Gaustadalléen 30 (map)
ARENA Working Papers
An Essay on the Foundations of Comparative
Historical Social Science
The relationship between history and sociology has been the subject of heated controversies ever since sociology was established as a discipline. These debates continue, and over the last decades, historical sociology has been at the core of debates spanning the whole spectrum from specific questions regarding quantitative studies to complex discussions of highly philosophical matters. This essay gives a background to and tries to draw some conclusions with reference to these recent debates.
I shall argue that comparative historical sociology is more than a narrow branch of sociology, rather it offers a comprehensive alternative for macro-sociology and can be linked to a specific synthesis in sociology, namely that of interactionism. Comparative historical sociology can aid the disciplines it involves in specific ways: It can save history from drowning in monographs, it can help sociology avoid suffocation in the thin air of grand �social theory�, and it can help social science reflect on its predicament of always having to interpret the present.
The discussion which follows links comparative historical sociology to some crucial divisions both within the philosophy of science and within the social sciences. Thus, we shall rely on a historical and systematic framework that emphasizes the relation between social science and the other sciences. We pay specific attention to sociology, but most of our argument have relevance for the social sciences more broadly. We start by providing a brief overview of this framework. 
A definition of theory
There are several conceptions of theory in the social sciences, but they mostly remain unspecified. In order to prepare such a specification, let us make some simple distinctions, drawing on the philosophy of science.
Let us define theory as accumulated knowledge, organized by the human mind, to be used for purposes of explanation. This definition can be specified in two directions, yielding two very different �everyday philosophies�. Given the restraints on knowledge any one research collective can master, we are all in the situation that we �want to know more�, and we want to explain new aspects of reality. But we may work from one out of two convictions: cognitive optimism and cognitive skepticism. These basic orientations are practical philosophies of social science. They surface when scientists � more or less consciously � try to legitimate their activities.
One direction is that of cognitive optimism. It holds that full knowledge of reality is possible, even if we presently do not possess it. It claims that theory mirrors the essential forces of reality (cf. the laws of nature in the classical mechanical worldview). This idea of directly representing the basic forces of nature allows a clear-cut demarcation of science (scientific theories) from all other kinds of knowledge (ideological knowledge, everyday knowledge). The latter are disturbing illusions that must be overcome.
With reference to the above general definition of theory, cognitive optimists hold that accumulated knowledge should preferably take the form of laws (as in experimental natural science). To the extent that there is a hierarchy of laws, there are certain fundamental laws to which other laws can be reduced. This gives a maximum concentration of knowledge; the more such a theory can explain, the more powerful it is.
As for the remaining part of the definition of theory, cognitive optimists hold that the ability of the human mind to organize this knowledge poses no problem. They are aware that implications are derived from theories by means of logic and mathematics. But they trust that these are neutral universal and consistent tools used by the human mind in uncovering the main forces of reality.
The other direction is that of cognitive skepticism. This view does not deny that a mind-independent reality exists, but skepticism is waged concerning our ability to establish secure knowledge of the full range of this reality. The focus is thus on epistemological self-reflection, on how the human mind organizes knowledge, and on the limitations on our knowledge due to our cognitive apparatus. Today, a whole range of sciences � from brain sciences and neurophysiology to linguistics and social science � are involved in such self-reflection. In this view, the demarcation of science from other forms of knowledge is always contested. What counts as �objective� changes historically. Demarcation is a question of mechanisms relating to power, trust, consensus, etc. Only within such a knowledge regime or paradigm, the sciences produce their facts and check their statements against �reality�.
It is not necessarily denied that there are areas where knowledge can be organized in quite concentrated ways, but skepticism is waged against the ideal that such clearly demarcated knowledge will eventually uncover all the basic forces of reality. Thus, due attention is waged also towards the part of the definition which says that theory is �knowledge organized by the human mind�. Cognitive skepticists are aware that there are several varieties of mathematics and logic and that the consistency of encompassing mathematical-logical systems has been contested in the 20th century. The unwillingness to take systems of logic and mathematics as self-evident tools for the organization of knowledge was typical of both American pragmatism and European phenomenology, two historical cornerstones from the time when cognitive skepticism first became a widespread influence.
The cognitive optimist position holds that at one future point we shall arrive at a complete understanding, knowledge will converge, illuminating ever greater parts of the matters we are studying (the social world, in the case of social science). Cognitive skepticists are not willing to assume such convergence. Thus, the pragmatist aspect, that humans always develop knowledge for a purpose, is seen as important to the knowledge generated. A reflection based on the sociology of knowledge is indispensable if one takes into account that knowledge is organized by the human mind and for a purpose. The creation and reproduction of knowledge is seen as a social process. Knowledge is related to how human beings learn in a variety of their everyday activities.
Most likely these two positions will not make much difference as long as they relate to controversies on concrete matters of research. After all, most cognitive optimists accept that at the moment, our knowledge is incomplete. But once scholars with different basic orientations take off into the thin air of philosophical discussions, they create all kinds of polarizations.
Most recently, we have seen the two positions of cognitive optimism and cognitive skepticism clash in the socalled �science wars� that have haunted history of science. In social science and particularly sociology, there has been a related clash between realists and constructionists (although that dichotomy is not the same as that between cognitive optimism and cognitive skepticism). Unfortunately � but not so surprising given the nature of such debates in the borderland between the scientific community and the broader public sphere � these debates mostly have led to polarization. The analysis below is based on a wish to avoid such polarization. But the analysis also shows that for the last decades, the cognitive skepticist position has gained ground, in most social sciences and in history. This essay too is part of this movement. We are convinced that the most interesting positions are those �closest� to the line separating cognitive optimism and skepticism.  We do want to take seriously the cognitive optimists that try to grapple with insecure knowledge, and we do not embrace the relativism of the post-structuralist school. We do not deny that science can be cumulative, but we do insist that the conditions of and limits to accumulation of knowledge in various fields of science must be investigated. A number of factors influence the development of knowledge in various fields, we do not want to reduce such developments to e.g. that nietzschean will to power invoked by many post-structuralists.
Some main lines in postwar sociology
While the attitude of cognitive skepticism was not uncommon following the impact of Darwinism in the decades preceding World War I, cognitive optimism became dominant since the inter-war period. This lead to the dominance of the covering law model of explanation (CL), corresponding to the high status of experimental physics in the first decades following World War II.  Although philosophers of science soon found certain problems in that model,  most social scientists retained a populist version of it: that causal explanation is secured by universal laws that turn initial conditions into causes of predicted outcomes. This general knowledge is tested in experiments. The CL-model became a cornerstone of postwar cognitive optimism: social scientists took the experiment as the paradigm for social science, promoting a mechanical worldview and the ideal of explanation by laws.
Merton�s 1945 discussion of �sociological theory� illustrates this: He first rejects three definitions (methodology, general orientations, conceptual analysis) which are obviously only preconditions of theories, not theories as such. This leaves him with three candidates, as listed in Table 1:
Table 1. Notions of theory according to Merton
Merton�s term Characteristics Forms a Label
set of inter-
(a) Post factum Similar to clinical (1) Implicitly Case
sociological observations, explains assumed reconst-
interpretations after the event (2) No, ad hoc ruction
(b) Empirical Simple aggregation, No Induction
generalizations correlations, no
(c) Sociological Derives fresh hypo- Yes Experi-
theory based on theses to be confir- ment
scientific laws med by new observa-
tions. Explanation =
There is probably general agreement that (b) cannot stand alone. To be part of theoretical knowledge, such regularities must be accounted for by either (a) or (c). We shall later return to (a) and its two varieties. At this point, it is sufficient to note that Merton clearly takes (c) as his ideal and only proper understanding of theory (symbolized in his use of the term �scientific�). He explicitly states that it allows the social scientist �the likelihood of approximating a �crucial� observation or experiment�.  Note that observation here is equated to experiment.
However, the expansion of social research in the postwar period was not based on experimental data, rather on non-experimental data. Data from sample surveys played a major role, and the (linear) statistical methods devised to analyze such data became the master examples of how social science produced empirical knowledge. The impression, created by variables-oriented social science (VO), was that statistical treatment of such data constituted quasi-experiments. The notion of �statistical testing� became associated with the notion of an experimental test, despite the fact that these are two quite different procedures, one mainly linked to a mechanical, deterministic worldview, the other to a probabilistic view of the world.
Now, social scientists also had to reflect on the peculiarities of the particular subject: human action, interaction in social structures and the role played by knowledge. Through the first postwar decades, the dominant paradigm was here functionalism (F), relying not on physics, but on the analogy of the organism, drawn from biology.
Table 2. Varieties of sociology on a cognitive optimist basis
Approach to Label Paradigm Worldview Key features
Theory [CL] Covering Experimental Mechanical Laws/forces of nature
law natural science causes, experimental
Empirical [VO] Variab- Statistical methods Probabilistic Statistical testing,
substance les-oriented as quasi-experi- (indetermi- prediction only
statistical ments on non- nist) within confidence
methods experimental data intervals
Frame- [F] Functio- Biological notion Organic Interpenetration of
work for nalism of organism interdepen- social systems and
the analysis dence subsystems
social [RC] Ratio- Cost/benefit and Rational Formal, idealized
structure, nal choice game theory action models, intentions
knowledge notions equi- equilibria as causes
These three different components define the early postwar orthodox consensus in sociology: CL/VO/F. Its leading spokesmen were Parsons and Merton. In terms of fostering productive relations with the discipline of history, that combination was particularly unsuited. History had been consolidated in Europe through the 19th century nationalist phase. Even in the postwar period, the disciplinary prestige and identity of historians was linked to archival work and to the production of non-comparative narratives of various aspects of national developments. Most historians despised both the functionalist and the variables-oriented style of research that dominated in sociology. But even in historical methodology, the CL- model was suggested as a norm, which only goes to show how far the dominance of the experimental ideal went. 
Rather than bringing sociology on to a track of cumulative growth of knowledge, the orthodox consensus soon came under attack. Some of the main challengers are noted in this Tables 2 and 3. One major challenger shared the basic attitude of cognitive optimism, supporting at least CL, mostly also VO, but they wanted to replace the biological analogy with analogies drawn from cost-benefit economics and game theory: the rational choice alternative was CL/VO/RC. The extent to which rational choice theorists have been able to analyse empirical material in a way what satisfies their ideal of theory will be discussed below, since rational choice is the only cognitive optimist approach that has been involved in the recent debates on comparative historical sociology.
But other currents within sociology saw the criticism of the orthodox consensus as part of a broader attack on the attitude of cognitive optimism. The alternative position of cognitive skeptisism was based on very different views on explanation and empirical substance. These views pointed back to early 20th century currents that had been displaced by the Parsons/Merton attempt to create a hegemonic synthesis: notably the Chicago school of sociology based on pragmatist philosophy, and partly also the European phenomenological tradition (Husserl/Sch�tz/Garfinkel). Table 3 provides a brief overview.
Table 3. Varieties of sociology on the basis of cognitive skepticism
Approach to Label Paradigm Key features
Theory [GT] Groun- Understanding Link to pragmatism. All theory is
ded theory of diversity middle range. Any distinction between
through scientific and everyday knowledge is
comparison relative and requires an understanding
of the community of researchers
Empirical [CS] Case Explanations Empirical weaknesses of �variables�-
substance studies based on under- based studies
Framework [I] Inter- No analogy that Action is seen as situated in time and
for the actionism is external space. Accumulation of knowledge
analysis to sociology/ restricted to specific research fields.
of action, social science
structure, [ST] Social Transcenden- Claims to define the conditions of
knowledge theory tal philosophy possibility for social science. Focus on
of action conditions of knowledge, not substantive
[PS] (Post-) Linguistic Cognitive nihilism.
structura- analogy No accumulation of knowledge.
