the contextualist approach to social science methodology · social science methodology ... the...

[09:52 24/11/2008 5247-byrne-ch02.tex] Paper: a4 Job No: 5247 Byrne: The SAGE Handbook of Case-Based Methods Page: 39 39–68

2The Contextualist Approach to

Social Science Methodology

L a r s M j ø s e t

When delimiting a case, we start from aproblem, then select a process towards anoutcome and finally define a context inwhich it takes place. We explain by tracingthe process within the context. These arethe three basic operations of a contextu-alist methodology. This chapter provides adetailed comparison of this methodologywith two other – the standard and thesocial–philosophical – methodologies. Thiscomparison emphasizes how important it isthat researchers in the qualitative traditiondo not simply subordinate their reflection onthe conduct of case studies to either of theother two methodologies. It also generatesmore general lessons on how we should thinkabout methodologies, theory and accumulatedknowledge in social science.

THE METHODOLOGIST’S DILEMMA

Methodologists are scholars who draw onthe professional philosophy of science to

develop general methodological guidelinesbased on experience with certain kinds ofresearch method. They direct students toselected professional philosophical literaturesabout what science is, and spell out howthe main research methods produce resultsthat confirm these principles. Methodologistsmediate between the professional philosophyof science and specific technical routinesof empirical analysis. Methodologists mostlycome in disciplinary varieties. Their role isvery authoritative and clear-cut, as they playa crucial role in the socialization of upcomingscholars.

Methodologists face a dilemma. In thefollowing, we present this dilemma withspecific reference to social science methodol-ogists, who present methodologies in whichtheir experience in using certain types ofsocial science method is generalized withreference to selected philosophical principles.These methodologies have to appear highlyconsistent, as they reflect and influenceviews on what should pass as science in


40 THE SAGE HANDBOOK OF CASE-BASED METHODS

various disciplines. But they can never behighly consistent because they are formulatedby social scientists who are not professionalphilosophers but are specialized in specificresearch methods and related notions oftheory.

The methodologist’s mediating rolebetween philosophy and empirical researchis the key to this dilemma: Philosophicaldiscussions draw towards fundamental,eternal questions, whether they are calledquestions of metaphysics, ontology, episte-mology or just ‘meta-theory’. However,any piece of empirical research must becarried out, reported and defended hereand now. Methodologists cannot matchthe professional philosophers. Rather, theirtask is to produce practical philosophies ofsocial science – methodologies that provideconcrete guidelines in research practice andcriteria to what knowledge shall count asscience. Such practical demands requiremethodologies to appear as clear-cut andconsistent as possible. However, if the linkwith philosophy were to be pursued system-atically, all kinds of reservations, reflectionsand conceptual specifications would threatenany clarity whatsoever. In the end, most coremethodological concepts become vague,with many proximate, even contradictorydefinitions.

We should not think about methodologyas a question of one set of impeccablenormative principles. Instead, we will takefor granted that a community of socialscientists exists, and that this communitydoes not allow just any kind of knowledgeto pass as science. Still, the fact thatthere are several schools of philosophy, andvarious research techniques in social science,makes it likely that there is more than onepractical philosophy of social science. As anymethodology is an uneasy balancing act, it isnot easy to single-out sets of preconceptionsthat define families of methodologies inrecent social science. Still, we make anempirically based attempt, claiming thatthese methodological standards are shared bybroad groups of contemporary social sciencemethodologists.

THREE PRACTICAL PHILOSOPHIES OFSOCIAL SCIENCE

Methodological controversies are clues todifferent methodological frameworks. Scan-ning older and recent controversies, we findthree broad clusters of methodologies (Mjøset2006b). However, we also need to rely on soci-ology of science concepts. We have definedthe concept of a methodologist, and othersconcern disciplinary communities, methodscommunities, local research frontiers, andthe external and internal relations of science.Whereas our preferred technical term for whatmethodologists do is ‘practical philosophiesof social science’, we also use other, shorterlabels, such as ‘methodological frameworks’,‘views’, ‘positions’, ‘approaches’, ‘researcherattitudes’ or – following Veblen (1919) –‘preconceptions’. We also use personalizedterms, such as ‘standard scholars’, ‘socialphilosophers’ and ‘contextualists’.

The three practical philosophies of socialscience are determined ‘from above’by philo-sophical orientations, and ‘from below’ bythe actual research methods their proponentsemploy. Looking ‘upwards’, methodologistscan orient in three different directions:towards the professional philosophies ofnatural science, social science and the human-ities, respectively. Looking ‘downwards’ tothe everyday practices of social research,the methodologist will basically be familiarwith either one of the clusters of empiricalmethods indicated in the lower part ofTable 2.1. Our main claim, then, is thatelectoral affinities between the downwardsand upwards orientations constitute the threepractical philosophies of social science.

The practical philosophies are not ‘fun-damental’ positions taken by professionalphilosophers, they are loose summaries of theviews promoted by three types of method-ologist. There may be both internal debateswithin each cluster, as well as typical debatesbetween them. Our threefold distinction formsan empirical sociology of knowledge-basedframeworks that might be useful as a meta-perspective on earlier and contemporarymethodological debates in social science.


THE CONTEXTUALIST APPROACH TO SOCIAL SCIENCE METHODOLOGY 41

Table 2.1 The three practical philosophies of social science

The philosophy of the natural sciences:twentieth-century traditionsemerging from logical positivism,Popper’s critical rationalism,Lakatos’ research programmes,Kuhn/Popper debate, analyticalphilosophy

The philosophy of the socialsciences: US pragmatism,European critical theory,standpoint epistemologies

The philosophy of the humanities,phenomenology, hermeneutics,structuralism, post-structuralism

The standard attitude The contextualist attitude The social-philosophical attitudeMathematical modelling. Thought

experiments/simulation. Statisticalanalysis of large data-sets. Thesemethods indicate a methodscommunity with the natural sciences

Qualitative methods implying director indirect involvement with thecases studied; ranging fromlong-term participantobservation, more or lessstructured interviewing,comparative work on distinctcase-histories. These aremethods that are distinct to thesocial sciences

Interpretative analysis of texts: formallinguistic and narrative analysis,discourse analysis, history ofconcepts, content analysis, lessformal methods of textual exegesis,use of classical texts in social theoryto build broad “philosophy ofhistory”-like interpretation of thepresent. These methods indicate amethods community with thehumanities

Two of these three clusters have becomevisible from time to time in ‘struggles onmethodology’ such as the critique of posi-tivism in 1960s sociology, the science warsin 1990s science studies or the ‘perestroika’debate in recent US political science. Thethird (contextualist) alternative has mostlyappeared as a less obvious third position.

Methodological debates have been immi-nent since the birth of modern social science.Historically, these controversies display apattern of recurrence: most of the topicsdebated today have appeared, disappearedand reappeared many times since the for-mation of modern social science morethan hundred years ago. In some periods,one methodological approach has held adominant position, gaining normative status.In other periods, some kind of dualismhas prevailed, namely a dualism betweengeneralizing (‘nomothetic’) natural sciences,and specifying (‘ideographic’) social/humansciences.1 Our contextualist third position hasbeen less frequently identified and discussedin its own right. One of the aims of this chapteris to argue that it is a most important one forsocial science.

The question here is not just the differencebetween methods used, neither is it solelyabout fundamental philosophical differences.

As the combination of philosophical positionsand research methods defines each practicalphilosophy, the relation between any pairof these cannot be reduced to conventionaldichotomies. For instance, the distinctionbetween standard and contextualist attitudesis not identical to the distinction betweenquantitative and qualitative methods, nor tothe philosophical distinction between realismand nominalism (constructionism). Rather,such traditional dichotomies are interpretedin specific ways within each framework.More than consistency, each framework has acertain logic in terms of how various famousdichotomies (realism/constructionism, real-ism/empiricism, explanation/understanding,etc.) are discussed.

The philosophical inputs to a methodology,as well as the mere intellectual energyrequired to master specific social sciencemethods, make it hard for one researcher sim-ply to switch between practical philosophies.In everyday academic life, methodologicalclusters mostly remain self-contained, espe-cially in disciplines largely dominated by justone framework. Even in multi-methodologysocial sciences (sociology, above all) themethodological camps mostly keep to them-selves. Sometimes, this results in a situationin which scholars with different attitudes



grapple with very similar problems withoutever discussing (at least not in a seriousmanner) the arguments of those who belongto other clusters.

We now turn to a more detailed overview,first sketching the standard and social philo-sophical methodologies. We then introducethe third, contextualist attitude, defining itsspecificity in comparison with the other two.We specify five aspects of each practicalphilosophy:2 (1) their master example ofexplanatory logic; (2) their popularization offundamental metaphysical questions; (3) theirexplanatory priorities given the autonomy ofsocial science; (4) their implied sociologyof knowledge; and (5) their assumptionsabout the relationship between the sciences.Furthermore, we define two notions of theorywithin each methodology, showing how eachof these notions imply distinct strategiesof specification and generalization, thusquestioning any unspecified generalization/specification-dualism.

THE STANDARD PRACTICALPHILOSOPHY OF SOCIAL SCIENCE

In earlier incarnations, the standard attitudewas entirely dominant in the early postwarperiod; it still seems to be the dominant frame-work. It emerges when practical experiencesfrom mathematical modeling and statisticalanalysis of large datasets are combined withselected items from the professional philoso-phy of the natural sciences.3 Its five charac-teristics are:

(1) Master example of explanatory logic: Thereis one logic of explanation. The most directversion of this logic is found in natural scienceexperiments. Natural scientists treat ‘nature’as something that can be rearranged forexperimental purposes. But in social science,the object (‘nature’) is a society of humans,into which experimental intervention is mostlyundoable. Whereas social science knowledgedoes not – in any significant sense – grow asa result of actual experiments and modellingrelated to experiments, experimental logic isstill held to be the best example of the kind

of explanatory logic pursued. Thus, standardmethodology regards statistical analyses of non-experimental data to be quasi-experiments,considers mathematical modelling as thoughtexperiments, or – as a minimum – employsconcepts originating from the conduct of exper-iments: dependent and independent variables.The experiment becomes the paradigm forreasoning both about past events that werenever produced as experimental outcomesand about ongoing processes into whichexperimental intervention is impossible. It isnearly always implied that the experimentalbenchmark applies in indirect and modifiedways. A sequence of internal methodologicaldebates revolve around the modification of thisideal; these concern the epistemological statusof thought experiments, and how statisticalanalysis of non-experimental data can emulatereal experiments (e.g. Lieberson 1985).

(2) Popularization of fundamental metaphysicalquestions: The standard attitude is basedon a broad set of convictions that gaineddominance in the early postwar period. It laterwent through several revisionist interpretations.One of its early roots was interwar logicalpositivism, which clearly pursued an anti-realistprogram (Hacking 1983), aimed at abolishingany metaphysics (any notion of unobservables,and of causality). But this purely syntacticdefinition of theory was philosophically hardto defend. As philosophers turned to semanticnotions of theory, the question of representationcould not be avoided. The main thrust sincethe late 1950s has been towards secularizedapproaches to the core metaphysical questions,mostly labelled scientific realism. Currently,there is something close to a consensus onthis view. Scientific theories represent inherentstructures, unobservables that lie below orbeneath the flow of empirical events. The basicentities of this structure are referred to asthe elementary particles or ‘atoms’ of socialscience. Internal disagreements relate to thenature of such a generative structure, e.g.whether it is based on individual beliefs anddesires, or rather on systems of unintendedconsequences that cannot be derived fromindividual intentions.

(3) Explanatory priorities given autonomy of thesocial science realm: Social science theoryrepresents the inherent structures of the realm ofaction and interaction. Reductionism (wherebyall of society would be made part of nature only)



is out of the question, at least in practicalterms (Elster 1989, p. 74). The social sphereis, or must be treated as, marked by emergentproperties vis-à-vis the realm of natural science.Neither concepts of utility (beliefs and desires)nor concepts of generative processes needto be further reduced to entities studied byneuroscience, biology or chemistry. The internaldiscussion concerns the kinds of explanatoryreduction – to individual or non-individualentities – one can pursue within the sciencesthat cover the realm of the social.

