Critical Realism and the Limits of Philosophy

Critical Realism and the Limits ofPhilosophy

Stephen KempUNIVERSITY OF EDINBURGH, UK

AbstractThis article critiques the idea that, by establishing a general frameworkwithin which research must be conducted, philosophical argument can ‘takethe lead’ in relation to research. It develops Holmwood’s work in this areaby examining the ontological arguments put forward by critical realists,which attempt to establish the fundamental characteristics of the socialrealm prior to the production of empirically successful research in that realm.The article draws on a contrast with ontological argument in the naturalsciences to demonstrate the illegitimacy of this manoeuvre, showing thatontological claims can be given some justification, but only when they arederived from research that is widely held to be empirically successful. Realistontological claims in the social sciences do not have this basis, and it isargued that Bhaskar’s alternative mode of justification for these claims isunconvincing. Archer’s view is also criticized that critical realist argumentsshould be given a strong regulatory role in relation to research, illustratingthe problems with this by critiquing Cruickshank’s ontologically driven analysisof unemployment and the underclass. The article concludes that socialscientific research should be conducted without philosophical legislation.

Key words■ Bhaskar ■ critical realism ■ general theory ■ Holmwood ■ ontology

The role that philosophy can be expected to play in relation to substantiveinquiry is an important, long-standing issue within social science. Can we expectphilosophy to take the lead, establishing an appropriate framework for socialscientific thinking prior to actual research activities? Or are other, more modestroles appropriate, such that, for example, philosophy may clarify issues arisingfrom research, or offer potentially productive speculations for researchers?Although there are, of course, many tendencies and counter-tendencies intwentieth-century philosophy, I would suggest that one important line ofthought has emphasized the limitations of philosophical argument, criticizing theidea that philosophy can lay down the rules of reasoning by which all other formsof thought must be judged (see, for example, Wittgenstein, [1953] 1978; Rorty,

European Journal of Social Theory 8(2): 171–191

Copyright © 2005 Sage Publications: London, Thousand Oaks, CA and New Delhi

www.sagepublications.com DOI: 10.1177/1368431005051762

04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 171

at PENNSYLVANIA STATE UNIV on March 5, 2016est.sagepub.comDownloaded from

http://est.sagepub.com/

1980). The influence of this kind of critique is apparent, for example, within thediscipline of philosophy of natural science, in which the adoption of a moresensitive, contextual position by many practitioners has involved an acknowl-edgement both that forms of reasoning within the sciences have changed overtime, and that different scientific domains often have different forms of reason-ing (see Feyerabend, 1975; Shapere, 1984). Accordingly, philosophers havebecome more cautious about the idea that a single, unchanging set of criteria for,say, a good scientific explanation, can be philosophically identified and then usedto judge the validity of any and all forms of scientific research. Indeed, previousattempts to do so have been exposed as false generalizations of the limited logicof one area of research (usually some part of physics) rather than as truly general,unchanging criteria of adequacy. This does not mean that philosophers of sciencehave given up on any attempt to assess scientific reasoning, but rather that themost persuasive of these assessments are located firmly within particular domainsof research rather than at a general philosophical level (see, for example, Sober,1984; Shapere, 1974).

In light of this anti-foundationalist move within philosophy, and the shifttowards more located, immanent forms of philosophizing, I would suggest thatour expectations of philosophy in relation to social science should be relativelymodest. However, as John Holmwood has pointed out, many social theorists haveother ideas. In his Founding Sociology? Talcott Parsons and the Idea of GeneralTheory (1996) Holmwood identifies a genre of theoretical writing in whichtheorists attempt to provide a unifying general framework for social science, thisframework establishing the basic characteristics of social life by philosophicalargument, independently of a consideration of substantive research. Such theoriesaddress issues such as the nature of social structure and social systems, thecharacteristics of human agency, the nature of power, and so on. Typical examplesof this kind of framework would be Talcott Parsons’s structural-functionalism(Parsons, [1937] 1949, 1951), Anthony Giddens’s structuration theory(Giddens, 1979), Jeffery Alexander’s neo-functionalism (Alexander, 1985a) andJürgen Habermas’s critical theory (Habermas, 1984).1 Although there is a degreeof variety in the motivation for developing these frameworks, one commontheme is that something must be done to go beyond the disagreements, bothphilosophical and substantive, that are rife within social science (see Alexander,1982). The claim is that adopting the theorist’s framework will resolve suchdisagreements, and provide the unified basis that social science has previouslylacked. On this view, philosophical argument will provide a guide for researchersthat leads them along a path towards successful empirical social science.