The cognitive optimist attitude gives priority to fundamental theory. Explanations are seen as derived from theory, either from explicitly established laws, or from thought experiments within idealized models (rational choice theory). The ambition is to test either concentrated empirical knowledge (laws) or internally consistent analytical models. In contrast, grounded theory starts from substantial research problems (e.g. driven by culturally significant questions), establishing conjunctions of factors to explain outcomes of interest. It opposes the preoccupation of conventional philosophy of science with the testing of preconceived theories, instead emphasizing the development of new knowledge, moving from successful explanations to careful, bounded generalizations. It proceeds �bottom-up� from knowledge of actor motivations as established by micro-sociology and ethnography mainly through various forms of fieldwork, e.g. participant observation. Even grounded theory consists of interrelated propositions, but the consistency of these sets is established at the level of the interactive interplay of actors with limited knowledge. Explanations are then used as analogies in a process of constant comparison with other interaction patterns. Such comparisons help establish the categories of the cases and allow limited generalizations. We shall discuss this in more detail later, here it is sufficient to note that the program of grounded theory emerged as a methodological reflection on case-based empirical research.
Social scientists with a cognitive skepticist attitude regard action and interaction as historical processes. There is here a link to the debates within analytic philosophy of history, where the covering law model of explanation was increasingly watered out as ever more singular events were allowed into the range of initial conditions.  These events had to be organized in temporal sequences, while the �general laws� of history faded out into the realm of tautologies. From the point of view of the experimental ideal, context became unmanageable. Researchers committed to the CL-ideal thus came to see historical explanations as �descriptive�, fully outside their conception of theory. In contrast, the cognitive skepticist attitude rather promoted sensitivity to context. In this essay, we shall thus use the term context as an alternative to the cognitive optimist notion of �initial conditions�. Context indicates constellations of particular causes not organized by laws.
Doubts about the cognitive optimist attitude spread in history. Similarly, several alternatives to nationalist history were launched: more historians would work on broader matters than what separate national archives allowed them to study, and they would deal with more contemporary topics, despite closed archives. Many historians took up social history, focusing on class, gender or minorities rather than on national developments. A more extensive presentation of the many �non-nationalist� approaches that spread within the discipline of history since the 1960s cannot be given here. It should just be noted that these historians could rely on earlier alternative traditions in history, such as parts of the German younger historical school, the French Annales school and the Anglo-American traditions that started off with the influence of Toynbee. 
These changes happened in history, while a grounded theory-based critique of the orthodox consensus spread in sociology. This opened up new options of fruitful interchange between history and sociology, based on a view of action and interaction as inherently historical. Interaction patterns could be reconstructed as sequences, that is as histories that could be understood after they happened, but not predicted in advance. This is the interactionist notion of situated action.
The turn to cognitive skepticism implies a rejection of the logic of the experiment as a norm for all science. Rather, it seems that reconstruction of events (cases) is the business of social science (Table 1). Experimenting in search of the laws of nature is a highly specific human activity, while reconstruction can be linked to a broader set of converging logics concerning the unfolding of human life: evolutionary trajectories, events that have developed, events that constitute problems that we need to diagnose in order to understand (and often improve) our present situation, or events that we need to know about in detail because we want to attribute responsibility (e.g. for a crime that has taken place). While the model role of cognitive optimism is that of the (human) natural scientist experimenting on nature, the model role of cognitive skepticism may be that of the detective, the doctor or the judge.
Let us return to our brief survey of postwar sociology. Three main scholarly views of action, interaction and knowledge share the cognitive skepticist attitude (see Table 3). The first one emerged in the U.S. We shall here deliberately define the interactionist position as broadly as possible, which means that it does not just include the 1950s �symbolic interactionism� of H. Blumer, as well as more recent positions that continue along the lines of the Chicago school, but even the ethnomethodologist school. A distinguishing feature of the interactionist position is that it uses only analogies from within its own realm, i.e. the realm of action and interaction. It also applies the pragmatist point that accumulation of knowledge is always specific to separate research fields! Interactionism combines all the elements GT/CS/I.
The remaining two positions developed as European criticisms of the orthodox consensus. By social theory we here understand a number of approaches within social philosophy, marked by their attempts to define a transcendental theoretical position, based on some kind of combination of the anti-positivistic philosophical traditions (phenomenology, hermeneutics, critical theory, etc.), which certainly sustain an attitude of cognitive skepticism.  While the idea of explanation by laws or idealized cost-benefit/game theory models constitute the �highest� or �most fundamental� level of contemporary cognitive optimist notions of theory, the impression was created that systems of �social theory� (transcendental theory of action, social structure and knowledge) was the parallel cognitive skepticist notion of theory at such a fundamental level.  This notion of theory does not satisfy the definition we gave above, since there is no accumulation of knowledge, the notion of transcendental theory simply gives a set of notions that are allegedly conditions of possibility for social science to exist at all. Thus, it is not at all clear whether social theory, as here defined, relates to GT/CS.
The final position is that of post-structuralism, which takes cognitive skepticism to the extreme, claiming that there is no foundation for objective knowledge. The only way to remain objective is to show that any approach (be it in philosophy or social science) which claims to rest on an objective position fails � on its own terms � to fulfill these claims. This is done by deconstruction, and any transcendental theory is a tempting candidate for such deconstruction. Here, even more than in social theory, there is no relation to GT/CS, since accumulated empirical knowledge is doubted.
Merton�s paradox and Skocpol�s dilemma
Our brief overview of postwar sociology may be related to the sub-field of comparative historical sociology through a paradox and a dilemma.
The paradox emerged as Merton republished his essay on sociological theory: four years after its first publication it appeared unaltered as part of his famous collection Social Theory and Social Structure (1949). But in the introduction to that collection, Merton launched the notion of middle range theories, a notion which is not at all discussed in the 1945-essay. In the third edition (1968) of the book, he expanded his discussion into a full-fledged essay on middle range theories. In his framework, these are negatively defined as theories in which the ideal of explaining by laws according to the experimental ideal (CL/VO) are not satisfied.
Through the postwar period, social research had expanded rapidly. It is tempting to interpret Merton�s increasing emphasis on a concept of middle range theories as a reflection of the fact that while the CL-ideal dominated, it was soon noted that the knowledge created was not really corresponding to the natural science ideal: social science found no solid, universal laws. What researchers declared to be laws turned out to be empty tautologies badly in need of contextual specification, while middle range theories with no secure basis in conventional philosophy of science seemed to explain quite a lot. To rescue his cognitive optimism, Merton declared that social science was still in its infancy, and that middle range theories would be replaced by full range theories � his ideal (c) � only when social science matured. Natural science had matured, but social science was at least two hundred years behind.  Judged by the experimental ideal, middle range theories were really ad hoc arguments (a2).
Merton�s paradox is a paradox within cognitive optimism: what sociologists mainly do in practical research is in Merton�s view without any justification from mainstream philosophy of science (CL). Strikingly, in none of his accounts, Merton told his readers which notion � out of those he had defined in 1945 (Table 1) � his notion of middle range theories corresponded. The same paradox between the cognitive optimist ideal and actual explanations reappeared � and was solved in the same way � when rational choice theorists coined the notion of mechanisms.  When rational choice scholars entered the debate on comparative historical sociology, they faced the challenge of bridging between the experimental ideal of explanation by laws and the empirical knowledge which seemed to be accumulated as mechanisms or middle range theories (see II in Figure 1 below).
The 1940s and 1950s saw the publication of the first works by a group of scholars that we shall call the first postwar generation of historical sociologists were published.  This was an important milestone in the relationship between social science and the tradition of nationalist history. These first generation scholars provided comparative case-studies of countries with reference to topics that had been at the core of the nationalist tradition of European history: how the separate nation states had formed. But they also addressed the dramatic 20th century consequences of this development: the struggles over democracy and dictatorship. The grand works of nationalist history were among the major inputs into this kind of historical sociological research. But these sociologists compared several cases. While the nationalist historians simply had asserted that their home countries had developed in unique ways (as e.g. in the notion of �the peculiar German development path�), the historical sociologists compared the uniqueness of the various nation states and their development along several dimensions (political mobilization, economic development, nationalism, democracy, etc), using categories drawn from social science. As social scientists, they provided detailed explanations, but were explicitly critical of at least two elements of the orthodox consensus: functionalist theory (F) and variables-oriented analysis (VO).  As for the ideal notion of theory as explanation by causal laws (CL), they remained silent � they did not want to challenge the dominant model of scientific explanation, but could easily see that they provided nothing more than middle range theories.
The 1970s saw a renaissance of such comparative historical studies, and in the early 1980s, Theda Skocpol published two essays, in which she aimed to codify the principles of comparative historical sociology.  She opposed functionalist approaches, claiming, quite to the point, that the application of such a basic analogy to empirical material did not at all yield comparative studies, only �parallel demonstration� of a theory already established. She also opposed an �interpretative� holistic account that she found in Bendix, seemingly regarding this as connected to the Chicago-school emphasis on conjunctures specified in time and space. As her own preferred approach, she suggested to follow the lead of Barrington Moore, claiming that his program was to arrive at macro-causal regularities. She chose a proto-version of the CL-ideal, John Stuart Mill�s methods of agreement and difference, mid-19th century generalizations of experimental procedures, claiming that comparative historical sociology could use these methods to derive �causal regularities in history�.
But at the same time, she claimed that macro-historical comparative studies avoided commitment to preconceived theories, which is clearly a characteristic of grounded theory: �The crucial point is that no effort is made to analyze historical facts according to a preconceived general model. (...) The investigator�s commitment is not to any existing theory or theories, but to the discovery of concrete causal configurations adequate to account for important historical patterns�.  Skocpol�s dilemma was this undecided position: a dilemma between cognitive optimism and cognitive skepticism within comparative historical sociology. She attempted to stand with one foot in the mainstream CL-camp, and the other in the interactionist, grounded theory position (GT/CS/I) � although she did not use these terms.
Three syntheses and one non-synthesis as solutions to Skocpol�s dilemma
Both Merton�s paradox and Skocpol�s dilemma are problems of linking theory and empirical substance. In the debates that followed Skocpol�s attempt to codify a methodology for comparative historical macro-studies, three syntheses were proposed (Figure 1). By synthesis we here mean, a thorough account of how concentrated accumulated knowledge may be linked to new knowledge that is generated as researchers try to explain and understand their research topics. The three syntheses rely on the well-founded social science traditions of rational choice, classical European sociology and interactionism.
One of these syntheses is rational choice theory. In recent debates on comparative historical sociology, rational choice is the only approach that sticks to the attitude of cognitive optimism. Thus, as a synthesis, we shall see that it offers a solution to Merton�s paradox, a union of CL/VO/R which claims easily to achieve the transition to specified, testable explanatory mechanisms. Social science knowledge is easily demarcated and built with reference to fundamental theory so the objective of interpreting the present has no bearing on the development of theory.
In contrast, for the two other syntheses, interpretations of the present play a major role. These syntheses both embody a cognitive skepticist attitude. Theories is seen as parts of social development processes, and since no demarcation of scientific knowledge can be absolute, the challenge of interpreting the present always impacts on theory formation in social science.