(4) Sociology of knowledge: Scientific beliefs arestabilized only from inside the scientific com-munity, and this is enough to demarcate sciencefrom non-science. As a research community,social scientists are driven by the urge toilluminate ever more of the underlying structuresbehind social events and regularities. Influencesfrom outside this community can certainly befound, but they are not relevant to the growthof social scientific knowledge. A sociologyof (scientific) knowledge is irrelevant in thisrespect. ‘The context of discovery’ is no topic forthe philosophy of science, and only logic mattersin the ‘context of justification’. Most statementsof the standard programme emphasize thatits ambitious theoretical ideals have not yetbeen realized in actual empirical research.Worries about this gap between theory andempirical research have haunted spokesmenof the standard framework since the earlytwentieth century.

(5) Assumptions about the relation between thesciences: Compared with the natural sciences,social science is still young and immature,encountering several kinds of barrier in its effortsto mature. A sociology of scientific knowledgemay be invoked to explain this. Disagreementsrevolve around how this is explained, whetherit is simply due to younger age of the socialsciences or due to the nature of their subjectmatter.

These five features do not go together in aconsistent system. Still, they appear again andagain in the writings of methodologists with astandard conviction. The preconditions can bespecified historically. For instance, the nom-inalistic inclinations were obvious in earlypostwar operationalism, when Vienna schoollogical positivism was still broadly influen-tial. Although one can distinguish different

varieties of the programme, the above spec-ification is sufficient for our purpose.

Within the standard approach, we find atleast two distinct types of theory (Mjøset2005). The idealizing notion conceives theoryas thought experiments using mathemathicalequation systems, investigating the implica-tions (in terms of equilibrium or disequi-librium) of assumptions on actors and theirinteraction. The law-oriented notion emergesfrom attempts to find law-like regularitiesin datasets or from the use of qualitativedata in ways that allow the researcher toinvestigate hypotheses about such regulari-ties. One version of the law-oriented notionis a regression equation calculated from alarge dataset, yielding the net effects ofthe relevant independent variables on thechosen dependent variable. Another versionis what Merton (1968) called middle rangetheory.

Throughout the 1990s, many syntheses(e.g. Goldthorpe 2000) were suggestedbetween the law-oriented and the idealizingnotions of theory. But recently, more emphasishas been placed on problems in both of thesecomponents. These intenal debates have ledto what we shall call a revisionist standardposition, such as, for example, Hedström’s(2005) programme of analytical sociology andPawson’s (2000) programme on middle-rangerealism.

The notion of causal mechanisms is crucialto this revisionist position, which developsfurther the ambivalence towards high theorythat was already built into Merton’s notion ofmiddle-range theory (Mjøset 2006b, p. 339f ).Given the autonomy of the realm of socialinteraction, mechanisms define the inherentstructure, and representation of this is countedas a satisfactory explanation. Mechanismsare elementary particles or driving forces.Elster (1989, cf. Elster 2007, p. 32, Hedström2005, p. 25) conceived mechanisms as ‘acontinuous and contiguous chain of causal orintentional links’ between initial conditionsand an outcome. Hedström (2005, p. 23) statesthat we ‘explain an observed phenomenon byreferring to the social mechanism by whichsuch phenomena are regularly brought about’.



As we will see, this standard concept ofmechanisms addresses some of the samechallenges as the contextualist notion of aprocess tracing.

THE SOCIAL-PHILOSOPHICALPRACTICAL PHILOSOPHY OFSOCIAL SCIENCE

Historically, the social-philosophical positionwas the first one to challenge the standardview in postwar debates. In Germany, second-generation Frankfurt school philosopherschallenged Popper’s attempt to transcend log-ical positivism. In Britain, philosophers con-cerned about Wittgenstein’s ‘linguistic turn’introduced the continental hermeneutic andphenomenological traditions into English-language methodological debates. AroundWestern Europe and the US. local varietiesof this effort countered the dominant standardmethodology, which was seen as an improperprojection of natural science principles onto the sciences of man. The leading figuresbelonged mostly to the humanities (especiallyphilosophy), but through the 1960s and 1970s,social science and the humanities were oftenclose allies in pursuing these arguments.

Social philosophical methodology can bedrawn from overview works presenting vari-ous personal contributions to social theory.4

To the extent explicit methodologies are pro-duced, they are treatises on the methodologyof discourse analysis, of conceptual history,etc (see Table 2.1). In the following, wepresent a stylized account, using the same fiveproperties as in the standard case.

Master example ofexplanatory logic

Interpreting the standard view as a logicof mechanical, causal explanation, socialphilosophers emphasize that in human inter-action the element of meaning cannot beignored. Social actors are reflexive. Expla-nations in social science must therefore bebased on a logic of understanding meaning.

Disagreements revolve around how commonthese intersubjective meanings are: do theyrelate to small communities or are theybroad discourses that the researcher cantap into in her capacity of being a partic-ipant in (e.g. Western) society or cultureat large?

Popularization of fundamentalmetaphysical questions

Whereas the standard position today in broadterms converges on scientific realism, thesocial philosophical position has a similarrealism/nominalism debate, but with nofull convergence around a constructionist/nominalist position. The social-philosophicalapproach specializes in fundamental ques-tions. The leading question is how socialscience is possible. Transcendental notionsof action, interaction, knowledge and struc-ture are necessarily assumed by anyonewho conducts empirical social research. Thephilosophical discussions about conditions ofpossibility are linked to empirical questionsby means of broad concepts characterizingthe state of the present social world: themost frequent core concept is modernity(cf. e.g. Habermas 1981, Giddens 1990), butother periodizing labels are also invoked.Modernity is mostly interpreted as a regimeof knowledge. There is, however, also amaterialist interpretation in which the core ofpresent-day society is seen as an underlyingstructure of unintended consequences. Thesedriving forces can be conceived in line withe.g. Marx’s analysis of the cycles and trendsof the capitalist mode of production. Still,there is a sense in which even the materialistinterpretation requires an ‘understanding’ ofcapitalism in its totality.

Explanatory priority given theautonomy of the socialscience realm

The exploration of transcendental condi-tions of social science is social philoso-phy’s demonstration that the study of the



social realm must be considered in its ownterms. Contrary to the standard reduction tobasic social entities, the social-philosophicalattitude mostly implies a holistic position.What drives modernity is either a regime ofknowledge (e.g., the rationalizing logic ofmodernization) or underlying driving forces(e.g., capitalist mode of production), whichimplies specific ideological patterns. Thus,a macro- to micro-connection is mostlyemphasized. For instance, cases are studiedas ‘expressions’ of a regime of knowledge(see Table 2.2, p. 000), just as the work of anartist synthesizes elements of the contempo-rary existential situation of mankind or somesocial group. The idea of individually rationalaction – whether it is defended or criticized –is a consequence of this regime, and thusnot the elementary particle of explanations.Again, there is a more materialist version ofthis argument: the real forces of technolog-ical and economic rationality in a modern(e.g. capitalist) society produce ideologiesthat influence the growth of social scienceknowledge on this society. These forces mustbe the basis of any explanation. Internaldebates revolve around the link between ‘realstructures’ and ‘regimes of knowledge’.

Sociology of knowledge

In the social-philosophical view, the fun-damental processes, modernity above all,supply a sociology of knowledge. Modernityis mostly interpreted as a knowledge regime.One can trace its impact in all social spheres,as well as in social science. In such a meta-perspective, the standard position expressesthe fascination with natural science’s instru-mental rationality in the modern world,specifically in Western academic culture.Alternatively, this rationality and the accom-panying ideas of ‘enlightenment’ are linked –via ideologies or cognitive structures – to theunderlying driving forces. In both versionsof the argument, the sociology of knowledgeis an external one: the preconditions impliedby social research communities are seen asexpressions of more encompassing regimes ofknowledge and/or ideologies.

The relationship between thesciences

There is an inclination to consider scienceas a disguise. The varying maturity of thedisciplines is not interesting. Rather, mostdisciplines play practical roles, they are partof a larger machinery of standardizationthat imposes discipline. Empirical research isoften seen as purely instrumental (as is sec-tional divisions, the many ‘partial sociologies’within sociology). Empiricist instrumentalismis overcome, either by existential accountsor by transcendental arguments about theconditions of social science. In the social-philosophical vision, social science showsits maturity and its superiority over naturalscience by being able to provide citizenswith comprehensive understandings of theirpresent predicament.

We can distinguish two social-philosophical notions of theory (Mjøset2006b, pp. 347–349), with correspondingstrategies of specification and generalization.Transcendental or reconstructionist theory isabout the transcendental conditions of socialscience: basic notions of action, interaction,knowledge and structure (Habermas 1981,Giddens 1985). This notion is general atthe outset, as it concerns pre-empirical gen-eral conditions. With such a starting point,considerations about modernity must beconsidered a specification (see Table 2.2,p. 000). The deconstructionist notion oftheory is its opposite, aiming to show that notranscendental conditions can be established(Foucault 1969, 1975, Seidman 1991). Thisimplies a far-reaching sociology of knowledgeassessment, denying any accumulation ofknowledge whatsoever in social science. Thatapproach has a periodization of modernity asits most general (see Table 2.2, p. 000) feature(as it doubts any transcendental foundations)and it suggests an exceptionalist strategy ofspecification close to the one we find withinthe discipline of history.

Most historians tend to claim exceptionalstatus for their single cases. Their disci-plinary socialization does not require themto conduct the explicit comparisons along

balamurugan.k

ED: Please provide page numbers.

balamurugan.k

000),

balamurugan.k

000).

balamurugan.k

000)



several properties that might tell them inwhich respects their case is specific. Thepredominant view that there are no ‘casesof the same’, leads historians to avoid thecomparisons that would have allowed themto see their cases in different lights. Theirperspective changes only when the ‘spiritof the time’ shifts. At that point, revisionistinterpretations crop up, only to be challengedat a later point by post-revisionists. Eachperiod’s historians, so goes the old saying,writes the national history anew.

It is true that many of the – at least WesternEuropean and American – communities ofhistorians have absorbed a lot of social scienceinspiration since the student revolt of the late1960s. However, since the 1980s, many ofthese communities have been inspired by themethodology of denationalized humanities,developed by the French post-structuralists.Thus, many contemporary historians con-verge with the deconstructionist branch ofsocial-philosophy. Both groups of scholars aresocialized into a style of research typical ofthe humanities: their focus on archives andwritten sources leads to a non-comparativefocus on the single case. Applied to history,the deconstructionist position would claimthat the sequence of revisionisms and post-revisionisms shows that there can be noresearch frontiers.

THE CONTEXTUALIST PRACTICALPHILOSOPHY OF SOCIAL SCIENCE

The contextualist framework is a thirdmethodological framework in social science.5

As outlined in Table 2.1, it differs from the twoothers in terms of philosophical referencesand through its reference to styles of socialscience craftwork that lacks methodologicalcommunity with either natural science or thehumanities.6 In the following, we define thisattitude with reference to the same five aspectsas the two attitudes discussed above.

The contextualist approach is closely con-nected to the conduct of case studies. Beforewe turn to the five properties of practicalphilosophies of social science, let us consider

some common sense understandings of a‘case’. If you are involved in a case incourt, you are attentive to the specificitiesof the singular case, rather than any generalfeatures. You and your lawyer are interestedin how your case fits a taxonomy of legalregulations. You are interested in how aspecific institution (court) within a legalsystem will classify your case in the lightof the specificities emerging from documentsand testimonies presented in court. Certaintypes of court case require judges/juries todecide on personal responsibility or con-textual determination (cf. the discussion inBarnes 2001). Ahead of some court caseslies police work on cases. Turning to otherexamples, the daily activities of therapists andsocial workers are work on cases. Whereascases in court refer to conflicting parties,cases here refer to single persons, personswhose life history has led to problemsthat may be eased through a participatoryrelation.

The common feature is that we isolatesequences of events towards an outcome asa case because we have an interest in theoutcome and thus also in the process. Ineveryday life, the importance of the casemight be quite personal, we make cases outof particularly important chains of eventsin individual life histories. When socialactors become involved with cases (e.g. incourt cases, police investigations, therapy),they need to be sensitive to the processleading to the outcome, either because exactunderstanding of why the outcome cameabout is needed, or because the interest is ininfluencing the outcome.