Having identified this strand of theoretical argument, Holmwood offers anincisive critique of it. In the first place, Holmwood argues, the idea that thecharacteristics of some features of society can be identified independently ofresearch, and will not be challenged and subject to revision by research in thearea, runs counter to the post-positivist emphasis on the historical, falliblecharacter of all knowledge (see Holmwood, 1996: Chapters 2 and 3). Second,adherence to such frameworks would be bad news for research, argues

European Journal of Social Theory 8(2)1 7 2

04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 172



Holmwood, because, in the reach for generality, the frameworks becomesomewhat indeterminate (1996: Chapters 4–6). Research expressed in the idiomof general theory loses its sharpness, and it becomes unclear what the empiricalclaims in question actually rule in and rule out. Finally, Holmwood points outthat the general theorists who attempt to set up a framework for research thatmust be accepted by everyone cannot even agree among themselves about theappropriate character of this framework, spending much time critiquing eachother’s positions (for examples of such debates, see Giddens, 1977; Alexander,1985b). Given the recurring problems with general theoretical frameworksidentified by Holmwood, I would follow him in arguing that social scientistsshould abandon them in favour of empirical explanation and problem-solvingand focus on producing compelling substantive theories about the social world.

In this article I would like to extend Holmwood’s critique of general socialtheory by considering the critical realist perspective developed by Roy Bhaskar([1975] 1997, [1979] 1998), Margaret Archer (1995) and Andrew Sayer (1992),among others.2 In important respects the critical realist framework is similar tothose critiqued by Holmwood. It is preoccupied with general questions about thenature of social structure and the character of human agency and intentionality.Like other general theorists, critical realists also place a great deal of faith in thepower of philosophical argument to establish the appropriate course for socialscientific research. Their account of the ontology of the social world elaborates aconceptual framework which, many realists argue, provides appropriate prin-ciples by which to regulate research. This framework is also presented, as in othergeneral theories, as a solution to the problems of an ‘angst-ridden’ social science,providing a basis for unifying research and moving beyond the present state ofdisputation (see Bhaskar, 1998: 16). It is true that in some respects critical realismappears to be different to other general theories, insofar as it rests on a seriousphilosophical account of natural science and calls on a distinctive mode ofargument, the transcendental deduction. However, I will be arguing that thesedifferences do not exempt it from the critique of general theory developed byHolmwood.

My critique of critical realism’s reliance on philosophical argument proceedsas follows. I suggest that critical realist ontological arguments are in many waysconvincing in the natural sciences, but that their persuasiveness is dependent onthe empirical success of the scientific arguments they are derived from. I thenturn to social science, arguing that critical realist ontology in this area is not sosoundly based, as it is not derived from an analysis of existing successful inquiry.Instead, critical realists attempt to derive an ontological framework from certaincommon-sense assumptions about the social world, a move I see as bothprocedurally flawed and practically unable to produce a coherent set of concepts.In the following section, I move on to consider Margaret Archer’s claim that theontological theories put forward by critical realists should be used to regulateresearch in the social sciences. This is of doubtful value both because of thelimited evidence for critical realist arguments in the social sciences, and becausethe idea of regulation suggests an unjustified foreclosure of research options. I

Kemp Critical Realism and Philosophy 1 7 3

04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 173



illustrate these points by contrasting an ontologically driven analysis of the under-class and unemployment, that put forward by Justin Cruickshank (2003), witha substantively driven approach to the same area, that put forward by RobertMacDonald (1994), arguing for the superiority of the latter. I then conclude byarguing that critical realism fails to demonstrate that philosophy can lead the wayin establishing ontological claims to be used as a regulatory framework forresearch. These matters cannot be resolved ahead of successful empirical research,and, therefore, a focus on philosophical ontology is not the way to resolve theangst of the social sciences.

Ontology in Natural Science

As is well known, critical realist analysis of natural science focuses on the issue ofontology, analyzing the most basic characteristics of the entities studied bynatural science. In contrast to logical positivists, critical realists argue that onto-logical claims can be rationally justified. In order to do so, Bhaskar appropriatesa Kantian mode of argument, the transcendental deduction, employing it in anovel way (Bhaskar, 1997). Whereas Kant used transcendental argument in orderto establish the conceptual preconditions for any act of knowing, Bhaskar uses itto identify the ontological preconditions required for the activity under analysisto be successful (1986: 11–12). His transcendental argument has two stages. Inthe first stage, the critical realist chooses a practice of inquiry that is generallyaccepted to be epistemically successful3 and identifies features of this inquiry thatare agreed upon in a range of different accounts. In the second stage, the realist4

performs a transcendental deduction in order to discover the ontological precon-ditions that must be present for a form of inquiry with these features to besuccessful. Such transcendental deductions are argued to be a reliable means toestablish ontological claims in a particular domain.

