Steinmetz, G. (1998) - Critical Realism and Historical Sociology. A Review Article. Comparative Studies in Society and History, 40 (01), 170-186.
Steinmetz, G. (1998) - Critical Realism and Historical Sociology. A Review Article. Comparative Studies in Society and History, 40 (01), 170-186.
Steinmetz, G. (1998) - Critical Realism and Historical Sociology. A Review Article. Comparative Studies in Society and History, 40 (01), 170-186.
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Cambridge University Press and Society for Comparative Studies in Society and History are collaborating with
JSTOR to digitize, preserve and extend access to Comparative Studies in Society and History.
http://www.jstor.org
Perhapsthe fiercest conflict within the social sciences today is one that is not
even articulatedas a recognizable "debate."Nevertheless, this conflict has
generatedbitter divisions between academics who only a decade ago might
have stood together on the left or the right and has forged equally strange
alliances. I am referring,of course, to the split between epistemological and
ontological positions which are usually described as radical constructivism
and realist positivism. The recent debate provokedby Alan Sokal's article in
Social Text' broughtinto sharpfocus the growing sense of distrustand anger
that divides these academic camps. And, yet, the heterogeneityof each supposed groupingsuggests thatthe dichotomousmodel masks what is actuallya
much more complex, multi-dimensionalfield. On the one hand we find a
motley assemblage of positions variously characterizedas constructivism,
culturalism,neo-Kantianidealism, and postmodernism;the other pole throws
togethera set of even strangerbedfellows, includingrationalchoice theorists,
survey researchers,and traditionalhistoriansalongside "realist"philosophers
of various stripes.
In this articleI will disaggregatethese artificialgroupingsin orderto clarify
the issues in the debate which are importantfor historical social research. I
start from the observation that the actual writing of most historical social
scientists is not accuratelycapturedby the descriptionof either of the two
camps. To make better sense of our own practical epistemological and ontological views, we need first to break open the two caricaturedpositions of
relativist discursivismversus hard-headedpositivism. Each of these actually
consists of several distinctpositions. The supposedlyunitarypositivist-realist
pole encompassespositivism and several varietiesof realism.The explanatory
I See Alan D. Sokal,
"Transgressingthe Boundaries-Toward a TransformativeHermeneutics
of Quantum Gravity," Social Text (Spring-Summer 1996), 217-52; and idem, "A Physicist
Experimentswith CulturalStudies,"Lingua Franca (May-June1996), 62-64. The key response
is StanleyFish's article,"ProfessorSokal's Bad Joke,"in TheNew YorkTimes(Opinion-Editorial
page, May 21, 1996); see also the responses in the July-August 1996 issue of Lingua Franca;
Michael Albert, "Science, Post Modernism,and the Left," in Z Magazine (July-August 1996);
Steven Weinberg,"Sokal'sHoax,"New YorkReview of Books (August 8, 1996), 11-15; and Tom
Frank,"TextualReckoning,"In These Times(May 27, 1996). I am gratefulto Michael Rosenfeld
for references to the latter two pieces.
0010-4175/98/1601-0900$9.50? 1998SocietyforComparative
Studyof SocietyandHistory
170
I7I
I72
GEORGE STEINMETZ
ALTERNATIVES
TO CRITICAL
REALISM
of cause and effect at the level of events" (Collier 1994:7). The untheorized
presentationof empirical regularitiesis preferredto so-called "metatheory."
As Somers (forthcoming) has argued, positivist empiricism needs to be
distinguishedfrom theoreticalrealism, a position most strongly associated in
the social sciences with rationalchoice theory (see Kiser and Hechter 1991).
On the one hand, theoretical realism is post-positivist (and Kuhnian) in its
acceptanceof the "primacyof theory"and of a strong ontology "thattells us
what really exists" (Somersforthcoming).Theoreticalrealism'sprioritizingof
ontology over epistemology and its acceptanceof unobservablecausal entities
are featuresthatit shareswith criticalrealism, as discussed in the next section.
6 For a recent defense of positivism against "postmodernrelativism,"see JonathanTurner
(1992).