Master example ofexplanatory logic

The contextualist framework is based onsensitivity to cases. (This is implied in theterm ‘qualitative research’). In the highlysimplified representation of Figure 2.1, a caseis an outcome preceded by a process thatunfolds in time. Delimiting cases in socialscience investigations, we imply a three-foldlogic of empirical research. First, we relate to



XP O

C

Past | PresentCase reconstruction

(CR)

CRPI

Past | Present | FutureProcess intervention

(PI)

Figure 2.1 Two varieties of case study. C, context; O, outcome; P, process.

our research problems, selecting the processand/or outcome to be studied. Second, wedefine the context, the elements that we treatas the environment of the process singled out.Third, we trace the specific links in the processwe have selected. Depending on the quality ofour knowledge about the case, we arrive at anexplanation of the case.

Sensitivity is necessary because investiga-tors or therapists interact with their own kind.Standard preconceptions tempts scholars toapproach experimental logic as closely aspossible. The case study, in contrast, can bedepicted as the opposite of the experiment. Inan experiment, the production of a predictedoutcome is just a means to arrive at generalstatements on the process. In case studies,outcome and process are significant in andof themselves. A case cannot be replicated atany time by any researcher anywhere. Somecases were the case in a particular contextand will not necessarily ever happen thesame way again. Other cases are producedagain and again by an ongoing process, butwe are then either interested in its specificcultural significance, or eager to evaluate itand possibly change it.

In an experiment, ‘similar context’ impliessimilar experimental set-ups. The fact thatthe processes studied are driven by humanslargely precludes the exact engineering ofexperimental controls and shielding in socialscience. We cannot build context, so wecannot produce the isolated workings ofparticular mechanisms. Instead, we either

have to reconstruct both context and processex post, or we intervene in an ongoingprocess in an existing context. Figure 2.1represents these two varieties of case study;both are related to contemporary problems.In case reconstruction, the research problemdetermines the selection of an outcome. Inprocess intervention, the research problemleads the observer/analyst into (actual orpotential) participation in the production ofan outcome.

In case reconstructions, the researcherreconstructs the process towards an outcomethat has occurred in the past – once or severaltimes. In social science, we often make casesout of historically individual and highly sig-nificant events in the development of a smalleror larger community. In an indirect sense, casereconstructions are also participatory.7 Forinstance, case reconstructions of significantmacro-events can contribute to the self-understanding of a community (9/11 in theUS). There are many examples of this, e.g.in the literature on the politics of ethnicidentity. This is particularly the case if theoutcome is not just significant but also remainscontroversial.

In process interventions, the researchertakes part in a process and the outcome is inthe future. This might be a once-off outcomeor, just as likely, a repeated outcome that hasoccurred many times before. Observation hererequires participation (interview, fieldwork,participant observation). Participation differsfrom experimental manipulation. In many



cases, the observer wants to play a passiverole, as the interest is in tracing the process asit would be without intervention. Anthropo-logical and ethnographic case studies mightnot be interested in particular outcomes,but sensitivity relates to the historical andcultural specificity of daily life routinesand cultural understandings. Participation,however, inevitably makes the researcher partof the creation of the case. This generatesdelicate problems of method. Still, theseproblems differ from problems of processintervention in cases that are inherently verycontroversial. These are more closely relatedto the ethics of research, and they mighteven lead the researcher to become part ofmobilization to change the outcome, giventhe ethical judgement voiced by a socialmovement. Below, we define this as criticaltheory.

We can consider these two varieties withoutsensitivity to case particularities. Processintervention would then be an experiment,the production of an outcome predictedby a theory as general as possible. Casereconstruction would be the selection of acritical case that can test a high-level theory,even without manipulation by the researcher.Both varieties are well known from the naturalsciences.

The interest in the specificity of casesgoes together with a focus on learning andintervention. The philosophical backgroundhere is in pragmatist (mainly US) andstandpoint (originally European) philosophies(Horkheimer 1937). Therefore, unlike thediscipline of history, the contextualist positionis committed to the explanation of single casesby means of comparison with other cases.The dual purpose is better specification of theoriginal case and development of contextualgeneralizations. But these generalizationsemerge through the analysis of specificities.They are important for learning. Learningestablishes an important link between processintervention and case reconstruction: in manysituations; case reconstructions explainingpast outcomes by means of comparisons mightbe important for the intervention in present-day problems.

Popularization of fundamentalmetaphysical questions

Pragmatist philosophy is distinguished by itscritique of the spectator theory of knowl-edge. Standpoint philosophies, such as earlyEuropean critical theory (developing Marx’s‘scientific socialism’) and postwar feministphilosophy and social science (Smith 1999)particularly emphasize the epistemologicalconsequences of social movements claimingequal rights in the development of society.In Hacking’s (1983) terms, both treat scienceas intervention rather than as representation.Judgements about the real and the constructedare made with reference to local settings.The pragmatist tradition has been invokedby both realists (pointing to Peircian processrealism) and by constructionists (there is evenan old empirical tradition studying socialproblems construction, closely connected tothe broader interactionist tradition that startedwith Chicago interwar sociology).

What unifies both realist and construction-ist interpretations is the view that accumula-tion of knowledge is linked to participation,intervention and learning. Pragmatism differsfrom the far-reaching empiricism of bothHume and the early twentieth-century pos-itivists who counterposed experience-basedscience to religion and philosophy. Thepragmatists instead tried to bridge this gap,aiming both to redefine the area of scienceand to bring philosophy and religion intoline with modern science. Acceptance ofthe Darwinian revolution was crucial totheir efforts to reconcile scientific reason-ing and religious belief (Skagestad 1978,pp. 21, 34f). Darwin transcended the Cartesiansubject/object dualism because he saw manas an organism that is part of the worldit develops knowledge about. Referring tothis interaction between the organism and theenvironment, Dewey (1920/1950, p. 83) wrotethat knowledge:

… is not something separate and self-sufficing, butis involved in the process by which life is sustainedand evolved. The senses lose their place as gatewaysof knowing to take their rightful place as stimuli toaction.



The contextualist position implies scepticismtowards generalizing statements on funda-mental features, whether they are aboutcultural deep structures or material ‘drivingforces’. Rather, it holds that statements onsuch properties must be linked by means of thesociology of knowledge to some participatory,interventionist and/or learning purpose. Likesocial philosophy, the contextualist position iscommitted to the analysis of a totality, but thetotality of a case, thus not sharing the socialphilosophical focus on the totality of drivingforces or cognitive deep structures.

Explanatory priorities givenautonomy of the social sciencerealm

As for reduction, the emphasis is on emer-gent properties and the practical contextof scientific knowledge. In contrast to thelargely theory-driven standard programme ofexplanatory reduction (cf. Hedström 2005,pp. 26–28, 36), contextualist research isproblem driven. Explanations are related tothe context relevant to the research questionat hand.8 There is no programme of reduction,not even within social science. The ideaof representing inherent structures betweensocial science elementary particles is absent.There may be many layers between the verymicro- and the very macro-levels, but thesedistinctions are drawn with reference to theresearch question, not to principal declara-tions about elementary particles. Explanationmight require analysis at lower levels thanthe outcome, but there is no general micro-to-macro problem. With contextualization asan independent element in the explanatorystrategy, the problem of micro-reduction doesnot emerge. Research may relate to variouslocations on a micro–macro continuum.A case may be singled out at any levelof aggregation: therapists deal with singleclients as cases, whereas macro-historiansdeal with nation states, macro-regions oreven historical epochs. Statements aboutinherent structures are always contextualizedand depend on the research question being

asked. Scholars committed to the standardframework sometimes discuss – and disagreeon – what outcomes social scientists shouldbe concerned to explain. Goldthorpe (2000,p. 203) insists on regularities, Hedström(2005, p. 67) on macro-level phenomena,and Elster (2007, p. 13) on events. In thecontextualist framework, such disagreementsseem odd. The problem at hand definesthe outcome(s) to be explained, whetherevents or regularities. Explanatory venturescan be plotted into a space, depending onwhere they are located on the micro-/macro-continuum, whether the researcher prefers apassive or active position, and whether theprocess/outcome studied is controversial ornon-controversial.9

Sociology of knowledge

The cluster of methods involving variouskinds of participantion define the contextualistframework ‘from below’ (see Table 2.1).Gaining knowledge by participation impliesthat, in principle (although not always inpractice), we interact with what we study.Sensitivity to cases implies that we acknowl-edge the knowledge of actors who are ‘in’ thecases. Standard researchers tend to judge thisjust as a source of bias. But given seeminglyinsurmountable problems of getting sounddata on beliefs (Elster 2007, p. 465), theethnographic pride in ‘being there’ has itsmerits. Specific to research into society is thatwe can enter into the very sphere where the‘mechanisms’ (even ‘elementary particles’)are supposed to be. The basic fact that westudy something that we are (or could be)ourselves, implies that there must be somerelationship between the way that we gatherknowledge, and the ways in which peoplelearn. Researchers may pursue knowledgemore systematically, but not in a qualitativelydifferent way.

According to the standard spectator theoryof knowledge, the scientific community rep-resents ‘nature’. In contrast, the contextualistview considers the scientific community as asociety (the internal perspective), embeddedin society at large (the external perspective).



As internal and external factors influencethe way researchers represent their researchtopics, not only the context of justification butalso the context of discovery (from biases infunding institutions, to the cultural problemsaddressed in the public sphere) is importantto understand why we get the knowledgewe get. This combined focus separates thecontextualist sociology of knowledge fromthe predominantly external perspective ofsocial philosophy. Research collectives witha high degree of (relative) autonomy canpursue strong programmes of basic research.Research collectives have their own internalprocesses, even fads and fashions. Still, eventhese collectives relate to current problems,often as defined by the agenda of fundinginstitutions.

Assumptions about the relationshipbetween the sciences

Although the social sciences are juniorpartners compared to the natural sciences, tocontextualists they are not immature. Rather,the view is that both are related to pockets ofrelatively robust, problem-related knowledge(local research frontiers). Such knowledgedoes not converge in one high researchfrontier. There is no methods community withthe natural sciences. Doing case studies, oneneed not feel bothered by ‘immaturity’vis à visthe natural sciences, there is no commitment tosome modified experimental master example.Social science is full of well-crafted andsuccessful case studies, many of which alsoserve as a basis for learning.

The contextualist notions of theory are twoways in which knowledge is drawn from andrelated to our ability to be sensitive to cases.

Explanation-based theory is knowledgeof contextual regularities accumulated fromexplanations of singular cases. These explana-tions are sharpened by means of comparisonsand further cases are sampled with referenceto the theory so far developed. Severalprogrammes in social science (e.g. in networktheory, Granovetter 1985) may be counted asexplanation-based theory, but in this chapter,we limit our discussion to the program of

grounded theory (Glaser and Strauss 1967) asa specification of such a notion.

We define critical theory as a special caseof explanation-based theory (Mjøset 2006a,p. 761). This definition is narrower thanthe broad definition implied by many socialphilosophers. The impulse towards criticaltheory emerges when social researchersparticipate closely with social groups thathave legitimate claims to social change.Some of these groups are organized as socialmovements, others might be marginalizedgroups with few capacities to organize. In suchcases, the position of the social researcher ina relatively autonomous research communitymay become problematic. In ethical terms(as argued, for example, in Habermas 1981)it might be untenable just to report researchon contextual regularities; moral imperativeslead the researcher to become part of therelevant social movement to change theseregularities.

The researcher then joins a movementwhose collective action might fail or succeed.In the latter case, it feeds back on thesingle case. The researcher does not simplyreconstruct an outcome but takes part inbroader efforts to change an outcome. This isprocess intervention, but outside the confinesof the research community. The kind of out-come differs depending on the kind of socialmovement. The most prominent examplesare movements that have even given nameto social science theories, such as ‘scientificsocialism’ (Marxism) and feminism. The kindor permanent structures they point to andwant to change differ: capitalist oppression ofworkers in classical Marxism, the position ofwomen at work and at home in the feministmovement. Critical theory need not be linkedto the macro-level: there are concepts of actionresearch relating especially to firms and inpsychology there is therapy relating to groupsof persons. Not all such process interventions,however, are critical theory. The line ofdivision between critical and explanation-based theory is a question of both ethicaljudgement and the social position of theresearcher. Our discussion below, however, ismainly focused on explanation-based theory.