The most persuasive example of this procedure can be found in Bhaskar’sanalysis of fundamental physics and chemistry, developed initially in A RealistTheory of Science (1997) but also discussed elsewhere (see Bhaskar, 1989, 1998).In the first stage of the transcendental deduction Bhaskar starts from the positionthat fundamental physics and chemistry are epistemically successful forms ofscientific activity which have the experimental investigation of the natural worldas a key feature. Characterizing experiments, Bhaskar argues that they consist ofan intervention in the pattern of events in the world. In such interventions,experimenters manipulate the conditions and control the surroundings so thatthey can exclude potentially interfering influences and produce a pattern ofevents that would otherwise not have occurred (Bhaskar, 1989: 15).

Proceeding to the second stage of his deduction, Bhaskar argues that ananalysis of the preconditions that must be in place for experiments to be success-ful reveals substantial conclusions about the ontology of the world that is studiedin fundamental physics and chemistry. He suggests that the very idea that controlis needed in an experiment to remove interfering factors or influences shows


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 174



that some concept of ontological depth is implicit in the activity. That is, in orderfor such experimental activity to be possible, it must be the case that there is anontological distinction between events and influences on events. If no suchdistinction existed, then there would be no need to attempt to control orexclude influences that the scientists were not interested in investigating at thatpoint. These arguments are used, then, to generate one of Bhaskar’s ontologicalclaims – that there are ontologically distinct levels, one of which is the actual(the level of events) and another of which is the real (the level of influences)(Bhaskar, 1997: 33–6). He goes on to develop an account of the real, arguingthat structures (or mechanisms) are the key constituents here. A structure is anenduring feature of the natural world that has a characteristic way of actingwhich influences events in the world when triggered (Bhaskar, 1997: 49–50).It exists even when inactive, and retains its power to influence events even whenthis power is not being exercised. Summarizing, we might then say that insofaras experimentation contributes to epistemically successfully science, a precondi-tion of this success is the ontological distinction between real structures andevents.

Bhaskar certainly succeeds in extracting convincing ontological conclusionsfrom the analysis of scientific activity. However, in considering the relationshipbetween ontological argument and substantive scientific theorizing, I want toemphasize the ways in which the validity of those conclusions is dependenton pre-existing substantive theorizing. One aspect of this dependency can beillustrated by showing that the validity of ontological concepts is limited to thedomain of the successful theories to which they apply. Bhaskar himself made thisdomain limitation clearer in his later work, arguing that the framework devel-oped within A Realist Theory of Science dealt only with ‘fundamental’ physics andchemistry, rather than all forms of investigation in the natural sciences (Bhaskar,1989: 183). In the second edition of that work, Bhaskar makes the point thattranscendental deductions rely on particular scientific practices as their premises(1997: 260). As he puts it, the result of a transcendental deduction is ‘domainspecific’ (1986: 12) and transcendental deductions will have to be conducted forevery specific science (1997: 260). This being the case, we cannot assume that aframework developed to account for fundamental physics and chemistry willprovide a convincing account of the ontology of other areas, such as biology. Thisemphasizes that convincing ontological theorizing follows on from successfulresearch in a domain, rather than leading the way. In a domain in which there isno research that is agreed to be successful, the transcendental deduction lacks itsfirst premise.

Another way to bring out the dependency of ontological argument on success-ful substantive theorizing is to explore the fallibility of ontological claims.Although some of Bhaskar’s early remarks imply that he sees himself as establish-ing through transcendental analysis the ontological characteristics that the worldmust have, his later remarks clarify that he does see critical realist analyses aspotentially fallible. This was acknowledged partly in response to critiques putforward by Alan Chalmers (1988) and Ted Benton (1981) which argued that the


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 175



validity of transcendental deductions of the kind used by Bhaskar depends on thevalidity of the scientific theories on which they are based. Chalmers puts the pointnicely by imagining the conclusions that a medieval scientist might have reachedhad he pursued a realist transcendental analysis of some version of Aristoteliantheory (Chalmers, 1988). Chalmers suggests that the likely ontologicalconclusion would have been that ‘the world must be a finite harmonious wholewith a centre’, because this would have been seen as the necessary preconditionfor the successful employment of distinctions in substantive Aristotelian sciencesuch as those between forced and natural motions (Chalmers, 1988: 19). Atranscendental deduction from Aristotelian science would thus have producederroneous ontological conclusions because of its derivation from mistakensubstantive conceptions. This suggests, then, that a key contributor to thefallibility of ontological theories is the fallibility of the substantive theories onwhich they are based.5

