest in permanent causes. The causes become characteristics of the units of analysis and can therefore not be isolated for experimental purposes. For this reason it is impossible to determine directly the real effect of a given factor. The focus on permanent causes and covariation analysis leads in the direction of many comparisons, the more the better, to control for contextual factors. In this respect, Rokkan's comparative strategy can be interpreted as being basedon the idea of 'social species', which can be observed empirically and classified objectively. Rokkan's supposition of a basic similarity between sociology and biology made him emphasize causes which are internal to the object of analysis, or permanent causes. Social causes have their origin within the internal milieu of societies.

The logic of analysis which is applied in comparative studies of permanent causes minimizes the importance of non-systemic causes related to historical context, or considers unexplained variance as 'culture' or 'noise'. The advantage of this strategy of comparison is its potential for generalization, at the expense of complexity. Explanations referring to connections between abstract variables encourage a neglect of the historically determined character society, i.e. institutional variation. This leads to a denial of history in the sense that the past can be considered as a cause of the present. A variable-based strategy seeks transhistorical generalizations, not concrete knowledge about specific units.

But it is precisely at this point that Rokkan represents a deviant and highly unique position. Rokkan does not accept the distance between theory and data in the Durkheimian tradition because he is interested in institutional variation rather than aggregated individual political values. Here Rokkan's emphasis upon region-specific models becomes crucial. These models and typologies are case-based in the sense that they are based on concrete historical knowledge of the regional context.

Rokkan's awareness of contexts and his concern with historically generated cleavages and conflicts are in this respect closer to the Weberian tradition since it was Weber who first presented such a conflictoriented 1985) and case-based research strategy (Ragin & Zaret 1983). Weber was also concerned with explanation and generalization, but in contrast to Durkheim's variable-based strategy, Weber's explanations were genetic, not functional, arid his generalizations were historically concrete, abstract and ahistorical (Weber 1949). The Weberian comparative is particularly suited for subjects that cannot be treated adequately by means of statistical strategies of comparison: questions pertaining to historical variation. Whereas the Durkheimian comparative strategy seeks generalizations by separating data and history, the Weberian strategy turns to history in order to establish modest generalizations about historical variation.

Turning to history in this way implies a difference in type of explanation and in degree of generalization. This difference has important implications with regard to the comparative method. In arguments about permanent causes, cause and effect are chained together in a continuous way. Variation in the one leads to variation in the other. Genetic arguments, by contrast, are combinatorial. Arguments about permanent causes tend towards the use of correlation analysis as a suitable analytical tool. Genetic arguments imply qualitative historical analyses based on 'logical methods'. Such methods are logical by nature because they are not used to explain variance in the statistical sense, i.e. they are not probabilistic (Rossi 1982, 1983; Ragin & Zaret 1983).

This interest in relations which do not vary, in different combinations of historical causes and outcomes, is what most sharply limits qualitative historical methods from the statistical. Such an interest in historical combinations at the heart of Rokkan's definition of the structural boundaries that give room for variation in mass behaviour, i.e. his explicanda. His qualitative-typological modernization model, which delineates his precondition intervening process variables, clearly builds upon combinatorical e.g. his focus on critical historical junctures in explaining structural variation. Processes are institutionalized through crucial which manifest themselves as cleavages and freeze into structures. Like Weber, Rokkan is highly sensitive to the long-term impact of decisions and alliance formations during the historical trajectories of individual countries. However, their dependent variables (explicanda) differed. focuses on one of Weber's cultural areas, namely Europe. Whereas Weber's concern was to explain the unique development of Europe in contrast to other main cultural areas, Rokkan's attention is directed towards qualitative variations within the European unit. The purpose of comparison for Rokkan is to arrive at parsimonious explanations of such region-specific variations. When Rokkan proceeds to provide concrete he comes closer to Durkheim mainly because his explicanda basically different from those of Weber.