Idealizing

Law-oriented

High(wide-range)theory

Middle-rangetheory Explanation-

based

The three-fold logic of casestudies:

1. Outcome/process-selection2. Contextualization3. Process-tracing

Reconstruc-tionist

Modifiedexperimentallogic

Deconstruc-tionist

Critical theory

The logic ofunderstanding

knowledge-regimes/material structures

Singlecase

Weak

Strong

Weak

Acc

umul

atio

n of

kno

wle

dge

Social-philosophicalContextualistStandard

Ran

ge o

f the

ory

Practical philosophy of social science

Figure 2.2 Practical philosophies of social science and notions of theory.

Figure 2.2 is the map we use for orientationin contemporary social science. It locates thethree methodologies and the six notions oftheory at different levels, and judges theirpotential for accumulation of knowledge. Inthe following, we use this map to discuss anumber of principal questions of relevance tothe conduct of case studies.

CASE STUDIES AND GENERALIZATION

How can the study of cases contribute togeneral knowledge? The question is a classicalone in debates on qualitative methodology. Inthe following, we deal with it in the light ofthe map in Figure 2.2. Generalization takeson different meanings depending on whatnotion of theory we refer to; not all of thesemeanings can be equated with ‘universal’.Table 2.2 relates one strategy to each of thesix notions of theory. The social philosophicalstrategies of generalization were definedabove, and are summarized inTable 2.2.As thediscussions on case studies and generalizationmainly concern standard and contextualist

methodologies, we do not relate to the social-philosophical framework in this section.

As for the standard notions of theory,the probability-based, statistical version ofthe law-oriented notion implies a segmentingstrategy of generalization: the aim is to extractgeneral relations within a specified field ofresearch particularly relying on large datasets.These datasets are special purpose ones:the Organization for Economic Cooperationand Development (OECD) makes datasetson growth available to the econometrician,research on innovation has its patent dataand innovation surveys, sociology has itsdatasets on social mobility, political sciencehas its electoral surveys, and so on. Thestrategy of generalization is to establishgeneral knowledge related to the socialsegment from which the data on a largenumber of cases are gathered: the theoryof economic growth, theories of innovativeupgrading of mature economies, the theoryof social mobility in industrial societies,the theory of voting behaviour and so on.The large number of cases contained in thedatasets allows the use of statistical methods



Table 2.2 Notions of theory and strategies of generalization and specification

Notion oftheory

Strategy of generalization Strategy of specification

Law-oriented Segmenting Cases are analysed as the locus of selecteddependent variables, which are explained bythe net-effects of selected independentvariables

Idealizing Insulating Cases as illustrationsExplanation-

basedFormal grounded theory (process-tracing,

mechanisms). Substantive grounded theory(contextualization by means of typologiesand periodization)

In combination, substantive and formalgrounded theory secure sensitivity to specificcases (this is most properly described as ajoint strategy of specification andgeneralization)

Critical theory Efforts to promote their world view,challenging the ‘model monopoly’ of present‘regimes of knowledge

Contributing to social change

Reconstructionist(transcen-dental)

Pre-empirical fundamentals Periodization referring to modernity or somephase thereof. Cases as expressions of such‘logics of the present’

Deconstructionist Periodization referring to modernity or somephase thereof

Exceptionalist strategy reminiscent of thatfound in the discipline of history

of generalization. However, the security ofthis method comes at the cost of segmentation,and findings based on patterns of correlationare not easy to translate back into a worldthat seldom is structured so that it givesrise to natural experiments. Given that thepurpose of the law-oriented theories is theestablishment of ‘as general regularities aspossible’, there is little concern for cases, theyare only the raw materials of large datasets.To the extent, however, that Merton-typemiddle-range theories are derived from otherempirical sources than large datasets, anotherstrategy of generalization may be involved.We return to this later.

The idealizing notion of theory implies aninsulating strategy of generalization. Rationalchoice theory claims relevance for all socialsegments. It is based on a general theory ofinteraction, practised as thought experiments.Cases may here serve to illustrate patterns ofinteraction modelled in thought experiments.Strong interpretations of the idealizing notionwould see game theory, etc. as the source ofsuch patterns, whereas softer interpretationswould analyze patterns with reference toa number of empirical sources, includingfolk wisdom, well-crafted case studies, etc.

However, the attitude is still the standardone, and thus a strategy of specificationis not considered important: the focus ison the thought experiments as generalized,widely applicable knowledge, and often the‘parsimonious’ nature of such knowledge isemphasized. In extreme cases, the aestheticlegitimation sometimes found in mathematics(the ‘beauty of a proof’) is simply taken over.

The contextualist strategy of generalizationis to generalize only within specified contexts(Mjøset 2006a). In this view, specification andgeneralization are not opposites. Specifica-tion is only possible through more generalknowledge. For instance, the exact features ofNorway as a welfare state must be assessedwith comparative reference to other casesof the same (in grounded theory, this iscalled theoretical sampling).As specificationsare made in this way, the results alsofeed back into more general knowledge:denser and broader typologies, concepts andmodels of contextualized social interactionpatterns. The use of comparison is the mainalternative to subsumption under as general aspossible concepts and theories in the standardconception, and also an alternative to the useof cases as ‘expressions’ of broad periodizing



notions (such as modernity) in the social-philosophical view.

It is not a contradiction to talk aboutgeneral theory in the contextualist framework,but then it must be distinguished fromuniversal range theory. Generalizations thatretain grounding cannot be taken further thanthe middle range, but we can have more orless general theories within the middle range.Universal-range theory must necessarily beungrounded. Theories can be established at acase level but should not be ‘overgrounded’,as in the case of the exceptionalist strategy ofspecification (see above).

Whereas the interwar Chicago School ofSociology mainly practised its ethnographiccase studies, its second generation in the 1950slaunched a criticism of variables as concepts(see, in particular, Blumer, 1969). Blumer’scriticism favoured case sensitivity, but thesensitizing concepts he promoted as an alter-native to ‘definite concepts’ alienated quanti-tative research altogether. A later generationof critics reflected the spread of statisticalinference, regression analysis in particular.Ragin (1986, 2008 chapter 10) claimed thatthe estimation of net-effects across the wholepopulation leads to notions of causal analysisthat are at odds with the sense of causationwe get from tracing processes that leads tooutcomes in cases. In contrast to Blumer’ssocial psychological orientation, the latergeneration was more concerned with macro-studies, especially historical sociology. Theircriticism pointed in two directions: to concernwith the methodology of macro-comparativesocial science (Ragin 1986, Mjøset 2000,referring particularly to Rokkan’s work in the1970s, cf. Rokkan 1999), or more generallyto exploration of various kinds of processtracing and network models (Abbott 1999,2001). These contributions generally agreedthat Skocpol’s (1984) attempts at method-ological synthesis in historical sociology didnot sufficiently cut the ties with standardpreconceptions (Mjøset 2006b).

The other direction was towards alter-native quantitative methods. One exampleis Ragin’s (2000) qualitative comparativeanalysis (QCA), based on set-theory instead

of probability. Not relying on correlations, thistechnique can also analyse small populations.Abbott (2001) proposed models based on thelogic of genetic sequencing. More broadly,Shalev (2007) urges scholars to rely on lesshigh-tech statistical methods. The contextual-ist orientation is more than just a legitimationof qualitative methods, it has recently alsoled to development of types of quantitativestudies that are designed to increase sensitivityto cases even among those who work withlarge datasets.

In the contextualist approach, the chal-lenge of generalization is the investigationof smaller numbers of cases explained byconcepts with high internal validity. Thework on qualitative macro-studies, and alsoon non-probablistic quantitative approaches,indicates that substantive generalization andcomparative specification can go hand inhand. The next two sections present someconceptual and methodological specificationsof such a strategy of generalization.

SUBSTANTIVE AND FORMALGROUNDED THEORY

The distinction between substantive andformal grounded theory, as well as the empha-sis on the operation of comparing, makesGlaser and Strauss’s (1967) programme ofgrounded theory – rooted in interwar Chicagosociology – a particularly rich source forinvestigations of explanation-based notionsof theory. In this section, we show how thisprogramme provides further insight into therelation between types of theory and strategiesof generalization. We can start from whatmight seem an inductionist credo:

Both substantive and formal theories must begrounded in data. Substantive theory faithful tothe empirical situation cannot, we believe, beformulated merely by applying a few ideas from anestablished formal theory to the substantive area.To be sure one goes out and studies an area witha particular sociological perspective, and with afocus, a general question, or a problem in mind.But he can (and we believe should) also study anarea without any preconceived theory that dictates,



prior to the research, ‘relevancies’ in concepts andhypotheses. Indeed it is presumptuous to assumethat one begins to know the relevant categories andhypotheses until the ‘first days in the field’, at least,are over. A substantive theory generated from thedata must first be formulated, in order to see whichdiverse formal theories are, perhaps, applicablefor furthering additional substantive formulations.(Glaser and Strauss 1967, p. 33f)

This is a statement in favour of substantivetheory, defined as theory developed ‘for asubstantive, or empirical, area of sociologicalinquiry’ (e.g. race relations, delinquency,research organizations). The opposite is for-mal theory, which is ‘developed for a formal,or conceptual, area of sociological inquiry’(e.g. stigma, deviance, formal organization,social mobility). Glaser and Strauss (1967,p. 32f) emphasize that both types of theory are‘middle range’ in Merton’s sense. However,towards the end of this section, we shallspecify differences (see Figure 2.2) betweenmiddle-range theory and explanation-basedtheory.

Strauss (1970) emphasized that the state-ment ‘without preconceived theory’ did notexclude reliance on earlier substantive theory,directly related to the field studied. Let us callthis the principle of substantive primacy; ‘dis-covering substantive theory relevant to a givensubstantive area (…), allowing substantive

concepts and hypotheses to emerge first, ontheir own’ (Glaser and Strauss 1967, p. 34).

We hold that this is a basic princi-ple in any explanation-based theory. How-ever, Glaser and Strauss never seem todiscuss how it relates to the distinction(grounded/ungrounded) that appear in theirbook title. We make this connection inTable 2.3, considering grounded/ungroundedas an account of how theory is discoveredand substantive/formal as a classification oftheory that has been discovered (whetherit is explanatory or not; its contribution toaccumulation of knowledge).

Glaser and Strauss’s programmatic for-mula – theory as grounded in data – refersto qualitative data emerging as researchersexercise sensitivity to cases. Such groundedtheory is based on systematic work that codesdata to establish core concepts and samplesnew cases to extend and develop earlierfindings. Glaser and Strauss also use the term‘ungrounded theory’, indicating theories thatare not grounded in this way (see Table 2.3).Both notions can be differentiated in linewith the substantive/formal distinction. In thefollowing, we discuss some results of sucha differentiation, relating to our typology oftheories (see Figure 2.2).

We argue throughout this chapter againstthe frequent accusation that grounded theory

Table 2.3 Taxonomy of grounded and ungrounded theories

How knowledge was discovered

Grounded Ungrounded

Types ofaccumulatedknowledge

Formal Stylized interaction patterns recurring inexplanations across various substantiveresearch fields (internal analogies)

Explanatory patterns drawn not from socialscience but from other domains ofscience (external analogies, e.g.mechanical, organistic)

Transcendental notions of theory.Methods-driven empirical analyses (e.g. net

effects, axiomatization)Substantive Case studies: case reconstruction/process

intervention in various fields of society inspecified periods. Boundedgeneralization developed throughcomparison of cases within an area ofresearch, using typologies andperiodization to specify context

‘Journalistic’ generalizations.Exceptionalist specifications.Examples selected from various fields put

forward as illustrations of (formal)theoretical claims.