Figure 1. Contending syntheses in contemporary sociology
One of the syntheses is suggested by social theory. Social theorists insist that the criticism of the experimental ideal must be based on a transcendental notion of fundamental theory. However, since they retain the attitude of cognitive skepticism, there must be some kind of link between their notion of theory and interpretations of the present. We shall see below that attempts to establish such a synthesis has led to a number of �theories of modernity�.
Finally, post-structuralism presents an extreme version of cognitive skepticism. Criticizing � that is: deconstructing � both the transcendental notion of theory and the idea of limited accumulation of knowledge, post-structuralism ends up by defining social science as nothing but interpretations of the present. Thus, it is a non-synthesis, located only in the bottom right corner of the triangle in Figure 1.
The remaining synthesis is based on the Chicago-school, interactionist �grounded theory� approach: its aims � more modest than the two others � are to produce good interpretations of the present by linking knowledge accumulated in the relevant � given problem to be studied � special fields of social science. Unlike the two other strong notions of theory, this is a less restrictive notion of theory (in Table 3 above: GT/CS/I).
In the rest of this essay we shall discuss these four positions. The major emphasis will be on the three syntheses. We start by presenting the first interactionist responses to Skocpol�s codification, then we discuss each of the four positions.
The first interactionist response
Linking sociology and history on a cognitive skepticist basis implies that sociology should be based on history. The notion of individual or singular causation, specified as conjunctions of causal factors � portraying situated actions � emerge as the common ground for history and social science.
This general insight was the core message from one group of scholars who adressed Skocpol�s main dilemma. Skocpol�s wish to rely on �no preconceived theories� could be read as a reference to the discovery of �grounded theory�, tracing causal conjunctures of historically significant situated action. Her interactionist critics insisted that such a reference was inconsistent with Skocpol�s wish to establish causal regularities. We shall here mention three of these critics: Ragin, Abbott and Tilly. Roughly, they can all be labeled interactionists, but in Tilly�s case, some qualifications are necessary, so we shall discuss his position after the two others.
Ragin and Abbott explicitly stand in the Chicago-tradition. Both insist that case-analyses have specific merits when compared to variables-oriented analysis. Both are skilled quantitative analysts, so they do not deny variables-oriented analysis any value, but they strongly oppose a generalization of variables-oriented analysis into a �populist� worldview � �general linear reality�, as Abbott called it. 
In line with the cognitive skepticism typical of Chicago school sociology, they also rejected the experimental ideal of explanation by laws.  Both clearly give case-studies of interaction situated in time/space a privileged place as a basis for social science. Ragin explored a specific notion of causality � multiple conjunctural causation � with reference to such studies, and Abbott similarly explored narrative sequences.  Both notions indicate sensitivity for context.
While it is not hard to trace some of this as an incarnation of old Methodenstreit cleavages, one should note an important new feature: as historically oriented critics of variables-oriented studies they promote close ties between sociology and history, but they also devised alternative quantitative techniques: Ragin developed his QCA-method to facilitate comparisons of causal conjunctures not just with small, but even with medium-sized populations. This method enables researchers to study causal conjunctures with up to 200 cases (the number of welfare states in the world, for instance, is between 20 and 30). Abbott explored the application of genetic sequencing models in sociology. The development of such models has been guided by the interactionist � one may even call it ontological � view that social reality happens in time/space specific processes. 
As for the relation to Skocpol, Abbott simply discarded her approach as a wholly mistaken attempt to legitimate comparative historical sociology with reference to the false unity between variables-oriented and causal generalizations. Ragin, on his part, demolished Skocpol�s counter-posing of causal explanation (Moore) and hermeneutic understanding (Bendix).  Ragin successfully defended his early view that there was one Weberian approach to comparative historical sociology.  Both Bendix and Moore, in this view, employed comparative case studies to investigate causal conjunctures. The QCA method extends such a logic of comparison beyond the maximum five to ten cases that can be tackled in qualitative case-comparison. It also diverges from standard variables-oriented studies in that it does not rely on probabilistic models or arguments.
Tilly had made similar points even earlier, relying mainly on the network theory of Harrison White, rather than from Chicago-type interactionism.  There are, however, many important convergent features. In some of his studies, Tilly valued variables-oriented analysis and rational choice theory more positively than Ragin and Abbott. He also claims that the postwar first generation pioneers of historical sociology were much too reliant on the historians of the nationalist phase. He insists that networks within nation states should have no privileged role, and more importantly, he emphasizes that networks connect many levels.  Finally, Tilly prefers to use the term mechanisms, which was avoided by Ragin and Abbott. But he legitimates his mechanisms-oriented studies with reference to an ontology of situated action, converging with the interactionists.
These three scholars thus proposed to reestablish historical sociology on the basis of a criticism of all three main aspects of the orthodox consensus (see Tables 2 and 3 above). They proposed � in our terms � to base historical sociology on the interactionist synthesis. Before we pay closer attention to the potentials of that synthesis, we shall review the alternatives.
The rational choice synthesis
The interactionist response to Skocpol suggested that sociology should be founded on the historicity of action, a view that was consistent with the broader cognitive skepticist attitude. Rational choice theorists agree that Skocpol lacks a theory of action, but they do not accept the interactionist alternative. Neither do they accept interactionism�s questioning of the experimental, law-oriented ideal (GT vs CL). Striving to define a new consensus in social science, they the need to counter an offensive based on cognitive skepticism and interactionist perspectives on action. Their program was to maintain cognitive optimism, replacing the functionalist theory of action with one based on cost-benefit and game-theory analogies. Skocpol�s dilemma can be solved, they claim, by cutting any ties to interactionism, fulfilling instead the programme of establishing causal regularities in history.
While interactionists took the historicity of action as the basis of social science, rational choice scholars rather warned about the corrupting influence of history. In 1991, British sociologist John Goldthorpe confessed his worries of how historical sociology thrives on the work of historians.  He proposed to reinforce disciplinary boundaries. Sociologists who want to do comparative history should simply convert to the discipline of history. In the same year, American sociologists Kiser and Hechter argued that history could simply be absorbed into rational choice social science. 
The restrictive notion of theory and explanation invoked by rational choice is CL/VO/RC: the union of the experimental ideal, variables-oriented empirical analysis and rational choice theory of action. To Kiser and Hechter, the rational choice program implies a wish to transcend time and space limitations: �the mechanisms derived from general theories are generalizable � they can be used in different substantive areas and historical periods�.  This emphasis on a deductive strategy of generalization leads to a view of action that is very different from the interactionist focus on action as situated improvisation in specific contexts. The latter view is rejected as pure descriptions that can be of no relevance to science.
The challenge to rational choice theory is to integrate their theory of rational action (RC) into the CL/VO-unity. Since CL is the experimental ideal and VO studies are interpreted as quasi-experiments, the obvious solution is to treat even action as an experiment! Although Kiser and Hechter never describes their program this way, we shall argue that this is what they propose. Treating action as a natural science experiment enable them to destroy the action/context link of interactionism, combining instead their focus on action with decontextualization.
Natural science experiments can be seen as paradigms of instrumental rationality. They are based on highly concentrated scientific knowledge, involving at least one universal law. But modern rational choice theorists � Kiser and Hechter included � are not reductionists. They do not want to reduce action to laws concerning the entities of natural science. They put no faith in reductionist claims, e.g. efforts to show that the human being is �nothing but a pack of neurons�. Still, they start with a notion of explanation by laws. They follow Elster in his modification of the CL model in line with recent debates in philosophy of science,  but still claim that from lawlike causal relations, one can derive mechanisms that provide satisfactory explanations. They simply claim that Merton�s paradox do not exist!
How can we claim that this is a projection of the experimental ideal of explaining by causal laws on to action! The general formula for this is actions (intentions) as causes. The covering law perspective, even its modified version, conceives of causes as causal laws. Thus, action in social science must be analogous to the (unobservable) laws of natural science: action mediates between initial conditions and outcomes. This implies that the mechanisms whereby social science explains amounts to the whole stock of rational choice decision models. Empirical research within a rational choice perspective becomes a hunt for the �correctly specified formal model�.
Kiser and Hechter conclude that the researcher generalizes by being able to fit such models to a variety of cases, far away in time and space. This is the rational choice notion of theory as �idealized models�. These models are not accumulated empirical knowledge, but knowledge idealized through mathematics with reference to certain assumptions that make such modeling possible. They are � one might say � pure products of the human mind! Only the theorist�s assumptions are involved, and they are not empirical, they are invented and influenced by analyst�s concern to make modeling easy.
We see here how action and decontextualisation is combined. To conduct a natural science experiment, instrumental control of the initial conditions (the context) is needed. But in the notions of situated action and the historical modification of the covering law model, context becomes unmanageable, as already noted. Context messes up the clear-cut distinction between (substantial) laws and (controllable) initial conditions that characterizes the experimental situation. The rational choice fear of history�s corrupting influence is a fear of such an unmanageable context, fear of having to do ever new case-studies!
Kiser and Hechter slide from a notion of universal laws, with real-life experimental initial conditions to a notion of decision making models that are employed as actors play the role of the natural science law, bridging between initial conditions and outcomes. But Rational choice models do not represent a context of concrete experimental initial conditions. Kiser and Hechter argue that there is no explanation to be gained from historical scope conditions.  As a solution to this problem, they launch the novel notion of �abstract scope conditions�. By means of this notion, Kiser and Hechter confirm that the only experiments that rational choice theorists can really hope to conduct are thought-experiments, which rely on pure logic alone. Abstract scope conditions are defined in contrast to concrete, �historically defined� scope conditions � a notion which can be equated with the more conventional notion of initial conditions � so a real life experiment can hardly be organized on such a basis.
With reference to physics, one might say that the laws of Newtonian mechanics relate to scope conditions that are less broad (less universal) than those of Einstein�s theory of relativity. But it would not make sense to dub the scope conditions of natural science abstract. The laws of natural science are substantial, they can be tested in experiments with concrete initial conditions. In that case, scope conditions and initial conditions are not opposite concepts. As Popper emphasizes, in the natural science experiment most of the action (knowledge) is in the laws, which organize the initial � or historical scope �conditions.  The problem in sciences dealing with human action is that laws are empty.
This is probably why Kiser and Hechter define the scope conditions of rational choice theory as abstract. The procedure is well known. After all, rational choice theory is the generalization to all social science of the neo-classical economics approach that has maintained a dominant position within the discipline of economics for more than 100 years. As thought-experiments rational choice models are pure products of the human mind: they are laws with no empirical content. The rational choice scientist provide the context: one imputes the preferences held by the actors of the model. After all, the knowledge held by real rational actors should be the same knowledge as the rational choice scientist disposes of. This fits the basic attitude in cognitive optimism: theory represents the major forces of an external reality. Kiser and Hechter are very close to claiming that when rational choice theory is applied in history, it works better the more incomplete the data are. 
Among philosophers of science, the assessment of these models vary. With reference to Popperian philosophy of science, it has been argued that neoclassical economic theory is based on strategies of immunisation. Its basic theoretical principles are not exposed to falsification at all. Neoclassical models generates �armchair social science�, a kind of �model-platonism�, hampered by a dramatic gap between theory and real life experience. 