As the discussion here shows, there is still much ambiguity with respect to the relation between theoretical approach and type of comparison. To the extent that Rokkan became increasingly aware of the importance of the geopolitical position and spatial location of a country within a larger region-specific context, his use of comparison, however, moves closer to the encompassing type, i.e. there is a close relation between the theoretical concern with explaining within system variation and the encompassing type of comparison.

Making Sense of Rokkan

Although the subject and level of Rokkan's explicanda change, the explanatoryaim

toryaimremains the same. It is evident that no matter how many variables or combinations of variables are included in the analysis, what is to be explained - i.e. structural variation at a given point in time - and the retrospective method remain constants in Rokkan's oeuvre. These premises were laid down once and for all and the way of phrasing the questions always remained the same. The choice of what is to be explained, the 'dependent variable' in other words, entails binding methodological consequences:explanations up as retrospective and teleological. The task of explaining structural differences presupposes that one heuristically chooses to 'freeze' the situation at a given point in time as a kind of 'end product' of earlier development, which is then to be explained. When all actors have been included, there is nothing more to be said (Galtung 1992). This contributes to the somewhat static character of the models and one is pushed in the retrospective direction. The purpose is to go back in history to find variables that can explain why the situation has turned out the way it actually has, and not to examine what the different alternatives were at different points in time.

Between the more abstract and the context-specific theoretical elements, there is a void in Rokkan's work (Flora 1992; Berntzen & Selle 1990). Hence, the main critics of Rokkan (Abrams 1982; Alford & Friedland 1974; Seip 1975) point to the wide gap between theory and empirical facts. The target for Abrams's critique, for instance, is Rokkan's cleavage model for explaining variations in the party systems of Western Europe. This 'explar ation', according to Abirams, is nothing more than a statement about covariation, where the possible causal relations are extremely difficult to grasp, except within the context of the historical development of individual countries. The paradigm is construed in such a way that there are almost as many theoretical sequences as there are empirical cases: each theoretical sequence covers only the case it is exemplified by, and each case exemplifies a different sequence among the plethora of possible combinations of relevant

An over-generalized and over-abstract conceptualization of key variables in the paradigm entails acute difficulties for the establishment of analytical control with the empirical variations of the historical world. Rokkan's cleavage structure/party system model allows him to explain a single country's political history in a relatively parsimonious way, given the importance of the relevant variables. Yet the model is incapable of postulating, alone proving, clearly defined causal relations between the variables, or of explaining the importance of relevant variables in a given country at a given point in time. In this vital sense history is taken for granted and escapes the theory.

Rokkan's interest in both the general and the specific is particularly
evident in his 'Conceptual Map of Europe'. Behind the 'Conceptual Map

of Europe' there is a general model of state and nation-building, a set of basic assumptions about the main processes and factors of nation-building. There is no clear distinction between Rokkan's general model of nationbuildingand 'Conceptual Map of Europe'. For analytical purposes, however, such a distinction should be made. The phases of integration: penetration, standardization, participation and redistribution seem to constituteelements the nation-building process by definition. The phases may occur simultaneously, have another succession or can be reversed. We have to do therefore not with a universal sequence of development, but phases derived from the actual historical development in Western Europe, and these phases serve purely analytical purposes as reference points for comparison and the discussion of deviances. It thus becomes even more important to identify clearly the territorial units which pass through the phases. But this is precisely the problematic point because territorial consolidation, centre formation and the definition of national boundaries are variables both in the 'Basic Model' and in the conceptual map of European geopolitical history (Østerud 1978, 136). The danger of making a circular argument is imminent: the units of analysis are adapted to characteristics of the phases of development.

Rokkan employs a basic organic metaphor in his analyses of societies and social phenomena. The different parts of Rokkan's model are not independent but are parts of an 'interdependent' system where they enter into a functional relationship. But does Rokkan's retrospective perspective also imply a theory of evolution? In our opinion a silent presumption of evolutional development is implied when one resorts to history to find more or less of, or different forms of, phenomena that already exist. This entails important consequences for Rokkan's concept of action. If one works within an evolutional framework, there is no need to know the intentions, motives or ideologies of the actors because one starts out with the results (Elster 1985, 318). In other words, from an evolutionary point of view one can unproblematically adopt a retrospective perspective.