Empirical cases as expressions of trendsemphasized in broad interpretations ofthe present



is inductive or descriptive only. In Table 2.3,we can locate description as substantiveungrounded ‘theory’. A pure discription ofthe flow of events is impossible, and thus‘unconscious’ principles of selection must beinvolved in any description. Moral panics andother forms of ‘journalistic’ generalization ofsingular cases are good examples. This is nottheory in any proper sense. The exceptionaliststrategy of specification (see Table 2.2 and‘The social-philosophical practical philoso-phy of social science’, above), which studiessingle cases independently of any comparisonwith other cases of the same, is alsoan example of substantive but ungroundedknowledge.

The relation between the two divisions ofthe ungrounded column in Table 2.3 illustratesthe gap in the standard framework betweentheory and empirical research (see ‘The stan-dard practical philosophy of social science’,above). We can distinguish four attemptsto bridge this gap. Two of these bridgingattempts try to close the gap going ‘down-wards’ from formal ungrounded theory, thusproducing substantive ungrounded theory.The other two attempt to find formal theorywith some reference to empirical research, letus say they try to move ‘upwards’.

As for the relationship between hightheories and empirical substance, there is astandard and a social-philosophical version(see Figure 2.2). The standard position impliesthe idealizing notion of theory, thought exper-iments in which the researcher establishesthe context via definitions that suits anaxiomatic system. The social-philosophicalposition implies reconstructive theory, whichis formal in the philosophical sense of claim-ing transcendental status. The first downwardsbridging attempt implies the quoting ofillustrative examples to substantiate idealizingtheory. For instance, the works of JonElster (e.g. Elster 2007) are avalanches ofsuch examples, used as raw materials forspeculations around his increasingly softversion of rational choice theory. The seconddownwards bridging attempt implies recon-structive theory approaching the study ofmodernity by linking transcendental notions

either to selected examples claimed to beexpressions of the core concepts (modernity,globalization) in their interpretations of thepresent, or to selected quotes drawn fromolder or more recent classics who tried tograsp the spirit of their age (‘iron cage ofrationality’, ‘anomie’).

As for the upwards bridging attempts,one is what we earlier discussed as thesegmenting strategy of generalization, yield-ing explanations in terms of net effects.The contextualist criticism summarized above(in ‘Case studies and generalization’) wouldconsider this methods-driven research, andin that sense formal (cf. Abbott’s criticismof ‘general linear reality’, and Ragin’s (2008chapter 10) criticism of ‘net-effects thinking’,which yields ‘vague theory’).

The second upwards bridging attemptwould be the Mertonian concept of middle-range theory. As noted (in ‘Three practicalphilosophies of social science’), the revision-ist standard notion of mechanisms can be seenas a contemporary update of this programme.It has emerged from frustration with thetwo standard strategies of generalization(insulating and segmenting), as none of thesehave been able to close the theory/explanationgap. Their proposed bridging solution lies inexplanation by mechanisms; however, beforewe consider it more closely, we need to discussthe purely formal, ungrounded theories inTable 2.3.

The history of social science has manyexamples of formal theories ‘ungrounded’by ‘external analogies’. A theory-drivenprogramme such as Parsons’ structural–functionalism of the 1950s and 1960s, reliedon the analogy of an organic system to developwide-ranging formal typologies. Despite astandard point of departure, Parsons’ theoryof action actually ended up quite close tothe transcendental position of reconstructivesocial-philosophy (see Figure 2.2). This kindof ‘grand theory’was the main contrast againstwhich Glaser and Strauss developed groundedtheory in the 1960s. There was no broad con-textualist breakthrough, however, as Parsons’theory was challenged simultaneously byeconomics-inspired rational choice and by



game theory. Here, ‘external’ formalism wasnot from another field of science but frommathematical notions of theory as an axio-matic system, ‘interpreted’ for action theory,thus requiring particularly strongly idealizingassumptions (cf. the economics notion ofcaricature models’, Mjøset and Cappelen2009). The rational-choice programme, whichcan also be considered to be methods driven,attracted a lot of intellectual energy for severaldecades, but recent standard revisionismrejects this attempt to close the gap (Hedström2005, Elster 2007).

This critical stance points in the direction ofa contextualist perspective. Ignorance of theprinciple of substantive primacy, wrote Glaserand Strauss (1967, p. 34), is in most instancesthe result of:

… believing that formal theories can be applieddirectly to a substantive area, and will supply mostor all of the necessary concepts and hypotheses. Theconsequence is often a forcing of data, as well asa neglect of relevant concepts and hypotheses thatmay emerge.

The rational choice criticism of functionalismwas only about replacing one ungroundedformal theory with another. Both differfrom the contextualist, interactionist traditionfrom Simmel, through Goffman and intocontemporary interactionist thinking, whichwe can conceive as formal grounded theory.This tradition was always close to and partlyoverlapping with the empirical ethnographicorientation of the Chicago School, which wecan count as substantive grounded theory. Canwe relate the revisionist notion of mechanismsto these notions.

Let us turn to the grounded column ofTable 2.3. The distinction between formaland substantive grounded theory decouplesgenerality and explanatory power. Substan-tive grounded theory is the basis of thecontextual generalization described above.Formal grounded theory leads to formalgeneralization that recognizes similaritiesbetween patterns of social interaction in manyfields of study. Such generalization mustbe grounded. It respects the requirement ofinternal analogies (see Table 2.3), because

these must be derived from substantivestudies in several fields of social research.It is not indexed to specific contexts, thusit is formal. It is general, but it explainsnothing before it is inserted into a context. Interms of the three-fold contextualist logic ofexplanation (see ‘The contextualist practicalphilosophy of social science’), a formalgrounded theory is an appendix to the thirdstep, as it isolates formal patterns visible inseveral process-tracings in different lines ofstudy. These formal patterns can be usefulas ‘components’ of explanations: they areexplanatory ‘modules’, which results fromresearchers’ efforts to spell out in some detailthe formal properties of selected interactionpatterns.

Glaser and Strauss (1967, p. 34) note thatthe principle of substantive primacy:

… enables the analyst to ascertain which, if any,existing formal theory may help him generate hissubstantive theories. He can then be more faithfulto his data, rather than forcing it to fit a theory.He can be more objective and less theoreticallybiased.

As an example, they note that it wouldbe wrong to apply Parsonian or Mertoniancategories at the start, it is crucial to ‘wait tosee whether they are linked to the emergentsubstantive theory concerning the issue infocus’. Formal generalization is not simply analternative strategy of generalization. Contextcan be specified only by means of substantivegrounded theory. Formal grounded theorydoes not relieve the scholar of doing thecomparative craftwork: exploring categories,core categories, properties and subproperties,devising typologies and periodizations.

This line of argument allows us to concludeon the second upwards bridging attempt,the revisionist standard notion of mecha-nisms. This can be seen as an attempt tomake formal theory substantive, without theproblem-related choice of outcome/processand contextualization. This kind of formaltheory isolates one part of the three-fold logicof contextualist analysis, namely the process-tracing logic. The revisionist notion ofexplanation by mechanisms requires tracing



of the chain of causal and intentional links(Elster 1989). Whether this formal theory isgrounded or ungrounded, however, dependson the source of the mechanisms. We haveargued that if they rely on external analogiesthey are ungrounded, but if they rely oninternal analogies drawn from social sciencestudies, they are grounded. In the latter case,they rely on earlier substantive groundedtheory in the form of explanations of causalprocesses, as in Elster’s 1989 understanding.

Interestingly, Elster in 1998 suggested arevised definition. The old definition he nowcalls ‘causal chain’, while according to thenew definition, mechanisms ‘are frequentlyoccurring and easily recognizable causalpatterns that are triggered under generallyunknown conditions’(Elster 2007, pp. 32, 36).With his two successive definitions, it seems,Elster rediscovered Glaser and Strauss’sdistinction between substantive and formalgrounded theory! If we trace a causal chainwe need to include the context, and thus,we have substantive grounded theory, theearliest definition. If we recognize patterns ininteraction across ‘conditions’ (contexts), wehave formal grounded theory, the most recentdefinition, in which conditions are unknown,i.e. unspecified. Elster has, however, notnoted the parallel, because he recognizes noother methodological frameworks than thestandard one.

MECHANISMS ANDPROCESS-TRACING IN THECONTEXTUALIST FRAMEWORK

By considering the components of groundedtheory, we can define a notion of mechanisms

within the contextualist framework. Glaserand Strauss (1967, p. 36) define categoriesand their properties as elements of a theory.A category ‘stands by itself as a conceptualelement’ and a property is a ‘conceptualaspect’ of a category. The two notions arerelative, thus an overall category might havemany properties and each of these propertiescan be seen as categories that have furtherproperties. This can be simplified into a three-level terminology of categories – properties –subproperties.

Let us start from a simple example oftypology construction. Size, flavour, juicinessand type of product are properties of thecategory fruit. Adding a question, specifyingthe dimensions along which the propertiesvary, and aggregating the two vertical rows,we get a two-fold typology (Table 2.4). Thetypology is related to the core category thatemerges when empirical data are scrutinizedwith the research question in mind (cf. thevarious types of coding and sampling ingrounded theory; Glaser and Strauss 1967).

Within this framework, a mechanism canbe defined as the pattern of social interactioninvolved as the properties of a categoryproduce outcomes along a dimension. We canfurther define causal chain or causal processwith reference to conjunctions of mechanismsinvolved as social interaction on the dimen-sions of several (or ideally all) propertiescreate overall outcomes. Thus, process tracinglinks mechanisms pertaining to the variousproperties of the core category we see ourcases as cases of. The contextualist approachexplains by means of many mechanismslinked in causal processes.10 Furthermore,once we consider causal processes thatgenerate recurrent outcomes, we must also

Table 2.4 An example of typology construction

Question Category Properties Dimensions

Why differentprices?

Fruit, specificallyoranges

SizeFlavourJuicinessProduction

Large ↔ SmallSweet ↔ Less sweetHigh ↔ LowEcologica ↔ Traditional

Types of orange High-price ↔ Low-price

Tables 2.4 and 2.5 are synthesized from Senghaas (1985).



pay attention to the possibility that cumulativechange occurs as the processes recur. Thesechanges might originate in contradictorydynamics or in small contextual changes,and cumulative processes might lead to moresignificant changes in context, either throughslow change or more sudden critical junctures(turning points). In the changed context, newprocesses might evolve.

In our simple example in Table 2.4,the mechanisms stems from biology andagro-science: specific species of oranges,fertilization, etc. But we can consider amore complex example from comparativepolitical economy. Senghaas’s (1985) largeproject on what developing countries canlearn from European development experi-ences can be reconstructed as a discoveryof explanation-based theory (Mjøset 2007).With reference to this research question,Senghaas established the core category ofauto-centred development, which is under-stood as the combination of economic growthand improvement in living standards for broadmasses of the population. Relying on a largeselection of relevant literature on economic

development – theoretical, quantitative andmonographic – Senghaas ended up with atable that specifies properties and subprop-erties of this core category. Table 2.5 is asimplified version.

Given the research problem, it is obviouslynecessary to close the explanatory chain(to establish context) way ‘above’what wouldbe Hedström’s (2005) fundamental level(beliefs, desires, opportunities). The numberof subproperties is, however, true to themiddle-range realist view about stratified real-ity (Pawson 2000). Each subproperty can bedimensionalized, some in quantitative terms,others just in qualitative terms. This examplealso allows us to see how a change fromdichotomization to more nuanced typologieswill increase the detail of the processes traced.The more distinctions we allow into thedimensionalization, that is, in Ragin’s (2000)terminology, we turn from crisp to fuzzy sets,the more historical context is allowed for. Thesame is the case if we increase the number ofproperties and sub-properties involved.