7 For an influential and typical example, see King, Keohane, and Verba (1994), who distinguish what they call "description"from "causalinference"and both of these from the searchfor
"causal mechanisms." Causal inference involves familiar multivariate statistical or quasiexperimentalmethods, in which the effect of some variable is discovered independentof other
variables.This independenteffect is termed"causal."They continuethat "we can define a causal
effect withoutunderstanding. .. the causal mechanismsinvolved" (1994:86). In the languageof
critical realism, the emphasis on identifying causal effects but not causal mechanisms is none
otherthan "description"of empiricalconjunctions.At a laterpoint, the authorsexpress skepticism
aboutthe value of concepts referringto unobservedtheoreticalentities (1994:109-110). Unobservable concepts are described as an inferioralternativeto empiricalones ratherthan as a fundamentally differentand equally importantpart of explanation(see part II).
I73
those who try to use purelypredictivetests of causal hypothesesin open systems . . . are liable to
be guilty of 'nafve falsification,' in which an anomaly due to interference from some other
mechanisms is treated as a falsification of the causal claim in question, e.g. "aeroplanesare
heavier than air but can fly, thereforethe law of gravity is refuted"(1992:212-4).
9 Along similar lines, see D'Amico (1992:142): "Postmodernismappearsto be an anti-realist
accountof knowledge and science because it treatsrepresentation,correspondence,and reference
as dependenton a conceptual frameworkor scheme."
10 It is inconsequentialfor science whetherrealityis arguednot to exist independentlyof human
thoughtor simply to be inaccessible.In eithercase, the connectionbetweenthe levels of the "actual"
and the "empirical"(see below) is severed. Some super-idealistsinsist that they believe in the
existence of a concept-independentexternalrealityand thatthe only questionis its accessibility to
the scientific observer.This counter-argumentis irrelevantor disingenuous,since eitheralternative
has the same implications for science. See Latour and Wollgar (1986) and Woolgar (1988) for
"super-idealist"argumentsabout science creating the objects it purportsto discover.
I74
GEORGE STEINMETZ
175
the social sciences, which I will returnto below. Idealists fail to distinguish
explicitly between ontology and epistemology, between "the . . . intransitive
objects of science . . . and the changing (and theoretically-imbued) transitive
of thought,
176
GEORGE STEINMETZ
naturalismwith its transcendentalrealism.14 Before turningto criticalnaturalism and the specificity of socio-historicalexplanation,however, we first need
to presentthe argumentsfor realismin some detail. Indeed,the key arguments
for the critique of positivism, theoreticalrealism, and conventionalistsuperidealism are alreadypresent in Bhaskar'sfirst book.
CRITICAL
REALISM
Bhaskar's"transcendental"
argumentfor realism borrowsfrom Kant its transcendentalform of questioning but inverts Kant's idealism, asking not what
must be trueaboutcategoriesof mind in orderfor synthetica priorijudgments
to be possible but insteadwhat must be true aboutthe world for science to be
possible.15More specifically, Bhaskarasks what must be trueaboutrealityfor
scientific experimentsto be intelligible.He notes, first,thatexperimentswould
be unnecessaryif the positivist's "constantconjunctionsof events"were not in
fact extremely rare in nature. Constant conjunctions have to be produced
artificially.Scientists rely on experimentsto createclosed systems, since most
domainsof nature-and not just social systems-are open systems. Science's
ability to apply knowledge gained throughexperimentsto the outside world
implies that the same causal laws are operativeinside and outside the laboratory.The experimenteris responsiblefor triggeringthe mechanismunderstudy
andpreventinginterferencefromothermechanisms,but she does not createthe
mechanism that is revealed by the sequence of events.
The argumentthus suggests an ontological differencebetween the operation
of causal mechanismsand the patternsof events they codetermine.Only if we
assume that underlyingcausal mechanisms are independentfrom the events
they generatecan we assume that they endureand continue acting outside of
the experimentallyclosed conditions which allow scientists to identify them
empirically.This out-of-phase-nessbetween causal laws and actualphenomena leads Bhaskarto differentiatebetween what he labels the domains of the
real and the actual, which correspondrespectively to the realms of mechanisms and events (see Figure 1). He furtherdistinguishesthe actual from the
empirical,noting that mechanismsmay be realized (at the level of the actual)
withoutbeing perceived(at the level of the empirical).A scientific experiment
aligns these three levels: Real mechanismsare isolated, assuringthat they can
produce actual events without their powers being overridden or combined
with those of other mechanisms;in a successful experiment,these events are
then recordedby the scientist.In sum, things have unexercisedpowers, as well
as powers that are exercised unrealized,and powers that are realized unperceived (Bhaskar 1975:19, 33).