If action plays the role of law, interacting actors select the initial conditions. The knowledge involved in decision making consists of knowledge of other actors� strategies and knowledge of other contextual factors. If all actors �rational and perfectly informed players�, and/or there were no problems of coordination, interaction would lead to an optimal outcome.  But this would imply that we had arrived at the end station of the cognitive optimist journey to a fully law-based science of human interactions. We know that social science has not at all given us this knowledge. Collectively, humans have discovered many laws of nature, but the human brains � working collectively with social science and cognitive science � has yet to spell out law-based knowledge concerning how human brains with bodies interact socially. It seems that we are left with concrete cases of situated action, with �historical scope conditions�. If we regard actions as experiments, we must realize that each outcome may be specific, we do not have the certainty of a natural science experimentor that a similar outcome will occur next time.  If humans were nature only, we could expect that such a discovery would eventually come, but the activity of humans take the form of interaction in cultural and social contexts, and it has proven hard to arrive at solid universal laws on this. Cognitive optimists claim this will follow when the social sciences mature, while cognitive skepticism is based on the belief that this state of knowledge will remain.
Let us return to the rational choice focus on action as an experiment. Despite their notion of abstract scope conditions, Kiser and Hechter also want to explain specific and unique outcomes. This requires �both general models and initial conditions � that is, particular features of the specific case�. But they never try to spell out how such particular features can be combined with the �abstract scope conditions� of their models! They just state that they want to derive them from �good models�, from �abstract ideal-types�.  They list three criteria of empirical relevance, but their account of how the rational choice theorist relates to empirical material � specification of models, investigation of possible mis-specifications, followed by an alternative specification � gives us no point at which knowledge on concrete, �historical scope conditions� can enter into their scheme! If the model does not fit the data, new thought experiments are invoked, implying consideration only of abstract scope conditions. What started out as an attempt at theoretical sobering of comparative historical sociology ends up as �parallel demonstration of theory�, which, as Skocpol noted, is not really comparative at all.
The logical positivists who dominated early postwar philosophy of science surrounding the covering law model were critical of any realist ontology. Postwar trends in analytical philosophy has brought a revival for so-called scientific realism. Kiser and Hechter declare themselves as realists. In fact, they link their rational choice theory to the empirical world by means of a strong kind of realism. Using Hacking�s terms, Kiser and Hechter suggest a realism about both entities and theories:  their realism concerns not just the fundamental entities (individual actors in rational choice theory), but also their stock of rational choice models. If they did not assume that their theories represent the basic interaction patterns of social life, they would have no link to the real world at all. Their fear of cases deter them from mapping specified initial conditions, so they claim that their stock of thought experiment models is really the essence of reality.
Before philosophy was differentiated from theology, realism was a way in which religious dogmatists secured cognitive optimism, implying that theory was accumulated knowledge, organized not by the human mind, but by God.  Kiser and Hechter have, actually, encountered the criticism that their realism is so strong that it is not really secularized, it has not cut its theological roots: it treats, so Somers� criticism, humans as angels. By imposing strict assumptions, the rational choice theorist takes the role of God, one might say.  God knew the world because he had created it, the rational choice theorist creates his world of thought experiments. None of them are really in need of any empirical input.
In fact, there is a strong tradition of legitimating economic models as normative standards rather than as empirical reconstructions. Even if they have scant empirical relevance, it is claimed that neoclassical economics tell you what to do if you want to be rational, if you want to maximize some objective function, such as welfare, profits, etc. But these models are then explicitly not empirical and explanatory (they are tools used by actors manouvering in economic life), and thus breaks with Kiser and Hechter�s ambition of providing superior causal explanations.
Kiser and Hechter of course oppose any criticism of this sort, claiming that there is an empirical input to their models, and insisting that even to them, history matters.  This seems unconvincing, because of the arguments given above, but also because there is considerable internal disagreement on these matters within the rational choice school.
Other rational choice theorists, namely, have also addressed theoretical and methodological questions relating to comparative historical social science. Among these, we do not find any trace of a notion of abstract scope conditions. It is easy to show that unlike Kiser and Hechter, these other rational choice theorists imply what Merton called middle range theories.
John Goldthorpe suggests a restrictive synthesis: a unity between rational choice and variables oriented studies (VO/RC).  Although he embraces the program of causal explanation, he contrasts the determinism he finds in the explanatory program of interactionist historical sociology to his preferred probabilistic world view. Concerning the relation between sociology and history, his position is opposite to that of Kiser and Hechter: as we noted, they seem more enthusiastic about rational choice explanations the weaker the empirical input is, while Goldthorpe rather wants to ban the explanation of historical singulars from sociology. In Goldthorpe, the fear of context results in a ban on explanation of singular events. He claims that historical sociology is unable to realize how biased historians� accounts are due to the incomplete nature of their empirical material. Kiser and Hechter, in contrast, see no such problems with historical sources.
Goldthorpe�s idea for a sober union between rational choice and variables-oriented analysis is to trace certain interesting macro-level regularities (lawlike, probabilistic statements) � e.g. certain patterns of social mobility � in order to explain them by means of rational choice theory. Still, he faces a theory/explanation gap similar to that of Kiser and Hechter, since he is also trapped in the problem of having to deal with uncontrollable initial conditions. He requires �some limitations of scope� for his analysis (e.g. all industrial societies), but he is not at all specific about how this context can be precisely delimited, and has to struggle with the dichotomy between cases that fit the general theory versus cases that emerge as outliers. 
Goldthorpe�s solution is unacceptable to Elster. Although Elster has only indirectly been involved in this debate, it is interesting to spell out his position here. Unlike Goldthorpe, Elster never cared much for the variables-oriented, probabilistic approach.  As a basis for his discussion of the relation between causal and intentional explanation he specifies a �causal ontology�, according to which causality is defined by determinism, in addition to local causality and temporal asymmetry.  Thus, the unity he can aspire to is between the experimental ideal of explanation by laws and the rational choice ideal of intentional explanation (CL/RC). As we have seen, Kiser and Hechter see no problems in moving from lawlike causal relations to mechanisms. The laws they claim for social science are the laws of intentional action. But Elster has always struggled with the fundamental question of whether intentional action can be connected to the notion of causal laws. In 1998, he made explicit a distinction between laws and mechanisms that had been evolving gradually in his earlier writings.  In Elster�s terms, explanation by mechanisms are similar to explanation by middle range theories in Merton. Furthermore, his 1983 discussion emphasized the difference between causal relations and intentions as causes, and in all his writings he has explored the limits to the latter type of explanation. 
More recently, Elster published a fierce attack on the group of scholars pursuing the program of Analytical narratives.  The position of this group is close to Kiser and Hechter, but they strive to relate rational choice models to specific cases. Elster criticize their work by emphasizing two of his main distinctions � laws/mechanisms, intentional/causal explanation � even more strongly than earlier. The following two quotes shows the two horns of Elster�s dilemma. He is not willing to let go of the strong notion of theory: �rational choice explanation is the only theory in the social sciences capable of yielding sharp deductions and predictions. The theory rests on maximization theory, the beauty of which is that it typically yields a unique maximum�.  But on the other hand � clearly contradicting both Analytic narratives and Kiser and Hechter � he holds that a firm link from such a theory � formal models of interaction between intentional actors � to historically and empirically oriented studies � causal reconstructions of events that have taken place � cannot be established: �two reasons make me believe that deductive history will forever remain impossible. First, the micro-mechanisms, if and when we find them, are likely to be of very fine grain. Second, however assiduously we search the historical record, the evidence is unlikely ever to match the fineness of grain of the mechanisms�. 
To Elster, the two statements are compatible: rational choice theory is possible only with strongly decontextualized assumptions, so that most attempts to analyze real life events will violate them. Thus, explanations by mechanisms, in contrast to laws, is the daily life of the social scientist.
This is clearly the rational choice version of Merton�s paradox. There seems to be two ways out of it: betting on the strong notion of theory (CL-ideal) would lead to a convergence with Kiser and Hechter � something that Elster seems eager to resist. The alternative would be to accept a less restrictive notion of theory (middle range theories), to explain via mechanisms, in Elster�s words. This would turn Elster�s causal ontology into an explicitly social one, since the emphasis on time (temporal asymmetry), space (local causation) and determinism can in fact be seen as close to the interactionist focus on situated action. But in order to make the move, Elster would need to resist the rational choice fear of context, and he seems too fond of the analytical precision of rational choice�s idealized cost-benefit/game-theory models make such a move. Thus, his position becomes a dualist one: decontextualization plus theory versus context plus explanation by mechanisms. He could chose the solution often suggested by cognitive optimists: unity some time in the future. But when he writes that rational choice history may �forever remain impossible�, he is remarkably close to cognitive skepticism.
The cognitive optimist attitude � as noted several times � entails the idea that the researcher (or the rational choice research community) should in principle be able to develop and command a total knowledge of the social world, one that allows comprehensive explanation and understanding via the stock of rational choice models. But the three attempts to pursue this program has led to very different conclusions.
Kiser and Hechter pursues the idea of reaching full knowledge by imposing abstract scope conditions. Elster establishes the fine grain of motives and knowledge as a standard, in order to conclude that since we have no way of getting law-based knowledge on this, rational choice theorists should drop formal modeling! Goldthorpe rather prefers to reserve the cognitive optimist attitude to the contemporary situation, which can be investigated by sociologists using standard variables-oriented analysis.
These very different conclusions indicate that the rational choice program of extending the cost benefit/game theory analogy, which was a quite united approach in the 1970s and 80s, has now reached a state of internal fragmentation. But all three rational choice scholars are able to unite against all varieties of cognitive skepticism: To them, interactionists, social theorists and post-structuralists are nothing but the same descriptive relativists that disobey the rules of science. We shall see, however, that the three cognitive skepticist positions are quite different, and that they have different implications for comparative historical sociology.
The social theory synthesis
Like rational choice, social theorists embrace a fundamental notion of theory, but on a cognitive skepticist, not a cognitive optimist basis. Modern social theory starts with Parsons. In his work, a transcendental notion of theory was launched on a cognitive optimist basis. Parsons held that his theory of action provided the basic concepts about the emergent properties of social life � notions that were preconditions for any social science aspiring to develop knowledge about society.  Later versions of this argument has always been built on a cognitive skepticist basis. Key terms are system/lifeworld in Habermas, autopoiesis in Luhmann, structuration in Giddens and habitus in Bourdieu.  Figure 1 illustrates the synthesis between (cognitive skepticist) fundamental theory and interpretations of the present.
Ever since Parsons� The Structure of Social Action, a major procedure in social theory has been to summarize the work of earlier theorists, particularly the classical, European sociologists. The idea is that in these works, the conditions of possibility of studying the social sphere has been realized. They can be eked out by means of a sociology of knowledge exercise in which contradictions in earlier reasoning are related to social, political and ideological developments, while a transcendental core remains. Major empirical achievements on the part of classics such as Weber og Durkheim are largely neglected. Parsons� idea was that the sociological tradition converged on the categories of analytical realism, a �social ontology� specified in his conceptual scheme of �the action system�. Later social theorists have provided their own particular syntheses along similar lines.
Social theorists pursue the most optimistic version of cognitive skepticism. They are convinced that a fundamental, transcendental theory of action, social structure and knowledge is needed. But nearly fifty years of such social theorizing has not brought any convergence! Apart from a general agreement on the position of cognitive skepticism, there is no philosophical convergence. The relations between the different fundamental concepts of, say, Habermas, Luhmann, Bourdieu and Giddens, are hard to spell out. While the promise of social theory has been to clarify the foundations of social science, their main accomplishment has been to lead social science into ever new debates between antagonistic philosophical positions. The larger the number of philosophically skilled social scientists that enter into these debates, the more uncompromising they get.