Systematic comparison led Senghaasto emphasize empirical indicators of the

Table 2.5 Properties, subproperties and dimensions of the core categoryauto-centred development

Properties Subproperties Dimensions

Agrarian property/social structure

DistributionInnovation-orientationCooperative movement

Egalitarian ↔ SkewedHigh ↔ LowStrong ↔ Weak

Distributional patterns IncomeIncome distribution/savingsWages and salaries’ share in net national product

Egalitarian ↔ SkewedPromote ↔ block innovationHigh ↔ Low

Economic institutions Firms (risk/innovation-orientation)Supportive banking systemNature of national system of innovationEducation – training (literacy)

Strong ↔ LowYes ↔ NoStrong ↔WeakHigh ↔ Low

Social mobilization Mobilization of farmersMobilization of workers (unions)

Strong ↔ WeakStrong ↔ Weak

Political mobilization Democratization (replacement of old elites)Nation building – sovereigntyClientilism in political parties

Strong ↔ WeakEarly ↔ LateLow ↔ High

State Administrative reformState provision of infrastructure

Yes ↔ NoConsiderable ↔ Low

Source: Mjøset (2007).Tables 2.4 and 2.5 are synthesized from Senghaas (1985).



egalitarian distribution of farmland andagrarian incomes, postulating a mechanismwhereby cash income in the agrarian sectoraccrued to a large share of the families inthe that sector. The context is one in whichthe majority of the population is in theagrarian sector. With egalitarian agrariandistribution, staple export incomes are spreadbroadly. This creates broad-based domesticdemand that stimulates small, but innovation-oriented manufacturing activities, the simpleproducts of which are bought mainly byfarmers.

Explanations of the development experi-ence of particular cases, however, requires usto look at causal chains connecting severalsuch mechanisms (or explanatory factors).The farmers joined in cooperatives thatassisted both with sales and with investmentsin agricultural equipment. There are manymechanisms here, and one would never reacha conclusion if one were to trace them allthe way ‘up’ from the level of beliefs, desiresand opportunities. Instead, they are cut off bymeans of contextualization. The institutionsor social structures that form the context arethe results of processes not further traced:agrarian mobilization, church/state relations,colonial history, etc.

The factors that have been deemed impor-tant by earlier substantive research are listedin Table 2.5. For each of the subproperties,the dimensionalization creates a scale or aspecified typology. These set the contextfor the mechanisms. But taken as such –at any level – mechanisms are patterns ofsocial interaction, routine behaviour and soon. Mechanisms are formal grounded theoriesthat cannot explain without context. Theinteresting feature is how various mechanismsconnect in specific contexts defined bytypologies related to properties/subproperties.This specifies how reality is layered (Pawson2000): the various cases of more or lesssuccessful overall outcomes in terms of auto-centred development, are produced and repro-duced thanks to conjunctures (Ragin 1986) ofsuch mechanisms in specific contexts.

The analysis is often in terms of ‘stylizedfacts’, which in such an analysis can be

specified as typical processes. For instance,Senghaas refers to Hirschman’s (1977) notionof linkage effects, which can be specified asvarious stylized constellations of early manu-facturing development and social structure.AsGlaser and Strauss (1967, p. 32) emphasize,substantive and formal theory differ only ‘interms of degree’. The theory of linkage effectscan be seen as substantive grounded theory,based on explanatory, monographic casestudies of economic development. However, itcan also be developed towards a more formalgrounded theory in the form of network mod-els. However, if the networks are modelledstarting from checkerboard structures, as inHedström (2005), we are in the realm ofungrounded formal theory and it is an openquestion as to whether a homogenous actionrule derived from a regression equation withcertain controls will actually ever lead us tosuch stylized processes.

To the extent that processes are cumulative,there will be change. This illustrates howtypologies are not just specific to researchquestions, but also to historical periods. Theymust also be revised with reference to majorcontextual changes, conceived as turningpoints or more gradual change. However,the formal grounded theory of linkages canbe used to trace the impact of manufactur-ing sector transformations through severalperiods, whereas the contextual specificationsvaries.

An important implication of this contextu-alist principle of substantive priority is thatwe cannot have accumulation of knowledgeat the high level. In terms of accumulationof knowledge (cf. Figure 2.2), high-levelformal theory in and of itself is as weak asungrounded descriptions of the flow of events.Another implication, of equal relevance tothe philosophy of the social sciences, is thatwe cannot have competing formal theories,only competing explanations. If explanationsare to compete, substantive theories mustbe involved. Competition between theories –in other words – requires agreement onproblems (outcome or process selection) andon context; it requires all three elements of thecontextualist logic of research.



LOCAL RESEARCH FRONTIERS

The claim that contextualist approachesviolate the principle that any observation istheory loaded is partly inspired by Popper’sphilosophy of natural science. As with anybroad principle, scrutiny by new generationsof professional philosophers (Hacking 1983,p. 171ff) leaves it in dire need of specification.Note that with reference to Figure 2.2, we faceat least six different interpretations of what‘theory-loaded’ might mean.

In the contextualist framework ‘theory-loaded’ means that discovery of new theoryrelates to earlier substantive grounded the-ory (Strauss 1970). We shall specify thisrelationship by means of the notion oflocal research frontiers (Mjøset 2006a). Itspecifies how explanation-based theory leadsto accumulation of knowledge just because itremains in the middle range. Social scienceknowledge is developed with reference to thevarious areas of society. The main argumentof the pragmatist philosophers was alwaysthat accumulation of knowledge takes placebecause it is important for the community.If many researchers ask the same researchquestions with reference to similar sets of dataand other empirical investigations, we get alocal research frontier. Such frontiers developwith reference to problems that are crucialto the community.11 Rather than believingthat we ‘observe’ in the light of some hightheory, we must realize that a problem arearequires the definition of a small numberof core categories. These categories havemany properties, and for each there might besubfrontiers of research, and thus, differenttypes of explanation-based theories.

Local research frontiers synthesize existinganalyses of relevant core categories and theirproperties. It also includes stylized facts,either quantitative (e.g. based on agreementon major indicators) or qualitative (e.g. asin certain commonly used typologies andperiodizations). All these components arerelated to what the community of researchersaccept as good explanations of relevant cases.The principle of substantive primacy applies,substantive grounded theory is the basis of

local research frontiers. Substantive theorywithout a local research frontier becomesungrounded substantive theory, unless thetopic/field is entirely unexplored, which isseldom the case. Only when substantive workis done can one check whether existing formal(grounded) theory might be useful in theconsolidation of the research.

We have shown (see ‘Case studies andgeneralization’) that contextualist researchershave ways of accumulating knowledgethat transcends the engrained generalization/specification dichotomy. Even a single caseanalysis can contribute to growth of knowl-edge when it is developed with reference toknowledge already accumulated in one ormore local research frontiers (Mjøset 2006a).Such analyses can also rely on monographs.Their empirical material might be ‘over-grounded’ due to the influence of ideas ofexceptionalist specification (see ‘The social-philosophical practical philosophy of socialscience’), but they can be regeneralized withdue attention to context. The specificationof one or more new cases feeds back intothe local research frontiers, adding to thegenerality of knowledge, even though its tiesto the context are not cut.

In the standard framework, one imagines‘basic’ theory – solving general problems in ahigh level research frontier – being ‘applied’to local problems. The social-philosophicalemphasis is on cultural problems that havean existential kind of generality. In the con-textualist perspective, one sees all problemsas specific and local, emphasizing how socialscience theory is cumulative only in localresearch frontiers. Local research frontiersshould not be confused with exceptionalistspecifications, which would narrow down aresearch frontier to what we know from earlierresearch on a particular case. Local researchfrontiers synthesize research on many cases.

Here is a contextualist notion of growth ofknowledge: with richer dimensionalization ofproperties or subproperties, research frontiersbecome increasingly mature. This shouldnot be conceived as ever more ‘correct’representations of basic features of reality,but rather as a growing consensus within



a broad social science research collectiveconcerning accumulated knowledge on socialstructures and processes in local, problem-related research frontiers. Even if somedegree of maturity has been reached, itmight not last forever, as underlying problemscan change. The knowledge might growfurther, be transformed in the light of newproblems or wither away. Social sciencetoday possesses knowledge in several suchlocal research frontiers, but this knowledge isnot converging into higher level knowledge(see ‘Mechanisms and process-tracing in thecontextualist framework’).

The accumulation of knowledge in localresearch frontiers has its own sociology ofknowledge. If the study relates to uncon-troversial cases, few researchers will returnto it, the public will not address it andresearchers will not judge this knowledge.However, real and persistent problems ofsocial development lead to so much attentionthat local research frontiers develop andpersist. A well-consolidated local researchfrontier requires funding enough to sustaina research collective over significant periodsof time. Certain clusters of problems arebetter suited than others as a basis fordurable collective and even interdisciplinarysocial science. In some cases, disciplinaryidiosyncracies might lead different disciplinesto study the same problems in relativeindependence of each other. In other cases,the nature of the problems is such that eveneconomists can work fruitfully with non-economist social scientists.

Research into the welfare state is a goodexample. In the Western world, many interestsconverge to sustain a local research frontier onthis topic. The collective of social researchersnow has at its disposal a literature addressingthe same cluster of questions by means ofcarefully maintained and updated databases,frequently used typologies, stylized facts,comparative case studies, models of explana-tion and converging discussions on historicalbackgrounds. Whatever a researcher mighthold in terms of high theory, he or she willhave to rely on this complex of middle-level knowledge, which is based on the best

explanations so far provided. This judgementis passed in the local research frontier,within which researchers judge knowledgeby drawing on it in further research. Thisknowledge is not insulated from broaderdiscussions in the public sphere on matters ofpolicy and strategy.

Local research frontiers can cluster andmight relate to each other in hierarchies.For instance, research into socioeconomicdevelopment patterns in Nordic countries(which might be relevant for policy learningby poor countries), the welfare state andthe position of women in Nordic societyare different research frontiers in terms ofoutcomes analysed. All can be related topolicy learning in several respects. Theyrequire us to trace different processes butthey might draw on overlapping contextualknowledge. We can imagine typological mapsin a hierarchy, where the highest ones containgeneral substantive knowledge that might berelevant for several local research frontiers.This is the notion of substantive generaliza-tion. Such general knowledge is still specificto certain areas, even geographically. Letus consider this further with reference toRokkan’s political sociology.

Rokkan’s contribution to political sociol-ogy was based on his contributions to subfron-tiers such as nation-building, state-formation,electoral behaviour and the structure of partysystems. In his last contributions (in the1970s, collected in Rokkan 1999), however,he developed a framework that integratedseveral such research frontiers. It remainssubstantive, though, because it is valid onlyfor Western Europe. The main outcome tobe explained is the structure of the WesternEuropean party systems in the postwar period(1950s and 1960s).

Rokkan developed a basic sequential modelas well as a multitude of typological maps.As grounded theory (Mjøset 2000), Rokkan’stheory draws on historical monographs, stud-ies comparing political institutions and on hiscontinuous work on electoral statistics. Hisstudy of Western Europe yields no generalformal theory to be applied directly (e.g.to Asia). But by doing a similar craftwork



of contextualization, relying on many of thesame properties, one could establish newsubstantive theory in the form of typologicalmaps of, say, the Asian region. This mightyield new formal theory, but some of theformal theories developed from WesternEuropean developments would surely beuseful, provided due attention is given to thedifferent contexts. We see the principle ofsubstantive primacy at work.

Given the clustering of research frontiers atdifferent levels, the three elements in the case-study logic often interact. Within each periodgiven in his sequential model, Rokkan traceseconomy – territory – culture ‘variables’.These broad contextualizations allow choiceof a number of more specified outcomes tobe explained, e.g. breakdowns of democraticregimes, the structuring of party systems,patterns of political mobilization. Specifiedtypological maps then provide more specifiedcontexts. The resulting explanations lead tosuccessive refinements, both to the specifiedtypological maps and the more basic sequencemodels.

Rokkan’s work also illustrates the com-bination of formal and substantive theory.Although his main strength was contextual-ization, process tracing was also involved.However, his formal theories were grounded.They were interaction patterns generalizedacross research on political and social history:patterns of mobilization, alliance formation,revolts in situations of scarcity, organizationbuilding, social movement formation. He alsodrew on the formal theories of others, e.g.Hirschman’s ‘exit, voice, loyalty’ triad andeven Parsons’s AGIL-scheme, trying to makethem serve him as formal grounded theory.He dealt with these formal theories only tothe extent he could put them to work togetherwith his substantive arsenal of typologies,periodizations and field-specific processesand mechanisms.