Rokkan's choice of explicanda (structural variation) and research strategy retrospective approach) push action and the content of politics into the background. Consequently, Rokkan presupposes the existence of political actors and introduces them as exogenous to his models. Rokkan uncovers objective structural sources of conflict, but does not clarify under what conditions different groups are able to fight for their interests. Given Rokkan's choice of explicanda and method, it follows logically that he is only interested in the results of politics ex post factum.

We do not argue that Rokkan disregards human action as a transmission belt between structural conditions and social outcomes. Rather he introducesaction festum as in classical pluralist group theory. But the question remains whether or not suitable actors always can be found to

perform the roles demanded of them by the structures. Rokkan is therefore unable to distinguish between why an action is possible and why it has been executed. Hence, Rokkan falls into the teleological ditch: he postulates for example the existence of the category of 'nation-builders' without telling us who they are, where they come from, and, when the nation-building process is 'completed', what happens to them afterwards.

Rokkan's approach makes it difficult, if not impossible, for him to defend the choice of independent variables theoretically as well as historically. This constitutes in our view the main limitation of the retrospective method embedded in a structural functional explanatory logic. When political action is introduced post hoc, and the content of politics is neglected, one becomes unable to argue theoretically and historically that a certain independent variable actually had or always has had the importance it is attributed retrospectively in accordance with the structural functional logic.

Even if we find Rokkan's contextual approach more fruitful than today's dominant micro-oriented trends, in short, it does not suffice to reveal historical and institutional influences. Such an approach must be coupled with the level of meaning, with cultural and ideological aspects. Cultural and social milieux must be explained and not only introduced from the outside. Preferences are neither self-evident nor do they emanate directly from structures. They can be fruitfully brought into the framework of analysis as a possible link in the relationship between the micro-level and the macro-level, which after all was the central theme in Rokkan's work.

Despite all of the possible criticisms, the historical, institutional and comparative aspects of Rokkan's approach remain highly relevant. Rokkan starts at the right end: by attempting to reveal how historical and institutionalconditions interests and actors' room for manoeuvre. Rokkan's point of departure is the self-evident, namely that we shape our history under circumstances not chosen by ourselves. Rokkan is an institutionalist in the true sense of the word. This is corroborated not only by his concepts such as thresholds, closeness, etc., but also by his subject matter - suffrage extensions, cleavage formations, state- and nation-building.Institutions between cleavages and mass mobilization. Institutionsare given and impose limitations on both action and the effects of action even if values may change rapidly. Rokkan was concerned with structurally rooted interests as sources of new cleavages, and he was a leading force in turning away from the socio-psychological orientation in political sociology. It is quite obvious that if one (as Rokkan) is concerned with the influence of structures, the purpose of comparison must be different from that of those who are concerned mainly with socio-psychological values and individual life styles. Since values exist everywhere, regardless of context, fixation upon them alone thus entails a lack of context control.

According to Rokkan, it becomes impossible to interpret values correctly
without such a control for context.

It is precisely Rokkan's deep understanding of the importance of contexts which makes him emphasize the strategy of region-specific models and which determines Rokkan's way of using comparisons. In this sense the Rokkan legacy represents both a highly relevant and a modern approach in the current efforts at understanding political process and conditions for development in both European and non-European contexts.

ACKNOWLEDGEMENTS

We thank Stein Kuhnle, Ulf Lindstrøm, Johan P. Olsen, Else Øyen, and two anonymous
reviewers for their comments.

REFERENCES

Abrams, P. 1982. Historical Sociology. Ithaca, N.Y.: Cornell University Press.

Alford, R. & Friedland, R. 1974. 'Nations, Parties, and Participation: A Critique of Political
Sociology', in Theory and Society, I, 1974.

Bendix, R. 1964. Nation-Building and Citizenship. New York: Wiley.

Bendix, R. 1974. Work and Authority in Industry. Berkeley: University of California Press.