Although their general attitude is cognitive skepticist, they end up with a problem similar to that of rational choice theorists: there is a gap between their fundamental theory and the knowledge accumulated in the various special branches of social science. As noted above, the position of cognitive skepticism implies that social science can only accumulate empirical knowledge in special areas of research. To social theorists, however, the knowledge gathered in many �applied� branches of social science is only �partial� knowledge. It seems that social theorists, like rational choice, have invested so much intellectual energy into their respective fundamental notions of theory that empirical work to both emerges as a question of description! Social theory has not been able to relate productively to the specific levels at which we can really find accumulation of knowledge in the social sciences. At these levels, it seems, researchers with very different fundamental convictions, can debate specific questions meaningfully. These debates further the growth of knowledge in that field, as long as they do not take off into the sphere of philosophy.
Like the cognitive optimist ideal of explanation by laws, the ambition of grasping the totality of conditions for social science deters social theorists from the middle range level. Still, social theorists aim at substantial analysis. As Figure 1 illustrates, the social theory synthesis is to link transcendental theories of action to totalizing interpretations of the present. Both Habermas, Luhmann, Giddens and many of their followers have been eager to develop interpretations of the present. The standard program is to develop the fundamental theory into an analysis of �modernity� defined as a historical period, the starting point of which is often just vaguely indicated, just as its geographical reach is often unspecified! As their basic transcendental notions do not represent accumulated knowledge, being derived from other theorists with explicit distance to all partial sociologies, their interpretations of the present suffer. Since the commitment to totality prevents social theorists from dealing with specific research problems, they often end up by dealing with highly abstract existential problems, such as the human predicament in the modern world. Social theory is trapped in the over-ambitious hunt for total interpretations, so typical of older philosophy of history. Their general interpretations of modernity are seldom put to any use by scholars at the various research frontiers of sociology. They may have many other advantages, such as providing world views to students and other young people, but in the light of our initial definition of theory they do not represent accumulated, explanation-oriented knowledge.
There is no denying that some of the social theorists have put much work into empirical founding of their interpretations. Given their attitude of cognitive skepticism, they all seem to accept a close relation between sociology and history. Thus, they claim that their work on modernity should be paradigmatic for historical sociology. Giddens, for instance, in his studies of the nation state, violence, the self and intimacy, have worked through an impressive amount of special studies.  But his results are vulnerable to criticism on behalf of the state of knowledge in the many special fields that he refers to: a specialist on the welfare state, or on feminist studies, will easily see that he has only scratched the surface of the research frontier in these various fields. Most strikingly, for our topic, is that Giddens� studies seldom are specific enough to rely on thoroughly comparative arguments. Even if Giddens is in crucial ways inspired by interactionist and ethnomethodological sociology, his contributions to historical sociology are in line with the social theory synthesis, not with the interactionist synthesis that we shall study below.
Similar critical surveys could be provided for Habermas and Luhmann. Habermas� early work included his study on the emergence of the public sphere in Western Europe, he later studied the legitimation problems of �late capitalism�, and even today strongly emphasizes the virtues of political economic studies.  Still, he derives an interpretation of the present from his general theory via the notions of an unbalanced institutionalization of the principles of lifeworld and system.  This analysis does involve distinctions between economic and political spheres (within the system), as well as integration and culture (within the lifeworld), but it is still far removed from the knowlege accumulated in the various special fields of social science, political economy included.
Niklas Luhmann�s work is even more marked by a gap between his transcencental systems theory and the various empirical implications he often drew. His more systematic empirical work consisted of studies in �conceptual history� (Begriffsgeschichte), which are often striking demonstrations of shifts in meaning of concepts (such as love, trust, etc), but they lack systematic comparisons and often fail to specify the links between concepts and practice. 
In this respect, Bourdieu is a border case. His philosophical background was blended by his work in anthropology, where the case-study plays a major role. We cannot here enter into a discussion of the links between his vast empirical work (in terms of empirical detail he is far the strongest among the four social theorists here mentioned) and his more abstract theoretical considerations. But it is striking that all his empirical work was on France (or the Francophone area). He never made any move to consider French developments in a comparative perspective. Thus, his contribution to interpretations of the present � in the late part of his career � were also tightly linked to the countering of both �americanization� and �europeanization� in the French context. 
The point here is not to discard all aspects of the work of these social theorists but to emphasize that their overly philosophical notion of theory is not the only one available to scholars with a cognitive skepticist attitude. Furthermore, their value in social science may not rest in their grand interpretations of the present, but in other aspects of their work. Habermas, for instance, has extensively elaborated the ethical dimension of social science. This is what makes his theory a truly �critical theory�. That aspect of his work is relevant for any social science notion of theory that accepts that the researcher herself is an actor whose knowledge is not easily demarcated from the knowledge of the actors studied.
Social theorists would counter the above argument by throwing their thrumph card of a question � �Is sociology at all possible without an answer to the basic philosophical problems�? So what if we do not want to give priority to a �final� solution at the transcendental level? It may even be that we accept the transcendental argument in principle, but that we may not wish to support one social theorists against all the others � what shall we do? Is it at all possible to defend oneself against the social theorists� claim that nobody can do without a view of totality?
What we can at least demand, is that these questions � of action, social structure and knowledge � be discussed with as close as possible a reference to interdisciplinary, substantial research in that field. Such research we find in contemporary cognitive science, which relates closely to neuroscience, social psychology, linguistics and other relevant natural and social sciences.  A strong feature of Luhmann�s work is the relation to cognitive science!
It seems that despite the considerable progress in cognitive science through recent decades, the knowledge we gain from it does not allow us to base social science on fundamental theory. Again, this leads on to the interactionist synthesis as defined in Figure 1. The capacity of the human mind allow only accumulation of social scientific knowledge in the light of fairly specific problems. Thus, we get no better interpretations of the present than those that combine accumulated knowledge from several relevant partial sociologies. This is surely heavy work.
The tradition of comparative historical sociology provides examples of shifting conceptions of cultural problems, often linked to the emergence of the modern state. For the first postwar generation of comparative historical sociologists the questions of fascism, communism and democracy were crucial. At the time of the student movement the study of revolutions and class relations emerged as new fields. Today, the Western world has been through a historical period of relatively peaceful relations, rising living standards and social differentiation. Politics of identity has attracted increasing attention. This has led both to a considerable diversification and generalization of cultural problems. As ever new groups develop the capacity to deal with their own roots, they start to accumulate knowledge on their own history and on their various specific problems. Simultaneously, the development of a more knowledge-intensive society gives rise ever more private units (firms, organizations, etc) that also command considerable (private) knowledge in their respective fields. One might speculate whether the inclination of social theorists towards purely existential problems is conditioned by this long stable, peaceful period of high living standards.
Let us consider the relation to the other syntheses. To rational choice theorists � who compare to their preferred range of cost-benefit/game theory models � social theory lacks analytic precision. Promoting to their CL/VO-commitment as the only possible approach to scientific explanation, they accuse social theorists of constantly reconceptualizing older writings, rather than pursuing testable claims.
In an interactionist perspective, rational choice and social theory emerge as two varieties of a too restrictive notion of theory, both blocking a productive relation to empirical research. To the extent that rational choice critics accuse social theory of being unable to contribute to the accumulation of empirical knowledge, interactionists would agree. But the interactionists would, as we have seen above, also question rational choice�s ability to accumulate empirical knowledge. Rational choice theorists often strike back by criticizing social theory, post-structuralism and interactionism jointly.  This is misleading, since, although they all incorporate a cognitive skepticist attitude, the interactionist notion of (grounded) theory is entirely different from the transcendental notion of theory invoked by social theorists, while the post-structuralists � as we shall briefly see � intend to deconstruct any notion of theory.
A short note on post-structuralism
Post-structuralism emerged as a response � by Foucault, Derrida, Barthes and others � to Parisian structuralism in the late 1960s. Since then, it has gained many followers worldwide. Its increased popularity is particularly based on translations spreading in Anglo-American academic circles, influencing especially the humanities, in the 1980s and 1990s. Post-structuralism cannot be subsumed under any of the three syntheses we here discuss. Quite often, post-structuralism is included as one variety of social theory, but it is distinct in the sense that it does not aim to provide a new transcendental philosophical foundation for social science. Post-structuralism rather claims that the only role an intellectual can take is to deconstruct any such fundament. Post-structuralism is a program to negate any notion of theory, and especially fundamental ones. It moves from cognitive skepticism into cognitive nihilism. Thus it also differs from interactionism, since it denies the notion of middle range theories. It takes skepticism against the accumulation of knowledge to the bitter end. Scientific knowledge cannot be demarcated in any way, it is not something that accumulates, it is always part of everyday actions. Post-structuralism see only fully discontinous knowledge regimes. It is thus left with a generalized sociology of knowledge in which any alleged social science theory is an interpretation of the present (cf. lower right corner of Figure 1). There is no notion of theory, although one may � critically � claim that post-structuralism �feeds� on transcendental social theory in a �negative� way!
Among followers of Foucault, knowledge regimes are seen as embedded in power relations. A knowledge regime is linked to the way in which institutions of power (the state at the macro level, various institutions as the prison, hospital, etc. at the meso level, and e.g. the family father at the micro-level) interpret their present predicament. �Interpretation� here indicates not primarily a broad picture of the present, but the many classificatory and standardizing schemes by which the state, the firms, and other institutions count, group and thus impose discipline on the members of society.
It is certainly important for social science to be aware of the fact that the data analyzed are often produced by the state. Even many of the categories and classifications they use have been defined by the state for practical purposes, that are often related to domination/discipline. Such a view may be highly appealing to social movements on the defensive. The more monolithic and uncompromising the established power structure seems, the more tempting it is to reduce any production of knowledge to the workings of this structure. Taken to the extreme, however, this becomes reductionism on the part of power-relations, denying any autonomy for the research collective. This reduction is not a functionalist one, it rather uses the structure of the language as a model for the impact of power/knowledge on the dominated subjects. The reliance on this one basic analogy often lead post-structuralists to overemphasize the role of cultural codes in history. the danger is that this leads to a sort of �imperialism of the humanities�,  incapable of comparative analysis and of interdisciplinary sensitivity. A more detailed assessment of whether post-structuralist studies on historical topics have avoided this danger cannot be given here.
The interactionist synthesis
Interactionism developed in postwar sociology predominantly as micro-sociology and social psychology. We have seen how Ragin and Abbott suggested a Chicago-school programme for historical sociology. Here, we shall show that also in the further development of that tradition, especially in Glaser and Strauss�s notion of grounded theory, there are elements that supports a distinct interactionist solution to Skocpol�s dilemma. This solution avoids a main problem of the two other syntheses, namely that restrictive notions of theory tend to displace comparative reasoning. The interactionist synthesis, namely, avoids both the two notions of fundamental theory that were invoked by rational choice and social theory respectively (Figure 1).
The interactionist synthesis operates at the middle range level. Unlike post-structuralism, interactionism retains a notion of theory as accumulated knowledge. But that notion of theory is less restrictive than both the experimental ideal implied by cognitive optimists and the transcendental notion held by social theorists. In line with pragmatism, interactionists claim that we can do without high level theory, but still gain partial knowledge, since humans learn and are able to relate knowledge to practical problems. We can reconstruct cases and learn from them. Such knowledge is not highly concentrated in the form of laws, but it is more than just legitimation of specific power-positions.