14 Bhaskar
adopted this self-description until his recent expansion into "dialectical critical
realism"(1993, 1994).
15 That science is
possible and successful is amply demonstratedby, for example, Hacking
(1983). Theory postulated subatomic particles and then found them, and now "you can spray
them."This does not imply that the currenttheory of subatomicparticlesis finally and eternally
truebut does suggest that realityexists independentlyand thatthe theorygraspssome aspect of it.
177
Domain
of Actual
Mechanisms
Events
Experiences
Domain
of Empirical
This analysis suggests that in open systems, unlike the artificial closure
characteristicof the experimentalsituation,mechanismscombine to produce
actual events conjuncturally,that is to say, in concert with other mechanisms
(Bhaskar 1975:17). As noted in the previous section, contingencyhere means
thatcomplex events arecodeterminedby constellationsof causal mechanisms.
Contingency also implies that such constellations are not repeatable in a
general way and also that the componentsthat make up the causally effective
constellationmay vary.16Recognition of the ontological reality of contingency thus rules out the search for constant conjunctionsof events as a normal
featureof science.17 But because events are caused, contingency is combined
with "necessity" (see Jessop 1990:12).18 Bhaskar provides a clear way of
visualizing the conjuncturaldeterminationof an event (or a nexus of events)
by a combinationof mechanisms, as reproducedin Figure 2.
Bhaskardoes not rejectthe term "causallaw" but defines it differentlyfrom
positivists. Within critical realism, a law is not a constant conjunction of
events but the characteristicpattern of activity, or tendency, of a mechanism.19More specifically, real structurespossess causal powers which, when
16 For an
explanationinvolving causal contingency in the naturalsciences (biological evolution), see Gould (1989).
17 Note that formal
positivists are sometimes willing to accept historicalnarrativeas a form of
explanation, provided that each consecutive event is amenable to rephrasingin terms of an
empirical covering law (e.g., Nagel 1961). Obviously historical narrativeconstruedin this way
does not reveal the form of causal contingencydescribedin the text. Whatis contingenthere is the
sequence of explananda,not the explanans.
18 As Jessop writes,"'Contingent'is a logical concept and concernedwith theoreticalindeterminability, 'necessity' is an ontological concept and refers to determinacyin the real world. Thus
'contingent' means 'indeterminablewithin the terms of a single theoretical system'; it can
properlybe juxtaposed to the notion of 'necessity', which signifies the assumptionunderpinning
any realist scientific enquiry that 'everythingthat happens is caused'" (1990:12).
19 Generative mechanisms are "tendencies"ratherthan "powers"because they are not just
potentialities but potentialities that may be exercised without being manifested (Bhaskar
1975:50).
178
GEORGE STEINMETZ
Conjunctural Determination
M1
M2
M2
M3
M4
M3
\M1
Eo
Eo
Case 1
Case 2
FIGURE
2. Conjuncturaldeterminationof the nexus of events by a totalityof mechanisms.Case
1 is the determinationof events in an open system. Case 2 is the determinationof a nexus of
events by a totality of causal mechanisms in an open system. SOURCE:adaptedfrom Bhaskar,
Scientific Realism and Human Emancipation, 100.