Drawing together these arguments, I would conclude that they illustrate thedependency of ontological argument on pre-existing scientific research, both tosupply the initial premises for transcendental deductions, and to supply validtheories from which valid ontological claims can be derived. Ontology is notsetting the agenda and leading empirical research in the natural sciences, but isfollowing behind this research, and requires correction when substantive theoriesare corrected.

Ontology in Social Science

Despite the many interesting aspects of critical realist analysis in the naturalsciences, there has been, to my knowledge, little uptake of this approach amongphilosophers of natural science, or natural scientists themselves. However, thesituation is different in social science. In this area, critical realism has a numberof prominent advocates including Bhaskar (1979), Archer (1995) and Sayer(1992), and there is growing interest in the doctrine among empirical researchers.What I would like to do in this section is consider whether realists can call onphilosophical analysis to establish the ontology of the social world.

When we compare critical realist analyses of natural and social science, thereis an obvious and striking difference. In many natural sciences, critical realistscan start from scientific theories that are generally (if not necessarily universally6)agreed to be successful, and develop their ontological arguments from thisstarting point. However, as Bhaskar points out in The Possibility of Naturalism(1998), there seems at present to be no line of social scientific inquiry which hasthe high level of agreement about its epistemic success found in many areas ofthe natural sciences (1998: 14).7 Therefore, it is not possible for realists tostraightforwardly apply the procedures used in the analysis of physics andchemistry to the social sciences. If one were to ask of a particular form of socialscientific inquiry the question, ‘given that this form of inquiry is successful,what must the social world be like?’, the contested status of the claim to social


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 176



scientific success would undermine the persuasiveness of the resulting ontologicalargument. What procedure, then, is open to realists?

Bhaskar puts forward two quite different responses to this question at differenttimes, one of which suggests strong limits on the potential for realism in this area,while the other is more bold in its scope. The first, more cautious, position canbe found in his postscript to A Realist Theory of Science (which first appeared inthe edition published in 1978). The guiding principle here is his remark that‘philosophy cannot anticipate the form of a successful scientific practice’(Bhaskar, 1997: 260). In other words, transcendental deductions are only validonce they have a successful form of scientific inquiry to start from. On this view,it would seem that a realist transcendental deduction from social scientificpractice is not possible. At best, realism can offer a negative contribution tounderstanding the ontology of the social world, counselling social scientists withan enthusiasm for natural science not to adopt a positivist ontology, because suchan ontology is not consistent with actual research practice in the natural sciences(Bhaskar, 1997: 260–1).

In The Possibility of Naturalism (first published in 1979), on the other hand,Bhaskar sets aside this caution, and offers a much bolder programme for criticalrealism in social science. In this book, he suggests that philosophical argumentcan actually establish the ontological properties of societies prior to the achieve-ment of successful empirical social scientific theorizing. This confidence in therole of philosophy is adopted by many other critical realists, often withoutextensive reflection on its validity (see, for example, Archer, 1995).8 The basis ofBhaskar’s revised, bold approach is the idea that a transcendental deduction neednot start from a successful investigation into the social world. Rather, it can beginfrom ‘more or less universally recognized features of substantive social life’(Bhaskar, 1998: 14) and work from these to establish the ontological features ofthe social world. It is this procedure that Bhaskar hopes will allow him to answerthe question: ‘What properties do societies possess that might make thempossible objects of knowledge for us?’ (1998: 25).9

For Bhaskar, the feature of social life that is universally recognized is the exist-ence of intentional agency, and his analysis begins with this and moves to showits preconditions (1998: 173). According to Bhaskar, intentional agency cannotoperate on its own. After all, it produces acts, speech and objects, and these donot appear out of nothing. Rather, agency must draw on social structures in orderto produce its outcomes. He states:

Thus consider saying, making and doing as characteristic modalities of human agency.People cannot communicate except by utilising existing media, produce except byapplying themselves to materials which are already formed, or act save in some or othercontext. Speech requires language; making materials; action conditions; agencyresources; activity rules. (1998: 34, original emphasis)

According to Bhaskar, then, transcendental analysis reveals that social structuresare an ontological precondition for intentional agency. Notice here that he is notattempting to reduce intentional agency to social structure, but to show that both


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 177



social structure and agency must exist, and, correlatively, both society and indi-viduals.