The implication that we should not strive for high or basic level knowledge in the social sciences, lead critics to claim that the knowledge gained in interactionist studies is purely descriptive. But anyone who reads interactionist micro-case-studies or comparatively specified historical-sociological macro-case-studies with an open mind will admit that such studies provide explanations. The claim of grounded theory is that we can accumulate somewhat more concentrated knowledge by carefully aggregating from explanations. This move makes the knowledge theoretical, but the knowledge is never cut loose from the specific context of interaction.
Scholars who maintain the CL-ideal, would object that this knowledge can not be tested. A grounded theorist would respond that alternative explanatory proposals can be tested in social science, but would doubt that alternative theories consisting of universal laws, or of idealized rational choice models can be tested. A test of alternative explanations relates to the same context. On such a basis one can generalize, but only in certain specific ways. Testing in order to investigate the degree of generality of a theory seems not to be a major way in which we accumulate knowledge in the social sciences. Rather, two other ways of generalizing play a more prominent role: case-typologies and formal grounded theory (mechanisms). We shall soon return to this topic.
Merton�s paradox was the clash between the experimental ideal of explaining by laws and the social scientist�s daily routine of explaining by middle range theories. We noted that Merton never specified into which of his proposed definitions of theory the later notion of middle range theories would fit! But it seems that post factum sociological interpretations (a in Table 1), which are diagnostic accounts of events that have happened, fit best. In Merton�s original account, however, this category had two varieties: one assumes a fully developed theory (a1), the other one is ad hoc (a2). Merton would probably much prefer to regard (a1) as middle range theories, but that does not square with the reference to a fully developed theory.
If, alternatively, we regard (a2) as middle range theory, we are close to the notion of grounded theory. Merton�s paradox implies that he does not give us any rules on how develop middle range theories. Glaser and Strauss� account of how grounded theory is discovered is much richer in this respect. A main reason is that, unlike Merton, they are not bothered by a philosophy of science ideal (CL) imported from the study of non-human nature. Glaser and Strauss� developed their ideas of grounded theory in opposition to the testing of �already established knowledge�. Emphasizing the discovery of new knowledge, they even suggested that categories should �not be contaminated by concepts more suited to different areas�.  Later, however, Strauss emphasized that one may well use existing theory from the start, provided it was also produced as grounded theory.  In this way, he confirmed that grounded theory can be accumulated into more concentrated knowledge.
Thus, middle range theories that are developed from substantial research (and many of Merton�s were!), should be seen as grounded theories. They are only tested in the sense that any explanation benefits from confrontation with other plausible explanations. In contrast, if a middle range theory is developed by application of certain basic analogies (cost-benefit/game theory in rational choice, organic interconnectedness in functionalism or structural linguistics in structuralism), it must be considered �ungrounded�. Also purely variables-oriented studies are ungrounded, since they isolate features measured by qualitative indicators (education, status, income, etc.) from the processes of social interaction.  If theories are �ungrounded�, we get parallel demonstration of theory, not comparison. A typical feature of interactionism, noted in Table 3, is its preference for analogies from within the fields studied. These analogies are used as part of a strategy of grounding the theory. Rokkan�s typological maps of Western European party systems � a major contribution to comparative macro-sociology � started from the quantitative analysis of just a couple of party-systems: the other party systems were integrated in the analysis through a method of constant comparison, increasingly focused on a qualitative tracing of contextual factors. The party system were used as analogies towards each other.
Let us return to the charge that grounded theory does not obey a common logic of inference. We have argued that grounded theory on specific conditions equals middle range theories (a2 in Table 1). Are they thus ad hoc? Ad hoc arguments are obviously absurd if we can conduct an experiment that test a preconceived theory consisting of at least one universal law from which we have derived a hypothesis: we find out directly if the hypothesis was a correct prediction or not (c in Table 1). But experiments are conducted with elements with which we cannot communicate. We cannot go back and check our deductive explanation with any of those entities involved in the study (whether these are elementary particles, the initial conditions, or the instruments used to measure the processes). In social science fieldwork, however, we can go back and check by asking, observing and interpreting. According to Glaser & Strauss, grounded theory is developed by going back and forth between concepual clarification and fieldwork. We can reconstruct social processes by refering to motives, routines and choices. The theories we discover do not hold a monopoly of knowledge vis-a-vis the �entities� they are about. In this respect, grounded theory is based on Verstehen, a notion mostly emphasized by scholars with cognitive skepticist attitudes.
But the social science researcher is not only an actor in the field. The researcher is also part of a community of researchers. The pragmatist perspective requires interactionists to combine the sociology of knowledge and the sociology of science: we must regard knowledge as related to practical purposes, and simultaneously see the scientists that accumulate knowledge about knowledgeable people as members of a collective which specializes in the generation of scientific knowledge. While this community does not possess strong, concentrated knowledge on human action, it does � as we have emphasized several times � possess partial knowledge accumulated in certain areas of research relevant to the research problems that led to the fieldwork. Furthermore, this knowledge has been organized in analytical ways.
The knowledge held by the actors may be more limited (possibly deeper, but more restricted to the actors� specific location) than that of the researchers. The researcher may find that actors are potentially wrong about important features of their situation, even that they are victims of unintended consequences caused by the way their actions are linked to other actions. Merton has many discussions of this,  and in line with his cognitive optimist attitude, he conceives of the researcher as a social engineer.  Within the interactionist synthesis, one would rather see the researcher as a participant who does not hold the law-based knowledge of an engineer, but who might give advice or in other ways try to influence the pattern of interaction. The researcher may relate to public bureaucracy, to representative institutions (trying to influence legislation) or turn to activism by becoming a member of a social movement. The ethical aspects of this situation has been thoroughly discussed in Habermas� work.
The important conclusion here is that grounded theory is more than just a reproduction of actor motivations. It may well cover reproduction of interaction patterns driven (partly) by unintended consequences. But it has no faith in the kind of generalizations about such patterns through the application of a fundamental and external biological analogy, which was what Merton suggested in his �paradigm for functional analysis�. 
We shall briefly consider the contrasts between grounded theory and the transcendental notion of theory. Social theorists would object that grounded theorists do not care about transcendental notions of action and that they cannot deal with the totality of society. One part of the interactionist response was already sketched in our section on social theory above: any theory of action and interaction should relate as strongly as possible to cognitive science.  Furthermore, we have several times noted that interactionism�s pragmatist perspective implies that no knowledge in the modern world is �total�, it is partial and practical. The social sciences do strive to establish interpretations of the present, but no such interpretation can ever be total, the best we can do is to piece together accumulated research in selected special fields of society. When launching such interpretations, we must be conscious that as researchers, we are participants in the society that we analyze. The knowledge we claim about the people we study may very well be challenged, not just by other researchers, but also by other (often more influential) participants.
But do we not still need notions of knowledge, action and interaction? This is a difficult question that we shall just briefly touch upon. Table 3 summarized some characteristics of the interactionist framework. It is interesting to note a trend towards convergence: within rational choice theory, scholars such as Boudon and Elster have for some time emphasized models of action and interaction which relax the rational choice assumption that actors� beliefs adequately represent social reality.  This can be termed a �thin� notion of rationality. Such a notion implies that the context of action and interaction must always be specified independently of the formal, rational choice models. We have already pointed out that in Elster�s work, it is possible to find an �ontology� of the social world as constituted by interaction.
As for the social theorists, their very point of departure has always been notions of action, interaction and knowledge. Given that there is little agreement on the philosophical details apart from a general focus on knowledgeable, interacting persons capable only of bounded (contextual) rationality, it makes sense to suggest that also in social theory, such notions should be kept as �minimal� or �thin� as possible. Such thin notions are consistent with interactionist grounded theory, as they imply that no notion of action and interaction can lead deductively to substantial explanations. A specification of the context � the interaction between the specific historical situation and actors� beliefs about what the situation is � is necessary. Such minimum notions of action and interaction becomes a �minimum ontology�.  The interactionist contribution to these converging notions would be the pragmatist emphasis on knowledge/action as processual. These very brief statements on complicated matters will convince neither ambitious social philosophers nor devoted rational choice theorists, but we shall not pursue such matters further here.
It has been claimed that grounded theory provides no �set of interrelated propositions�, and thus is description rather than theory. But we have already noted that it does provide such relations. They are, however, not �lifted high up� by strong external analogies: the interrelations are provided at the level of interaction. Here is a methodological focus on interacting individuals, a focus which does not impose a combined realism of entities and theories, that does not impose the whole set of rational choice models on actors, conceiving them as angels. At the most, we here have a realism of entities only.  But this is no methodological individualism, since interactionists may very well accept the collective aspects of knowledge (both everyday knowledge and the partially systematic knowledge accumulated by research collectives). Furthermore, it is problem-oriented, so it can easily establish an analysis at the level (e.g. macro) which is appropriate for the question asked.  It is only in the context of cognitive optimism�s quest for a full range theory which monopolizes knowledge on action and interaction that a decision on individualism or not becomes a burning question in methodological debates.
We can now return to the ways in which knowledge is accumulated in social science. We have doubted that social science possess much knowledge in the form of universal laws or formal rational choice models. The minimum ontology of action tells us that situated action always take the form of specific processes, but little more than that. Still, when they launched the programme of grounded theory, Glaser & Strauss claimed that social science could move from substantive to formal grounded theory. The latter cuts across substantive areas. They referred to their own study of dying as a �non-scheduled status passage�, and claimed that from such a starting point, they could proceed via comparisons to deal with the prosess of becoming a student, marrying, etc. to develop a formal theory of status passages. 
Using an example from macro-studies, one could turn from a comparative study of social revolutions to a formal theory of one factor that often appears in broader conjunctions of causes that explain revolutions: e.g. �rapid changes in living standards�. In this understanding, a formal theory would be similar to the notion of mechanisms. Such theories are generalized �modules� that can be combined in many specific explanations. They are not knowledge accumulated within specific fields, but similar components found in explanations in many fields. But if � in line with the �minimum ontology� outlined above � we regard all social situations as conjunctures of a range of specific factors, it follows that such formal theories only yield explanations when related to specific contexts. As knowledge, they are �formal�, they do not represent very concentrated knowledge. It may very well turn out that such formal theories/mechanisms are particularly useful in micro-sociology, but that they are of less use in small-n macro-studies.  There is no space here to reflect on the complicated philosophy of science questions that are involved in the scholarly discussion on mechanisms and formal grounded theory. 
While we have so far presented grounded theory with no specification of the level of analysis, we shall in the following turn to the specific features of such studies at the macro level. We shall claim see that the interactionist synthesis provides a framework for historical macro-studies. Such studies can be seen as the result of the application of a grounded theory strategy in a situation where it is impossible to conduct fieldwork.
Through fieldwork and participant observation, we can develop an understanding of actor motivation. Depending on the duration of the fieldwork, and on the extent to which additional research and evaluation is conducted, one may be in a position to make statements on unintended consequences. In the case of historically oriented macro-studies, fieldwork is evidently out of question. Furthermore, long-range consequences (including unintended ones) are immediately part of the problem. Certain historical problems emerge as important to our interpretations of the present, they become cultural problems. Particularly, this is the case when we suspect that certain events may constitute turning points or critical junctures that established important conditions for present development trends. Barrington Moore�s ambition of studying the roots of both democratic and authoritarian political systems is a typical case in point.  Even when dealing with less dramatic topics, historical sociological macro-studies often specify critical junctures with stable periods inbetween. Critical junctures thus establish historical scope conditions for regularities specific to distinct historical periods. 