Bendix, R. 1978. Kings or People. Power and Mandate to Rule. Berkeley: University of
California Press.

Bendix, R. 1984. Force, Fate, and Freedom. On Historical Sociology. Berkeley: University
of California Press.

Berntzen, E. & Selle, P. 1990. 'Structure and Social Action in Stein Rokkan's Work', in
Journal of Theoretical Politics 2(2), 131-150.

Bonneli, V. E. 1980. 'The Uses of Theory, Concepts and Comparison in Historical Sociology',
in Comparative Studies in Society and History 22(2), April 1980.

Campell, A. & Rokkan, S. 1960. 'Citizen Participation in Political Life: Norway and the
United States of America' in International Social Science Journal 12 (1), 69-99.

Collins, R. 1985. Three Sociological Traditions. Oxford: Oxford University Press.
Elster, J. 1985. Making Sense of Marx. Cambridge: Cambridge University Press.

Flora, P. 1992. 'Stein Rokkan's Makro-Modell for Politisk Utvikling i Europa: Et Rekonstruksjonsforsøk',
Hagtvet, 8., ed., Politikk mellom Økonomi og Kultur. Internasjonale
Streiflys over Stein Rokkans Politiske Sosiologi. Oslo: Ad Notam.

Galtung, J. 1992. 'Om å bygge på Stein Rokkan', in Hagtvet, 8., ed., Politikk mellom Økonomi
og Kultur. Internasjonale Streiflys over Stein Rokkans Politiske Sosiologi. Forthcoming.

Hirschman, A. 1970. Exit, Voice, Loyality. Cambridge: Harvard University Press.

March, J. G. & Olsen, J.-P. 1989. Rediscovering Institutions. The Organizational Basis of
Politics. New York: Free Press.

Merton, R. 1968. Social Theory and Social Structure. New York: Free Press.

Moore, Jr., B. 1966. Social Origins of Dictatorship and Democracy. Boston: Beacon Press.

Powell, W. W. & Dimaggio, P. J., eds. 1991. The New Institutionalism in Organizational
Analysis. Chicago & London: The University of Chicago Press.

Przeworski, A. & Teune, H. 1970. The Logic of Comparative Social Inquiry. New York:
Wiley-Interscience.

Ragin, C. & Zaret, D. 1983. 'Theory and Method in Comparative Research: Two Strategies',
in Social Forces (3), 731-754.

Rokkan, S. 1955. 'Party Preferences and Opinion Patterns in Western Europe', in International
Sociological Bulletin 7 (2), 247-252.

Rokkan, S. 1962. 'Valgsamfunnet og organisasjonssamfunnet: problemer for norsk og
europeisk forskning, in CMI, Beretninger XXXVII, 2.

Rokkan, S. 1966. 'Norway: Numerical Democracy and Corporate Pluralism', in Dahl, R.,
ed.. Political Oppositions in Western Democracies. New Haven: Yale University Press.

Rokkan, S. 1967. 'Geography, Religion and Social Class: Crosscutting Cleavages in Norwegian
Polities', in Lipset, S. M. & Rokkan, S., eds.. Party Systems and Voter Alignments. New
York: Free Press.

Rokkan, S. 1968. 'The Structuring of Mass Politics in the Smaller European Democracies. A
Developmental Typology', in Comparative Studies in Society and History, X, no. 2, 1968.

Rokkan; S. et al. 1969. Comparative Survey Research. Paris: Mouton.

Rokkan, S. 1970a. 'Citizen Participation in Political Life', in Citizens, Elections, Parties. Oslo:
Universitetsforlaget.

Rokkan, S. 1970b. 'The Mobilization of the Periphery', in Rokkan, S., ed., Citizens, Elections,
Parties. Oslo: Universitetsforlaget.

Rokkan, S. 1970c. 'Nation-Building, Cleavage Formation and the Structuring of Mass Polities',
in Rokkan, S., ed., Citizens, Elections, Parties. Oslo: Universitetsforlaget.