CRITICAL
REALISM
AND
HISTORICAL
SOCIOLOGY
I79
I80
GEORGE STEINMETZ
real the the actual have not been aligned throughexperimentalclosure, the
task of explanationdiverges sharplyfrom the work of theory.In open systems
and the practical or applied sciences, "explanationis accomplished by an
account of the . . . mode of combination or inter-articulation, in some specific
Although centered on the naturalsciences, Bhaskar's transcendentaldeduction in Realist Theoryof Science is crucial to his subsequentdiscussions of
social science, in both a negative and a positive sense.24By dispelling various
positivist, theoretical realist, and idealist misconceptions about the natural
sciences, it underminesan entire series of restrictivenotions aboutthe sort of
knowledge the social sciences should aspire to. Most important,if regularity
23 Bhaskar
(1975:38). This step is obviously the one that has been most conflicted and
analyzed since Kuhn in terms of the categories of irrationalityor incommensurability.Bhaskar
criticizes the notion of incommensurabilitywithin paradigm shifts (1975:191ff) and argues
againstjudgmentalrelativismas follows: If Theory 1 can explain more significantphenomenain
terms of its descriptionsthan Theory 2 can explain in terms of its descriptionsof the same or
overlappingtheory-independentworld, then there is a rationalcriterionfor choosing T1 over T2,
whether or not it is 'newer' or even if they are 'incommensurable"'(Bhaskar 1994:51).
24 Thereis not
enough space here to go into Bhaskar'ssubstantivesocial theory,except to note
that much of it clearly relates to and possibly anticipatesthe work of Giddens,while differingin
significantways (see Archer 1995 for a systematiccomparisonof the two). I also cannot discuss
Bhaskar's project of "moral realism," which is presented in part as a resolution to the
communitarian-liberalimpasse within political philosophy.Bhaskardissolves the fact-value distinction in favor of a "moralrealism"(see Bhaskar 1991; see also the recent development of a
non- or post-Hegelian "dialectics"in Bhaskar 1991, 1993, 1994).
CRITICAL
REALISM
AND
HISTORICAL
SOCIOLOGY
I8I
I82
GEORGE
STEINMETZ
social science may have over, say, survey research.The objects of survey or
ethnographicresearch are determinedat least in part by the social scientist
(Burawoy forthcoming). By contrast, while our knowledge of the past is
constantly changing, the past itself is "existentially intransitive and determined" (Bhaskar 1994:72). Of course, past historical structuresmay have
been partly-or even wholly-constituted by past forms of (scientific)
knowledge, and the latterwill thereforeneed to be reconstructed.But it will be
more difficult, ceteris parabus, for a survey researcheror social psychologist
to perceive the causal impact of her own theories than for a historical social
scientist to discern the causal impact of earlier theories on her object of
explanation.This is only a generalrule, since historianswhose beliefs closely
resemble those of past actors may also fail to recognize the causal role of
those actors' beliefs. Reflexivity is also requiredof historicalresearchers,but
that is reflexivity of a different sort.
A second key difference between the social and naturalsciences concerns
the greaterspace-time specificity of social as opposed to naturalmechanisms,
such that even the underlying tendencies they ground may not be invariant
across more than a limited period or territory.This provides yet another
argumentagainstthe enthronementof universallaws along eitherpositivist or
theoretical realist lines. And this difference vindicates in another sense the
historical social sciences, which typically have been much more prone to
emphasizechanges in causal structureacross time and space thanto view such
changes as a barrierto scientificity.28
The obvious drawbackto the social as opposed to (most of) the physical
sciences is the impossibility of true experimentation. The use of the
RRREI(C)model (see above) becomes more difficult in the social sciences.
Since experimentsare implausible,thereis less securityaboutthe existence of
causal mechanisms,which complicates the passage throughthe third, fourth,
and fifth stages in particular.In place of experiments,Bhaskarhas suggested
the use of "transcendentalarguments from premises familiar from social
practice"(Collier 1994:167), for example, argumentsabout"whatmust be the
case for the experiencesgraspedby the phenomenalforms of capitalistlife to
be possible" (Bhaskar 1979:51). More recently, Bhaskar(1994:94) has suggested that an alternativemodel which he proposedin an earlierwork (1986),
28 Such variations in causal structureare a central,
explicit part of my account of local and
national social policy development in nineteenth-centuryGermany, for example (Steinmetz
1993). Reviewers inclined toward positivism or theoreticalrealism have misunderstoodthe explanatorycomplexity which results from explicit acknowledgmentof conjunctural,multi-causal
and historically shifting causality as a kind of theoretical "polymorphousperversity"(one reviewer's particularlyrevealing choice of terms, given Freud's status as a preeminentcritical
realist thinker),or less colorfully, as a "toolbox"approachto theory.Accountingfor the determination of complex objects in open systems necessarily involves an "eclectic" mix of theories
relating to the relevant causal mechanisms.This is quite differentfrom the empiricist "variablism" found in much multivariatestatisticalresearch,where variablesare connectedto theoretical
mechanismsin a loose and ad hoc way.