In contrast, however much drama they experience in the course of their fieldwork, fieldworkers seldom have the chance of experiencing a critical juncture. The point is that even if the historical sociologist sensed that presently, the course of events was about to create a new critical juncture in her field of study, she would hardly be able to study that juncture through direct fieldwork: the kind of decision making contexts that historical sociology mostly focus on involve elites and collective actors, and in such contexts, access is hardly granted to fieldworkers. 
The challenge to a historical macro-sociology based on the interactionist synthesis is thus as follows: we possess nothing but a �minimum ontology� of action and interaction, we must pursue a grounded theory approach, but we rely on fieldwork. The entities of social science are relations between individual and collective actors, but we cannot travel back in history to ask them about their beliefs and intentions. Historical macrosociology can thus only be conducted through a working relationship with the discipline of history. Certainly, historical sociologists can do archival work themselves, and also collect all kinds of quantitative and qualitative information from as many sources as possible. Still, the scope of the questions asked, and the need to build grounded theory by means of constant comparisons, requires reliance on historical monographs, since these are bound to be main sources of knowledge on context and motives. They are in many respects macrohistorical sociology�s substitute for fieldwork. The historical macrosociologist here also has one advantage vis-a-vis the fieldworker: consulting several monographs is less time-consuming than doing fieldwork in many contexts.
Goldthorpe � as noted above � sees the reliance of historical sociology on historical monographs as a main problem for this kind of qualitative research.  Many of the controls that a researcher may pursue during fieldwork cannot be pursued by historical sociologists. But they have one important option: The interactionist synthesis emphasizes the link between the partial theoretical knowledge and interpretations of the present. It also claims that knowledge is inherently social, so it gives the sociology of knowledge an integral role in theory formation. By relying on the interactionist synthesis, historical sociologists will be able to expose historical monographs to critical reflection. The sociology of knowledge and science help us understand the phenomenon of historical revisionisms. We need only think of the generations of nationalist historians in Europe�s many nation states to understand that we need to reflect how historians are �children of their times�. Their accounts are always embedded in a knowledge regime and even if their interpretations are highly partisan in favour of certain social interests, the sociologist should be able to reflect on this and thereby avoid uncritical borrowing from historicans.
Of course, also sociologists are children of their times, they too must reflect their own position in society. But the use of constant comparisons with selected cases allow historical sociology to confront different historical revisionisms and post-revisionisms at the same time, thus evaluating critically the historians who much too often provide just one explanatory narrative with no confrontation against other potential explanations. The difference is one of degree, not a difference in kind.
The pursuit of comparisons enable the historical sociologist to engage in generalization and specification simultaneously. With reference to the interactionist synthesis, to conduct historical sociology in a comparative fashion is not a choice, it is a necessary feature! Even studies that appear to be non-comparative are implicitly comparative, and the comparisons involved are specific analogies from fields studied empirically by social science, not external ones inspired by biology, game theory or linguistics.
This, finally, lead us to the conclusion that the main way in which historical sociological macro-studies accumulate knowledge is through case-typologies. Whether the studies are small-n studies or investigate a somewhat larger number of cases (e.g. through Ragin�s �Qualitative Comparative Analysis�(QCA)-method), the major results follow as reconstructions of causal conjunctures that explain specific outcomes. We may for instance analyze a number of revolutions, outlining different conjunctures by means of constant comparison. While the conjunctures are different, they often contain many of the same factors. However, since the number of cases is small or at least fairly limited, we are less interested in isolating specific mechanisms across these conjunctures, i.e. to create formal grounded theories. One way of retaining the specificity of cases is to map them in a space defined by two or more factors. Such a procedure is followed in Rokkan�s typological maps, in which the various party systems of postwar Europe are plotted with reference to factors such as distance from Rome, landwards-/seawards economies and strength of territorial centers and city networks. 
Alternatively, an analysis of revolutions could establish types of revolutions, based on the extent to which they emerged from farily similar conjunctures. These conjunctures can then be stylized, creating what Weber saw as ideal-types. In contrast to the idealized models of rational choice, Weberian ideal types are empirically grounded. The idealization involved is based on the capacity of humans � implied by the �minimum ontology� � to reconstruct sequences. This entails a bounded rationality that generate �stylized facts�, but not a perfect rationality that generates models capable of deductive explanations of human interaction based on some kind of concentrated knowledge.
Both these precedures � regular typologies and typological maps � allow us to generalize without loosing our sensitivity to context. At the same time they represent the specificity of cases. They relate cases to contextual factors involving only a �thin� notion of actor rationality. The essence of the most reliable historical monographs is retained, but generalizing ambitions are satisfied through the display of the major factors that enter into the various causal conjuntures. By establishing typologies or drawing maps for various periods, the emphasis on the time/space-situation of human action is retained. The knowledge accumulated is not universal, it is pragmatic, based on the specific questions raised. As already Weber emphasized, his enormous edifice of typologies was constructed in order to throw light on one fundamental research problem: the specificity of the West. In order to establish that explanation � which was in effect an interpretation of the present � he had to combine many fields, which became his partial sociologies of economic relations, power, law, and religion. Had he asked other kinds of questions, he would have had to establish different typologies and connect fields of specialized knowledge in other ways.
The interactionist synthesis offers the following solution to Skocpol�s dilemma: It denies any deduction of macrocausal regularities from fundamental theories. Such regularities can only be established for specific contexts established via grounded theory. Such theory builds knowledge from the bottom up, often without any preconceived theories, but it also allows reliance on more concentrated theoretical knowledge (middle range theories, mechanisms, formal grounded theory) provided that this knowledge has been established in a grounded way. We can conclude that thanks to its avoidance of any kind of fundamental theory, the interactionist synthesis is the one that allow us to develop a historical sociology that is truly comparative.
Since Skocpol first tried to specify what comparative historical macro-sociology was all about, the debate has moved from macro to micro, and from comparison to just �theory�.  Our overview has covered all these aspects, but in the end it is crucial that we bring the discussion back to the ground, to empirical studies that are macro-oriented and comparative. We have argued that one must not be afraid of neither comparisons nor context. We have indicated that any approach based on a cognitive optimist attitude has problems in developing productive relations with the discipline of history. It is probably not by chance that few historians have chosen rational choice! Rather, it is rational choice social scientists such as Kiser and Hechter who insist that that only by importing their restrictive notion of theory, history can be saved as a science.
We have studied three approaches that all seem to accept the attitude of cognitive skepticism: social theory, post-structuralism, and interactionism. They are all much better suited to interdisciplinary cooperation with history. Historians have been attracted by the first two: some want to rely on the social theorist of their own choice (Luhmann, Habermas, Giddens, Bourdieu), while others want to rely on poststructuralism (especially Foucault). The link between history and interactionism has been less pronounced. Still, it is our conclusion that the latter link holds most promises for both history and social science. As of today, there are still too many scholars in historical macrosociology who are content with formulas like those launched by Skocpol some twenty years ago. Our discussion of the controversies triggered off by her contributions has, however, shown that the dilemma that she worked herself into can best be solved by founding historical macrosociology on the interactionist synthesis.
 Lars Mj�set, Sosiologiens framgangsm�ter, forthcoming, Oslo 2002.
 Lars Mj�set, �Realisms, Constructivisms and Environmental Sociology�, Sosiologisk Tidsskrift, 9:1-2, 2001.
 Karl R. Popper, The Logic of Scientific Discovery (1934), New York 1959, p. 59, Carl G. Hempel & Paul Oppenheim, �Studies in the Logic of Explanation� (1948), in C. G. Hempel, Aspects of Scientific Explanation, New York 1965.
 Wesley C. Salmon, �Four Decades of Scientific Explanation�, in Philip Kitcher & Wesley C. Salmon, eds., scientific Explanation, Vol. 13 of Minnesota Studies in the Philosophy of Science, Minneapolis 1989.
 Robert Merton, �Sociological Theory�, American Journal of Sociology, 50, May, 1945, 462-73, quoted from the reprint in Merton, Social Theory and Social Structure, 3rd ed. New York 1968, p. 153.
 The standard reference was Carl G. Hempel, �The Function of General Laws in History� (1942), in C. G. Hempel, Aspects of Scientific Explanation.
 A classic in this tradition is Arthur C. Danto, Analytical Philosophy of History, Cambridge 1965; for overviews, see Friedrich Eberle & Eike Hennig, �Anmerkungen zum Verh�ltnis von Theorie und Empirie�, Gesellschaft, 2, 1974, and J�rn R�sen, Rekonstruktion der Vergangenheit. Prinzipien der historischen Forschung, G�ttingen 1986.
 Cf. e. g. W. H. McNeill, The Rise of the West, Chicago 1963; William H. McNeill, The Pursuit of Power, Oxford 1982; for a brief overview of this tradition, see Lars Mj�set, �Raymond Crotty�s world history�, in Raymond Crotty, When Histories Collide: The Development and Impact of Individualistic Capitalism, Walnut Creek 2001, p. xx-xxi.
 In a wider sense, social theory is often used as equivalent to sociological theory generally. It is important to note we give the term social theory a much more restrictive meaning: it refers to the group of scholars � they can also be labeled social philosophers � which subscribes to the transcendental notion of theory.
 R. Audi, ed. The Cambridge Dictionary of Philosophy, 2nd ed., 1999, p. 925: �transcendental argument� is one �that elucidates the conditions for the possiblity of some fundamental phenomenon whose existence is unchallenged or uncontroversial in the philosophical context in which the argument is propounded.�
 Robert Merton, �On Sociological Theories of the Middle Range�, in Merton, Social Theory and Social Structure, 3rd ed., p. 47.
 Jon Elster, Explaining Technical Change, Oslo 1983, p. 24, Jon Elster, �A plea for mechanisms�, in Peter Hedstr�m & Richard Swedberg, eds., Social mechanisms, Cambridge 1998.
 A selection of these works are: Barrington Moore, jr., Soviet Politics � The Dilemma of Power: The Role of Ideas in Social Change, Cambridge, Mass. 1950; B. Moore, Political power and social theory, New York 1958; Reinhart Bendix, Social Science and the Distrust of Reason, Berkeley 1951, R. Bendix, Nation-Building and Citizenship, 2nd ed. 1964; Stein Rokkan, Citizens, Elections, Parties, Oslo 1970, C. W. Mills, The Sociological Imagination, New York 1959. Theda Skocpol, editor, Vision and Method in Historical Sociology, Cambridge 1984 is an influential collection of presentations of some of these scholars.
 Barrington Moore, �Sociological theory and contemporary politics�, American Journal of Sociology, 61: 2, 1955, pp. 107-15.
 Theda Skocpol and Margaret Somers, �The Uses of Comparative History in Macrosocial Inquiry�, Comparative Studies in Society and History, 22:2, 1980, Theda Skocpol, �Emerging Agendas and Recurrent Strategies in Historical Sociology�, in T. Skocpol, editor, Vision and Method in Historical Sociology, Cambridge 1984.
 Skocpol, �Emerging Agendas and Recurrent Strategies in Historical Sociology�, p. 373.
 Andrew Abbott, �Transcending general linear reality� (1988), in Abbott, Time Matters, Chicago 2001. Such a skepticism was originally launched in Herbert Blumer, �Sociological analysis and the �variable�� (1956), in Blumer, Symbolic Interactionism, Berkeley 1969.
 A. Abbott, �The Causal Devolution� (1998), in Abbott, Time Matters.
 C. Ragin, The Comparative Method, Berkeley 1987; C. Ragin, Fuzzy-set social science, Chicago 2001; A. Abbott, �From Causes to Events� (1992), in Abbott, Time Matters.