A comparison of (perhaps) the two mostquoted Norwegians in international socialscience gives a striking result: Rokkan followsthe principle of substantive primacy andhardly ever uses empirical material as mereexamples; Elster, by contrast, pursues formal

theory only, and examples is all he has in termsof empirical content.As their implied practicalphilosophies of social science diverge in thisway, it is not surprising that Elster hardly everfinds it interesting to refer to Rokkan’s work –so far.

Rokkan provided social science with accu-mulated knowledge, not just in the formof mechanisms, not as insulated relationsbetween variables, but in the form of contex-tualizing maps that might be useful to severalsubfrontiers of research on European polit-ical developments. These were substantivegeneralizations: they allow later researchersto contextualize also with reference toother significant outcomes, and they can beimproved and extended. His work has notrace of idealizing models and no interest inconnecting his arguments back to elementaryparticles (beliefs, desires and opportunities).Compared with social-philosophy, Rokkan’swork is too disaggregated: it is ‘below’ thelevel of modernity.

Besides Weber’s wideranging typologicalwork in Economy and Society, Rokkan’smodel and maps are some of the most workedout examples we have in the social scienceof substantive generalization. Within thecontextualist framework, we can understandWeber’s various ‘sociologies’ as typologicaldiscussions of the properties (law, domina-tion, religion, economy) of ‘Western devel-opment’. These ‘sociologies’ are subfrontiersrelated to a broad local research frontier.The overall explanation, as among otherssuggested by Collins (1986), ties the variousproperties together in a complex cumulativeprocess, one that is singular and relevantto many. The explanation traces processesthat tie the various properties together.This interpretation challenges both standardand social-philosophical interpretations ofWeber.12 Social philosophers are very fondof Max Weber’s few paragraphs on the ‘ironcage’ nature of modernity. But contextualistsare more interested in the main contentsof Weber’s work, namely his enormousweb of typologies contained in each of hissociologies. These various typologies, hewrote in Economy and Society, serve to



‘create conceptual points of orientation forparticular purposes’. There was no intentionof completeness, no intention of ‘forcinghistorical reality into schemes’ (Weber 1922,p. 154). Like the pragmatists, Weber clearlyconsidered social science knowledge to beproblem oriented.

Both from the vantage point of the standardnear-consensus about scientific realism, andwith reference to the social-philosophicalpreference for deep structures (real or cog-nitive), one might claim that processes aretoo often traced at the event level, ‘realscience’ is to unveil more fundamentalprocesses. Returning to the Senghaas project,for instance, one might claim that an analysisof several development experiences couldbe reduced to a common deep structuralprocess: the development of capitalism. Thetypology of different kinds of developmentoutcomes would disappear, being rejected as‘empiricism’, and there would be one deepstructural driving force. Alternatively, onemight refer to the Nordic development pattern,claiming that it did not result from specifiedcumulative processes but that deep-downdemography/family structure predeterminedthe role of women in the Nordic area (Todd1985, 1987).

Both would bring us closer to or even into‘philosophy of history’ kind of moderniza-tion approaches. The contextualist approach,however, is sceptical of such statementsof deep structures, suggesting instead asociology of knowledge reflection: differentresearch communities converge on certainstylized processes that seem to be the bestanswers to their research questions. Witha variety of research questions, we alsohave a variety of claims about ‘fundamental’forces.

Typologies should not be turned intoessential features of reality. Although theyare empirically grounded, typologies are stillconstructions. The degree to which they‘represent’ is up for discussion, at least if thereis a well developed local research frontier.They most probably have to be changed, asWeber emphasized, if we turn to a new set ofresearch questions.

Amore conventional term for local researchfrontier is ‘the literature’. But given the questfor high theory in the standard framework,this concept plays no important role inthat methodology: in particular there is noreflection on the fact that it is local, i.e. limitedto substantive areas. As we have shown,within the standard framework, researchfrontiers are defined with reference to formaltheory only. Elster’s notion of a ‘toolbox’is adequate, but such a toolbox is irrelevantwithout the substantive elements contained inlocal research frontiers.13 In the contextualistframework, a research frontier consisting offormal theory only is not possible. Thisalso ties in with the pragmatist emphasis onknowledge as problem driven.

Here we reach a conclusion for the practicalsocial research: we need more emphasis onsubstantive types of accumulated knowledge.We have implied that a basic weakness ofboth standard and social-philosophical high-level notions of theory is the denial ofcontextualization as a research craftwork.Typology construction by means of compar-ison is a main way of specifying context.Typologies synthesize available knowledgein a form that allows further comparisonwith reference to a set of research questions.They are maintained, revised and improvedby updating of cases and addition of newcases. A social scientist must commanda repertoire of typologies (logic of con-textualization) as much as they need arepertoire of formal theories (logic of processtracing).

The kind of research craftwork that yieldssubstantive theory is underrated in thecommunity of social scientists. The pursuitof typologies – or substantive theory moregenerally – is weakly institutionalized. Mosttypologies are sketchy and hard to find; andthey give low status publication wise! Manyscholars regard typologies as static. But theyneed not be. They become static if they are notmaintained, upgraded, revised and indexed toperiods. Whereas statistics and econometricmodels are well taken care of in economists’sresearch institutions and statistical offices,and large databases in similar institutions,



typologies are not cared for in the same way;they should be!

We can here specify our first implicationfor the philosophy of the social sciences (seethe end of ‘Mechanisms and process-tracing inthe contextualist framework’): social sciencetoday possesses knowledge in several localresearch frontiers, but this knowledge is notconverging into high theoretical knowledge.Neither substantive nor formal groundedtheory converge in one overarching researchfrontier; the former because it is tied tospecific research fields, the latter because itis formal only. Substantive general theory inthe contextualist sense is also not converging,at least so as long as we require that hightheory be explanatory. The highest theorythat is still explanatory might very well betypological maps such as Rokkan’s, whichare applicable to several outcomes that canbe chosen for explanation. But even thistheory is clearly delimited to a context. Theremight , as we have seen, be several generaltheories. Researchers must learn to manouvreand know how various frontiers emerge asrelevant depending on the research questionasked.14

But even if there is no high-level conver-gence, there might be relations of aggregationand overlap between local research frontiers.Topics can rise to dominance and then fade,but certain clusters might emerge and theremight be synergies. There is no way forsubstantive generalizations in a local researchfrontier to be replaced by formal theory.A logic of falsification is not of much help:rather, substantive generalizations fasten asparts of a local research frontier becausethey are used and improved in the researchcollective.

CONCLUSION

No drawing of distinctions is innocent! Onemight object that our linking of the method-ological frameworks to natural science, thehumanities and social science implies the con-clusion that only the contextualist frameworkis adequate for social science. Admittedly,

our discussion has focused on strong featuresof the contextualist framework vis à vis theother two.

However, this chapter is also the workof a methodologist, and the starting pointwas, after all, that a methodology cannot beconsistent! Although we are inclined to claimthat contextualism has less of a gap betweenideal and reality, theory and explanations, thanthe standard perspective, and that it avoids thepersonal, social-philosophical preoccupationwith fundamental or existential problems thatdeflect attention from thorough empiricalresearch, we do not claim that contextualismis without problems or that it can be taken asentirely consistent in philosophical terms.

If we want to make a plea for contextualism,then, it must be consistent with our intro-ductory discussion of the methodologist’sdilemma! Our claim, therefore, is that if onewants to understand the social sciences, athree-fold division is better than any of theconventional dualisms. It is a major strength ofthe contextualist position that it falls betweenthe two others: it is empirically oriented, as aremany scholars within the standard approach.It can engage seriously with discussions onempirical methods, comparing the differentlogics of qualitative and quantitative empiri-cal research: but it is also capable of reflectingin sociology of knowledge terms and ofdiscussing various kinds of contextualization.It is thus it is on speaking terms with social-philosophers.

We have tried to map, as thoroughly aspossible, the comparative specificity of thecontextualist framework. This is important, asit is so often rejected or embraced as being partof either one or the other two. The third posi-tion must guard against polarization betweennatural science and humanities, which all toofrequently degenerate into mutual parodiesthat serve only to bolster self-righteousidentities. Intervening to undermine suchmethodological polarization, the contextualistposition can temper each of the extremes.It has the potential to inspire more fruitfulapproaches to triangulation of methods andcooperative interdisciplinary work in an erawhen disciplinary divisions are challenged.



There is, then, first a contextualist lessonconcerning the relation between empiricallyoriented fellow social scientists: Triangula-tion of methods should be performed withreference to real differences in the wayresearchers work and how they legitimatetheir research strategies. The message tosocial scientists who employ qualitativemethods is that they should think twice beforethey buy into either a standard or a social-philosophical methodological style.

Second, there is a contextualist mes-sage to professional philosophers of socialscience. The contextualist methodologicalaccount, unlike the standard and the social-philosophical, gives professional philoso-phers of social science impressions of thevariety of social science procedures.

Although this chapter was not written toprove that all social science methodologymust converge on a contextualist framework,it presupposes a contextualist methodology.The framework we have used to discernthree methodologies and six notions oftheory has been a contextualist one. Thethree methodologies have been comparedas cases of the same: mediation betweenmethods and selected philosophical elements.By contrast, a standard methodology wouldbe based on normative arguments in favour ofone framework, while a social-philosophicalmethodology would be ripe with accountsof personalized theories. Neither normative,nor personalized, our account traces whatresearchers do when they conduct researchand when they argue about what they do. Bymeans of typologies, contextualized interac-tion patterns, and sociology of knowledge wecontextualize the case of social science in theearly twenty-first century.

The research behind this chapter is – at leastimplicitly – based on a coding of propertiesof academic social science research, reflectingthe author’s ‘participant observation’ in Nor-wegian, Nordic, European and US academicspheres over more than 25 years. The authormaps the contemporary situation in socialscience, clearly accepting a role as participantin this research community. Social scienceconcepts, we know, are ‘interactive kinds’

(Hacking 1999); it does matter how weclassify ourselves and our fellows. Hopefully,an increasing number of social scientists willfind it useful to think of themselves as beingguided by a contextualist methodology.

NOTES

1. The dualism was coined by German Neo-Kantian philosophers (cf. Collins 1998 chapter 13),but later appeared in other academic cultures (cf. e.g.Snow 1959).

2. Some of these can be traced back to Hacking(1999 chapter 3), but we rely more strongly on thesociology of knowledge.

3. This definition partially converges with thedefinition of the ‘received view’ in Hands (2001).Examples of standard methodology: Friedman (1953)in economics; King, Keohane and Verba (1994),Geddes (2003) in political science; Stinchcombe(1968), Goldthorpe (2000) in sociology; Pelto andPelto (1978) in anthropology; Shadish, Cook andCampbell (2002) in psychology.

4. Such as Lyotard (1979), Habermas (1981),Giddens (1985), Alexander (1983). One out of manyoverviews is Best/Kellner (1991).

5. Abbott (1999 pp. 196ff) uses the term‘contextualist paradigm’ in his plea for a returnto the programme of the interwar Chicago schoolof sociology. Even in the 1940s, Stephen Pepper(1942 pp. 232ff) distinguished contextualism as oneout of four ‘world hypotheses’, referring mainly toUS pragmatist philosophy. I earlier used the terms‘pragmatist’ and/or ‘participationist’ (Mjøset 2005,2006a, 2006b), but the term ‘contextualist’ is a moreneutral label and avoids the identification with anyparticular philosophical school.

6. Examples of this position: Hands (2001),Hoover (2001), Mirowski (2002) in economics; Barnes(2001), Abbott (2001), Ragin (2000), Mjøset (2006a,2006b) in sociology; Cicchetti and Rogosch (1996),Gottlieb and Halpern (2002), Biglan (2004) in psychol-ogy. In much recent political science, one often findsstrong elements of contextualism within frameworksthat try to retain standard features. Examples arePierson (2004), Goertz (2006), George and Bennett(2005), and several contributions (e.g. McKeown2004) in Brady and Collier (Eds) 2004. Unfortunately,a closer discussion of these various combinations isbeyond the scope of this chapter.