Rokkan, S. 1970d. 'Cross-National Survey Research: Historical, Analytical and Substantive
Context ', in Rokkan, S., ed.. Citizens, Elections, Parties. Oslo: Universitetsforlaget.

Rokkan, S. 1972. 'Forskningsstrategi', in Statsvitenskapelig Tidsskrift no. 1.

Rokkan, S. 1973a. 'Centre Formation, Nation-Building and Cultural Diversity: Report on a
UNESCO Programme', inEisenstadt, S. N. & Rokkan, S.,eds.. Comparing Nations. Vol.l.
Beverly Hills: Sage.

Rokkan, S. 1973b. 'Entries, Voices, Exits: Towards a Possible Generalization of the Hirschman
paper, ISSC-IPSA Seminar on Economic vs. Cultural Models in Comparative
Politics, Bellagio, June 1973.

Rokkan, S. 1974a. 'Entries, Voices, Exits: Towards a Possible Generalization of the Hirschman
in Social Science Information no. 13.

Rokkan, S. 1974b. 'Politics Between Economy and Culture. An International Seminar on
Albert O. Hirschman's Exit, Voice, and Loyalty', in Social Science Information, no. 13.

Rokkan, S. 1974c. 'Strukturendringer i Norden: Synspunkter på historisk-sosiologiske sammenligninger',
Nordic Congress of Sociology, Geilo.

Rokkan, S. 1975a. 'Dimensions of State Formation and Nation-Building: A Possible Paradigm
for Research Variation within Europe', in Tilly, C, ed.. The Formation of National States
in Europe. Princeton: Princeton University Press.

Rokkan, S. 1975b. 'Sentrum og periferi, økonomi og kultur: Modeller og data i kliometrisk
sosiologi', in Studier i Historisk Metode, no. 10.

Rokkan, S. 1975c. 'Votes Count, Resources Decide: Refleksjoner over territoritalitet vs.
funksjonalitet i norsk og europeisk politikk', in Makt og Motiv. Et festskrift til Jens Arup
Seip. Oslo: Gyldendal.

Rokkan, S. 1976. 'A Geoeconomic-Geopolitical Model of Sources of Variation Across the
Territories of Western Europe', ISSC-MSH symposium over Wallerstein and Anderson in
Bellagio.

Rokkan, S. 1977. 'Towards a Generalized Concept of Verzuiling, in Political Studies Vol. 25,
December.

Rokkan, S. 1980. 'Eine Familie von Modellen fiir die vergleichende Geschichte Europas', in
Zeitschrift fur Soziologie IX (1980), no. 2.

Rokkan; S. & Urwin, D. 1983. Economy, Territory, Identity. Beverly Hills: Sage.

Rossi, I. 1983. From the Sociology of Symbols to the Sociology of Signs. Toward a Dialectical
Sociology. New York: Columbia University Press.

Seip, J. A. 1975. 'Modellenes tyranni', in Studier i Historisk Metode no. 10.

Skocpoi, T., ed. 1979. States & Social Revolutions. Cambridge: Cambridge University Press.

Skocpol, T., ed. 1984. Vision and Method in Historical Sociology. Cambridge: Cambridge
University Press.

Tilly, C. 1984. Big Structures, Large Processes, Huge Comparisons. New York: Sage.

Torsvik, P. 1981. Mobilization, Center-Periphery Structures and Nation-Building. Oslo: Universitetsforlaget.

Wallerstein, I. 1974. The Modern World-System. Capitalist Agriculture and the Origins of the
European World-Economy in the Sixteenth Century. New York: Academic Press.

Weber, M. 1949. The Methodology of the Social Sciences. New York: Free Press.

Wildavsky, A. 1987. 'Choosing Preferences by Constructing Institutions: A Cultural Theory
of Preference Formation', in American Political Science Review 81 (1), 3-21.

Østerud, 0. 1978. Utviklingsteori og historisk endring: en kritisk fremstilling av utviklingsteoretiske
i samfunnsforskningen. Oslo: Gyldendal Norsk Forlag.