CRITICAL
REALISM
AND
HISTORICAL
SOCIOLOGY
I83
the so-called DREI model, may be applicable in the social sciences. DREI
stands for the "description of law-like behavior; retroduction, exploiting analogies with already known phenomena, to possible explanations of the behavior; elaboration and elimination of alternative explanations; issuing (ideally)
in the empirically controlled identification of the causal mechanism(s) at
work" (Bhaskar 1986:68). Collier (1994) is justifiably doubtful about the
applicability of the DREI model in the social sciences, however, citing in
particular the impossibility of the final stage which is "tied to experimental
closure" (1994:163). He suggests instead an amalgam of the two models
(1994:163-4). We begin with the first stage of the RRREI(C) model, resolving a complex event into its components through abstraction (see Sayer 1992,
for a good discussion of abstraction in this sense). This step would seem to be
necessary in the social sciences, since it is not only true that the mechanisms
producing events are multiple but also that events themselves are "overdetermined" in Althusser's (and Freud's) sense (see Eo in case 2, Figure 2). The
next three stages are drawn from the DREI model: description, retroduction,
and elaboration and elminiation of alternative explanations. "Elimination" is
the final stage, however, since definitive identification is impossible.
There is nonetheless a "compensator" for the lack of experiments. This
compensator starts from the "protoscientific" theories about society held by
social actors but transforms these into theory. Theoretical transformation of
this sort involves "retroductive" transcendental arguments "from premises
familiar from social practice" or from actors' own understandings of society
(Collier 1994:167). A retroductive argument is one that necessitates "the
building of a model of the mechanism which, if it were to exist and act in the
postulated way, would account for the phenomenon concerned" (Bhaskar
1986:61). But we cannot stop here, without empirical testing, without falling
into the trap of neo-Kantian super-idealism:
Whether or not the postulated mechanism acts in the postulated way cannot of
course be decided by theory alone, since in general a pluralityof possible explanations
will be consistentwith the phenomena ... So the realityof the conjecturedmechanism
must be empirically ascertained,and the variety of plausible alternativeexplanations
sorted,elaboratedand eliminateduntil the explanatorymechanismat work has been, in
the fallible judgement of the scientists concerned, successfully identified and adequately described (Bhaskar 1986:61).
CONCLUSION
Bhaskar notes that critical realism "transposed to the human sciences appears
immediately liberating." It allows social scientists who are attracted to cultural
theory and complex conjunctural forms of explanation to defend themselves
against being lumped together into undifferentiated categories of "postmodernism" or "eclecticism." Arguing for the reality of a historical "postmodern
condition" qua "cultural dominant" (Jameson 1984, 1991) does not require
adoption of "postmodern" judgmental relativism, any more than understand-
184
GEORGE
STEINMETZ
ing post-1789 Prussia requires that one become a Hegelian.29 By the same
token, arguing for the powerful causal effects of discourse on social practice is
not equivalent to the claim that society is a text for which all interpretations
are equally valid. And scaling back the predictive claims of sociology does not
entail relinquishing explanation.
Critical realism is especially "liberating" for historical sociology. It provides a rebuttal to the positivist and theoretical realist insistence on the
dogmas of empirical invariance, prediction, and parsimony (see Bhaskar
1989:184). Critical realism guards against any slide into empiricism by showing why theoretical mechanisms are central to all explanation. At the same
time, critical realism suggests that contingent, conjunctural causality is the
norm in open systems like society. Yet critical realism's epistemological relativism allows it to accept the results of much of the recent history and sociology of science in a relaxed way without giving in to judgmental relativism.
Historical social researchers are reassured of the acceptability of their scientific practice, even if it does not match what the mainstream misconstrues as
science. Critical realism allows us to safely steer between the Scylla of constricting definitions of science and the Charybdis of solipsistic relativism.
29
Similarly, arguing that scientific discourse codetermines social objects is different from
arguingthat science creates the only world we can know.
REFERENCES
CRITICAL
REALISM
AND
HISTORICAL
SOCIOLOGY
185
i86
GEORGE STEINMETZ