 A. Abbott, �Of Time and Space: The Contemporary Relevance of The Chicago School�, Social Forces, 75:4, 1997.
 Ragin, The Comparative Method, A. Abbott, �History and Sociology: The Lost Synthesis� (1991), in Abbott, Time matters.
 Charles Ragin & David Zaret, �Theory and Method in Comparative Research: Two Strategies�, Social Forces, 61, p. 731-754; Skocpol, �Emerging Agendas and Recurrent Strategies in Historical Sociology�, p. 378.
 Charles Tilly, Big Structures, large processes, huge comparisons, New York 1984.
 Charles Tilly, �Means and Ends of Comparison in Macrosociology�, Comparative Social Research, 16, 1997.
 John Goldthorpe, �The uses of history in sociology: reflections on some recent tendencies� (1991), in John Goldthorpe, On Sociology, Oxford 2000.
 Edgar Kiser & Michael Hechter, �The Role of General Theory in Comparative Historical Sociology�, American Journal of Sociology, 97:1, 1991, 1-30.
 Edgar Kiser & Michael Hechter, �The Debate on Historical Sociology: Rational Choice Theory and Its Critics� American Journal of Sociology, 104:3, 1998, p. 796.
 Elster, Explaining Technical Change,p. 26.
 Kiser & Hechter, �The Debate on Historical Sociology�, p. 797. The term �scope conditions� is not a standard term in the mainstream CL-literature. Among the most eager adherents to the CL-approach in social science, attempts have been made to distinguish between initial conditions and scope conditions, see Lee Freese, �The Problem of Cumulative Knowledge�, in L. Freese, editor, Theoretical Methods in Sociology, Pittsburgh 1980, p. 34. This notion seems quite parallell to Kiser & Hechter�s notion of abstract scope conditions, which we present below. The notion of historical scope conditions corresponds to what we defined as context in the cognitive skepticist framework, and thus also to initial conditions in the cognitive optimist framework.
 Karl R. Popper, The Open Society and its Enemies, London 1942, p. 264. Cf. also Lars Mj�set, �Theory, Understandings of in the Social Sciences�, in Neil J. Smelser & Paul B. Bates, chief editors, International Encyclopedia of the Social and Behavioral Sciences, Amsterdam 2001.
 Kiser & Hechter, �The Role of General Theory in Comparative Historical Sociology�, p. 10: �When data are fragmentary and hard to come by � as often in the case of comparative-historical research � only a theory with high analytic power, and thus low input requirements, can be tested.�
 Hans Albert, �Modell-Platonismus: Der neoklassische Stil des �konomischen Denkens in kritischer Beleuchtung�, in Albert, Marktsoziologie und Entscheidungslogik, Neuwied 1967, Marc Blaug, The Methodology of Economics, Cambridge 1980. � Kiser and Hechter tries to rely on Lakatos� �post-popperian� philosophy of science, emphasizing the unfalsifiable core of certain theories in order to make a virtue of such platonism.
 Elster, Explaining Technical Change, p. 78-79.
 There is also the ethical problem that we probably would not like to live in a society where some humans possessed cpncentrated knowledge (laws) that allowed them to predict the actions of all human beings.
 Kiser & Hechter, �The Debate on Historical Sociology�, p. 805.
 Ian Hacking, Representing and intervening, Cambridge 1983, p. 27 ff.
 Veblen�s writings on the sociology of economic science is a fascinating development of this point, cf. Thorstein Veblen, The Place of Science in Modern Civilization, New York 1919, and they were developed further by Swedish institutionalist �kerman, cf. Johan �kerman, Das Problem der sozial�konomischen Synthese, Lund 1938, and Lars Mj�set, �Johan �kerman�s dualistische Synthese�, pp. 31-55 in Gottfried Eisermann, Geoffrey M. Hodgson, Lars Mj�set, Johan �kerman�s �Das Problem der sozial�konomischen Synthese�, Vademecum zu einem Klassiker des skandinavischen Institutionalismus, D�sseldorf 1997.
 Margaret Somers, ��We�re No Angels�: Realism, Rational Choice, and Relationality in Social Science�, American Journal of Sociology, 104:3, 1998, pp. 722-84.
 Kiser & Hechter, �The Debate on Historical Sociology�.
 John Goldthorpe, On Sociology, Oxford 2000.
 John Goldthorpe, �A Response to the Commentaries�, Comparative Social Research, 16, 1997; Goldthorpe, On Sociology, p. 23 & ch. 8 for instance grapples with Sweden as an outlier case.
 Elster, Explaining Technical Change, p. 44-48.
 Elster, Explaining Technical Change, p. 27-30.
 Elster, �A Plea for Mechanisms�.
 Elster, Explaining Technical Change, p. 75-88.
 Elster, �Rational Choice History: A case of Excessive Ambitions�, American Political Science Review, 2000.
 Elster, �Rational Choice History: A case of Excessive Ambitions�, p. 693.
 Elster, �Rational Choice History: A case of Excessive Ambitions�, p. 694.
 This is a permanent feature of his work, from Talcott Parsons, The Structure of Social Action, New York 1937 to Talcott Parsons, Action Theory and the Human Condition, New York 1978. � The neofunctionalism developed by Jeffrey Alexander, Richard M�nch and others reestablish Parsonian sociology on a cognitive skepticist basis. � The critical remarks that follow in this section may well be seen as a variation on themes introduced in Mills� famous attack on Parsons in The Sociological Imagination.
 J�rgen Habermas, Theorie des kommunikativen Handelns, Frankfurt a.M. 1981, Niklas Luhmann, Soziale Systeme, Frankfurt a.M. 1984, Anthony Giddens, The Constitution of Society, Cambridge 1984, Pierre Bourdieu, Le sens pratique, Paris 1980.
 Anthony Giddens, The Nation State and Violence, Cambridge 1985, Anthony Giddens, Modernity and Self-Identity, Cambridge 1991, Anthony Giddens, The Transformation of Intimacy, Cambridge 1992.
 J�rgen Habermas, Strukturwandel der �ffentlichkeit, Neuwied 1962, J�rgen Habermas, Legitimationsprobleme im Sp�tkapitalismus Frankfurt a.M 1973; J. Arnason, �Globalism and Traditions: Interview with J�rgen Habermas�, Thesis Eleven, 63, 2000, p. 1-10.
 Habermas, Theorie des kommunikativen Handelns.
 Niklas Luhmann, Liebe als Passion, Frankfurt a. M. 1984, Niklas Luhmann, Gesellschaftsstruktur und Semantik, 3 vols, Frankfurt a.M. 1993.
 Pierre Bourdieu, Sur la t�l�vision, Paris 1996, and his various criticisms of the monetary unification via the Euro in the late 1990s.
 Cf. e.g. G. Lakoff and M. Johnson, Philosophy of the Flesh, New York 1999, a study that contains illuminating analyses of questions also grappled with in the more philosphical discussions of action in the work of the social theorists.
 This is evident in both Kiser & Hechter and in Goldthorpe.
 J. G. Merquior, From Prague to Paris, London 1986, s. 210.
 Barney G. Glaser & Anselm L. Strauss, The Discovery of Grounded Theory, New York 1967, p. 37.
 Anselm Strauss, �Discovering New Theory from Previous Theory�, in Tamotsu Shibutani, editor, Human Nature and Collective Behavior. Papers in Honor of Herbert Blumer, Englewood Cliffs 1970, p. 46-53.
 Andrew Abbott, �What do cases do?� (1992), in Abbott, Time Matters.
 Cf. even his famous early paper, which does not employ the functionalist analogy: Robert K. Merton, �The Unanticipated Consequences of Social Action�, American Sociological Review, 50:6, 1936.
 Robert K. Merton, �Manifest and latent functions�, in Merton, Social Theory and Social Structure, 3rd ed., p. 135.
 Merton, �Manifest and latent functions�, p. 104 ff.
 Barry Barnes, Understanding Action, London 2000.
 Raymond Boudon, �Social Mechanisms Without Black Boxes�, in Peter Hedstr�m and Richard Swedberg, editors, Social Mechanisms, Cambridge 1998. Elster�s switch from a restrictive (CL-based) to a less restrictive notion of mechanisms was related to his work on emotions, cf. Jon Elster, Alchemies of the Mind, Cambridge 1999. The question is, however, to which extent Elster�s mechanisms still have a priviledged relation to the basic cost-benefit/game theory analogies, which in the interactionist perspective would be external analogies.
 Andrew Abbott, �Things of Boundaries� (1995), in Abbott, Time Matters.
 Hacking, Representing and intervening.
 For a similar view by a prominent rational choice theorist, see James S. Coleman, Foundations of Social Theory, Cambridge, Mass. 1990, p. 5.
 Glaser & Strauss, The Discovery of Grounded Theory, Ch. IV.
 The recent �historical institutionalist� approach in macro- and meso-oriented sociology and political science advocates generalization with reference to mechanisms, while comparative strategies play a relatively minor role. Cf. e.g. James March & Johan P. Olsen, Rediscovering institutions, New York 1989; Paul Pierson, �Increasing Returns, Path Dependence, and the Study of Politics�, American Political Science Review, 94:2, 2000. There are crucial overlaps (not often realised) with the interactionist synthesis. The degree of overlap depends on the extent to which the various historical institutionalists give priority to evolutionary analogies: the less restrictive they are in this sense, the greater the overlap. There is no space for a further discussion of this here.
 For instance, it would be necessary to discuss more thoroughly Glaser & Strauss� argument that formal grounded theories are established via constant comparisons. Furthermore, Burawoy�s promotion of his own �extended case method� as an alternative to grounded (formal) theory is relevant in this context, cf. Michael Burawoy, �Reconstructing Social Theories�, Ch. 2. in Burawoy, Ethnography unbound, Berkeley 1991. We shall here only note our disagreement with Burawoy�s claim that grounded theory is based on acceptance of both variables-oriented analysis (p. 303) and the covering law model (p. 8). As we have shown, it is developed in polar opposition to both. As for Burawoy�s extended case method, it should be seen as a kind of, rather than as an alternative to grounded theory. This is so, at least if our argument below is accepted, that comparative macrohistorical sociology utilizes a grounded theory procedure, since obviously, such macrosociological studies are needed to establish the macro-context which Burawoy must include in order for his case-analyses to be �extended�!
 Moore, Social Origins of Democracy and Dictatorship.
 E.g. in welfare state studies, studies of economic policy regimes, or more generally in the recent rush of studies of �varieties of capitalism�, cf. e.g. Robert Boyer & Yves Saillard, Regulation Theory. The State of the Art, London 2002. It should also be noted that there is a long tradition for case studies in economic history and business studies. Alfred Chandler, The Visible Hand, Cambrigde, Mass 1977 and further work by him and his students on the period of U.S. managerial capitalism is particularly significant, since these case-studies was a mai point of departure for Oliver Williamson�s work on transaction costs within rational choice theory.
 A rare case is G. Ross� study of the negotiations of the EU Maastricht treaty in 1991, through which he followed the staff of the Commission president as a participant observer: George Ross, Jacques Delors and European Integration, Cambridge 1995.
 Goldthorpe, �The Uses of History in Sociology�.
 Lars Mj�set, �Stein Rokkan�s thick comparisons�, Acta Sociologica, 43:4, 2000.
 The rational choice response pointed to the micro level, Somers� response to Kiser & Hechter pushed the discussion far into the philosophy of science.