7. This is a way to state the ‘Thomas theorem’;cf. Merton (1968, chapter XIII).

8. In contrast to Hedström, Coleman (1990p. 5) explicitly denies that ‘for a given purposean explanation must be taken all the way tothe individual level to be satisfactory’. He insteadinvokes a ‘pragmatic’ criterion: ‘The explanation is



satisfactory if it is useful for the particular kinds ofintervention for which it is intended.’ In that case,the whole debate on methodological individualismversus methodological collectivism must be judgeda rather uninteresting polarization between thestandard vision of elementary particles and thesocial philosophical totalizing orientation. Standardformulations are ripe with reservations accepting thatthe ideal of methodological individualism can seldombe realized. Contextualism simply accepts this.

9. We have no space to pursue this further here,but the controversial/non-controversial distinction isimplied in our discussion of both critical theory andlocal research frontiers below.

10. This should fit Ragin’s (1986) notion ofmultiple, conjunctural causation, as well as notionsof equifinality and multifinality in the open-systemsliterature (Cicchetti and Rogosch 1996), but there is nospace to expand on this hunch further here. Note thatin standard accounts (Hedström 2005), the inclinationis often to talk about mechanisms in singular.

11. Although we have no space to pursue it here,the discussion of various types of problem (problemsof social engineering, conflict-related problems, exis-tential problems, theoretical problems, etc.) would bea fruitful specification.

12. The standard interpretation is, for instance,Coleman’s (1990, chapter 1) model of explanationin social science, and the social-philosophical inter-pretation is in terms of collective belief systems,civilizations, etc.

13. Elster (1989, 2006) suggests a toolbox offormal mechanisms. We suggest a tool-shed ofgrounded theories. In this shed there are many shelvesfor substantive theories: typologies, periodizations,stylized facts. There is also a place for Elster’stoolbox, but these mechanisms can only provideexplanations in contexts established by substantivetheory. Elster tends to leave the tool-shed carrying thetoolbox, forgetting to bring any of the substantivetheories along. In this way, he retains the standardpreoccupation with formal theory, although he oftendisplays an unhappy consciousness that somethingimportant has been left behind.

14. Social philosophers are more interested inthe conditions of knowledge than in the actualaccumulation of knowledge in local research frontiers.Transcendental theory is personal, and so are theresulting interpretations of modernity. This is theposition of the literary intellectual, the ability toexpress the deepest concerns of a generation (orsome other unit – a nation, a people, etc.) in anindividual synthesis (the work of art). It is interestinghere to consider Flyvbjerg’s (2001) programme ofphronetic social science. This parallels contextualistsocial science in that both are problem driven.However, thanks to its notion of local researchfrontiers, the contextualist approach avoids a majordilemma in Flyvbjerg’s account, namely that only

the high-theory ideal of the standard approach isconsidered to be theory. Flyvbjerg claims that becauseof ‘context-dependence’, ‘cumulative’ and ‘stable’research is not possible in the social sciences. Thecontextualist framework outlined here does not implythat analysis of context precludes theory. One shouldnot grant the standard position a monopoly onnotions such as science and theory, but rather definethese with reference to local research frontiers. Thisnote is relevant given Laitin’s (2003) use of Flyvbjergas a proxy for the unknown ‘Mr Perestroika’ in recentcontroversies within US political science. The resultwas yet another polarization (Flyvbjerg 2003 was theresponse) between standard and social-philosophicalpositions, in which the real merits of a contextualistposition gets lost.

REFERENCES

Abbott, A. 1999 Department and Discipline. Chicago:University of Chicago Press.

Abbott, A. 2001 Time Matters. Chicago: University ofChicago Press.

Alexander, J.C. 1983 Theoretical Logic in Sociology(4. volumes). Berkeley, CA: University of CaliforniaPress.

Barnes, B. 2001 On Agency. London, Sage.Best, S. and Kellner, D. 1991 Postmodern Theory: Critical

Interrogations. New York: Guilford Press.Biglan, A. 2004 ‘Contextualism and the development

of effective prevention practices’ Prevention Science5(1): 15–21.

Blumer, H. 1969 Symbolic Interactionism. EnglewoodCliffs, NJ: Prentice Hall.

Brady, H.E. and Collier, D. (Eds) 2004 Rethinking SocialInquiry. Lanham, MD: Rowman & Littlefield.

Cicchetti, D. and Rogosch, F.A. 1996 ‘Equifi-nality and multifinality in developmental psy-chopathology’ Development and Psychopathology8: 597–600.

Coleman, J.S. 1990 Foundations of Social Theory.Cambridge, MA: Harvard University Press.

Collins, R. 1986 Weberian Sociological Theory.Cambridge, Cambridge University Press.

Collins, R. 1998 The Sociology of Philosophies.Cambridge, MA: Belknap.

Dewey, J. 1920/1950 Reconstruction in Philosophy(1950 edition). New York: Mentor.

Elster, J. 1989 Nuts and Bolts for the Social Sciences.Cambridge, Cambridge University Press.

Elster, J. 2007 Explaining Social Behavior. Cambridge,Cambridge University Press.

Flyvbjerg, B. 2001 Making Social Science Matter.Cambridge, Cambridge University Press.



Flyvbjerg, B. 2003 ‘A perestroikan straw man answersback’ Politics and Society 32(3): 389–416.

Foucault, M. 1969 L’archéologie du Savoir. Paris:Gallimard.

Foucault, M. 1975 Surveillir et Punir. Paris: Gallimard.Friedman, M. 1953 ‘The methodology of positive

economics’ in M. Friedman (Ed.) Essays in PositiveEconomics. Chicago: University of Chicago Press,pp. 3–43.

Geddes, B. 2003 Paradigms and Sand Castles. AnnArbor, MI: University of Michigan Press.

George, A.L. and Bennett A. 2005 Case Studiesand Theory Development in the Social Sciences.Cambridge, MA: MIT Press.

Giddens, A. 1985 The Constitution of Society.Cambridge: Polity Press.

Giddens, A. 1990 The Consequences of Modernity.Cambridge: Polity Press.

Glaser, B.G. and. Strauss, A.L. 1967 The Discovery ofGrounded Theory. New York: Aldine de Gruyter.

Goertz, G. 2006 Social Science Concepts. A User’sGuide. Princeton, NJ: Princeton University Press.

Goldthorpe, J. 2000 On Sociology, Oxford: OxfordUniversity Press.

Gottlieb, G. and Halpern, C.T. 2002 ‘A relational viewof causality in normal and abnormal development’Development and Psychopathology 14: 421–435.

Granovetter, M. 1985 ‘Economic action and economicstructure. The problem of embeddedness’ AmericanJournal of Sociology 78: 1360–1380.

Habermas, J. 1981 Theorie des kommunikativenHandelns. Frankfurt am Main: Suhrkamp.

Hacking, I. 1983 Representing and Intervening.Cambridge, Cambridge University Press.

Hacking, I. 1999 The Social Construction of What?Cambridge, MA: Harvard University Press.

Hands, D.W. 2001 Reflection without Rules, EconomicMethodology and Contemporary Science Theory.Cambridge: Cambridge University Press.

Hedström, P. 2005 Dissecting the Social. Cambridge,Cambridge University Press.

Hirschman, A.O. 1977 ‘A generalized linkage approachto deveopment, with special reference to staples’Economic Development and Cultural Change, 25(supplement) (Festschrift for Bert F. Hoselitz).

Hoover, K. 2001 The Methodology of EmpiricalMacroeconomics. Cambridge: Cambridge UniversityPress.

Horkheimer, M. 1937 ‘Traditionelle und kritische Theo-rie’ Zeitschrift für Sozialforschung VI(2): 245–292.

King, G., Keohane, R.P. and Verba, S. 1994 DesigningSocial Inquiry. Scientific Inference in Qualita-tive Research. Princeton, NJ: Princeton UniversityPress.

Laitin, D. 2003 ‘The perestroikan challenge to socialscience’ Politics and Society 31(1): 163–184.

Lieberson, S. 1985 Making It Count: The Improvement ofSocial Research and Theory. Berkeley, CA: Universityof California Press.

Lyotard, J.-F. 1979 The Postmodern Condition (Englishtranslation, 1984). Minneapolis, MN: University ofMinnesota Press.

McKeown, T.J. 2004 ‘Case studies and the limitsof the quantitative worldview’ in H.E. Brady andD. Collier (Eds) Rethinking Social Inquiry. Lanham,MD: Rowman & Littlefield, pp. 139–167.

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

Mirowski, P. 2002 Machine Dreams. Cambridge:Cambridge Unversity Press.

Mjøset, L. 2000 ‘Stein Rokkan’s thick comparisons’ ActaSociologica 43: 4.

Mjøset, L. 2005 ‘Can grounded theory solve theproblems of its critics?’ Sosiologisk Tidsskrift 13(4):379–408.

Mjøset, L. 2006a ‘A case study of a case study. Strategiesof generalization and specification in the study ofIsrael as a single case’ International Sociology 21(5):735–766.

Mjøset, L. 2006b ‘No fear of comparisons or context: onthe foundations of historical sociology’ ComparativeEducation 4(5): 337–362.

Mjøset, L. 2007 ‘An early approach to the varietiesof world capitalism: methodological and substantivelessons from the Senghaas/Menzel-project’ Compar-ative Social Research 24: 123–176.

Mjøset, L. and Cappelen Å. 2009 ‘Economics andthe others’ Nordic Journal of Political Economy(in press).

Mjøset, L. and Clausen, T.H. 2007 ‘An introduction tothe comparison of capitalisms’ Comparative SocialResearch 24: 1–17.

Pawson, R. 2000 ‘Middle range realism’ ArchivesEuropeennes de Sociologie 41(2): 283–325.

Pelto, P.J. and Pelto, G.H. 1978 Anthropologicalresearch. The structure of inquiry. Cambridge:Cambridge University Press.

Pepper, S.C. 1942 World Hypotheses. Berkeley, CA:University of California Press.

Pierson, P. 2004 Politics in Time. Princeton, NJ: PrincetonUniversity Press.

Ragin, C.C. 1986 The Comparative Method. Berkeley,CA: University of California Press.

Ragin, C.C. 2000 Fuzzy-set Social Science. Chicago:University of Chicago Press.

Ragin, C.C. 2008 Redesigning Social Inquiry: FuzzySets and Beyond. Chicago: University of ChicagoPress.



Rokkan, S. 1999 State Formation, Nation-Building, andMass Politics in Europe. Oxford: Oxford UniversityPress.

Seidman, S. 1991 ‘The end of sociological theory: thepostmodern hope’ Sociological Theory 9: 131–146.

Senghaas, D. 1985 The European Experience.Leamington Spa, UK: Berg.

Shadish, W.R., Cook, T.D. and Campbell. D.T. 2002Experimental and Quasi-experimental Designs forGeneralized Causal Inference. Boston: HoughtonMifflin.

Shalev, M. 2007 ‘Limits and alternatives to multipleregression in comparative research’ ComparativeSocial Research 24: 261–308.

Skagestad, P. 1978 Vitenskap og Menneskebilde.Charles Peirce og Amerikansk Pragmatisme. Oslo:Universitetsforlaget.

Skocpol, T. 1984 ‘Emerging agendas and recurrentstrategies in historical sociology’ in T. Skocpol(Ed.) Vision and Method in Historical Sociology.

Cambridge: Cambridge University Press, pp.356–391.

Smith, D. 1999 Writing the Social. Toronto: University ofToronto Press.

Snow, C.P. 1959 The Two Cultures. Cambridge:Cambridge University Press.

Stinchcombe, A.L. 1968 Constructing Social Theories.New York: Harcourt, Brace, Jovanovitch.

Strauss, A.L. 1970 ‘Discovering new theory fromprevious theory’ in. T. Shibutani (Ed.) Human Naturesand Collective Behavior. Papers in Honor of HerbertBlumer. Englewood Cliffs, NJ: Prentice Hall.

Todd, E. 1985 The Explanation of Ideology: FamilyStructures and Social Systems. Oxford: Blackwell.

Todd, E. 1987 The Causes of Progress: Culture, Authorityand Change. Oxford: Blackwell.

Veblen, T. 1919 The Place of Science in Civilization.New York: Viking.

Weber, M. 1922 Wirtschaft und Gesellschaft, Tübingen,Mohr.

the contextualist approach to social science methodology · social science methodology ... the...

Documents