His insistence on the need to acknowledge social structures raises the issue ofhow they are to be characterized. According to Bhaskar, social structures arefundamentally relational in character. Social science should thus be concernedwith ‘the persistent relations between individuals (and groups), and with therelations between these relations (and between such relations and nature and theproducts of such relations)’ (1998: 28–9, original emphasis).

Although he is not entirely clear on this (see Bhaskar, 1998: 42–4), mostcritical realists take it that the relations which constitute the most fundamentalbuilding blocks of society, and are thus persistent, are internal and necessary ones.An internal and necessary relation is one in which the individuals (or groups)involved are intrinsically linked to one another, such that ‘the existence of onenecessarily presupposes the other’ (Sayer, 1992: 89). For example, within Marx’stheorizing, the relation between workers and bourgeois factory owners is aninternal and necessary one. Such relations, particularly internal and necessaryrelations, are the essence of sociality and the prime focus of social explanations(Bhaskar, 1998: 41).

Lack of space precludes me from elaborating Bhaskar’s ontology of the socialworld any further. Instead, I would like to consider whether Bhaskar is success-ful in his attempt to establish the ontology of the social world through philo-sophical argument. I would argue that realist claims about the ontology of thesocial world do not have as strong an epistemic warrant as the ontological claimsrealists derive from analyzing fundamental physics and chemistry. The latter arebased on a successful form of scientific practice, and gain support from theirability to account for the preconditions that make this success possible. Realistanalysis is particularly powerful in these areas because of its focus on a quitespecific aspect of inquiry, that is, experiment, so that it can move to show whatmust be the case for experiment to be a successful and necessary part of inquiry.Bhaskar’s choice of ‘intentional agency’ as a starting point for a realist analysis ofthe ontology of the social world does not give rise to a similarly persuasive analysisfor two reasons.

First, as Benton points out in his critique of The Possibility of Naturalism, it isnot clear that ‘intentional agency’ is a neutral and uncontested starting point foranalysis (1981: 16). That is to say, as some approaches to social science do notaccept that intentional agency is a central feature of the social world, it cannotbe taken as a generally agreed upon premise for a transcendental deduction.Oddly, however, Benton also argues that this starting point is particularly con-genial to Durkheimian and Marxist approaches, and will skew any ontologicalconclusions in the direction of these approaches. This rather lets Bhaskar off thehook, allowing him to respond that his starting point is actually more favourableto anti-naturalist approaches that focus on agency and meaning rather thanDurkheimian and Marxist approaches which have tended to have a more struc-tural focus (Bhaskar, 1998: 173). While Bhaskar identifies the bias of his startingpoint more accurately than Benton, the acknowledgement of this bias reinforces


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 178



the initial difficulty: that intentional agency is not a neutral, generally agreedupon feature of social life. Indeed, I would suggest that its existence is denied ina variety of research programmes including certain Marxist, Durkheimian, struc-turalist and post-structuralist approaches.10 Because intentional agency is notaccepted by a range of social scientific approaches, a transcendental deductionthat begins from it cannot have the same force as a transcendental deduction fromthe widely accepted premise that fundamental physics and chemistry are success-ful sciences that rely on practices of experimentation.

Second, even if we were to grant Bhaskar the legitimacy of his starting point,it is by no means clear that all, or indeed most, of the realist ontology is in somesense derived from or entailed by it. For example, the claim that social structuresare relational appears, in Bhaskar’s account, to be argued for quite independentlyof his starting point in intentional agency, emerging, rather, from a critique ofother ontologies (Bhaskar, 1998: 28–30). This disconnection becomes even moreapparent as realists such as Archer expand further the number of concepts calledupon to describe the ontology of the social world (see Archer, 1995). For manyrealist ontological concepts, then, there can be no genuine claim that transcen-dental argument from a generally agreed upon starting point is underpinningthem. As such, these claims should not be taken to be underwritten by atranscendental deduction.

Now, it could be argued that while these criticisms highlight proceduralproblems with the justification of the realist ontology, they do not demonstratethat the ontology is, in itself, problematic or misguided. Those who have greaterfaith in the powers of philosophical argument than I do may claim that suchargument has, in fact, produced a coherent account of the basic features of thesocial world. While I do not have the space to offer a detailed critique of realistontological arguments here, I would like to counter this claim by highlighting arepresentative problem with the coherence of realist ontological categories.Although Bhaskar’s arguments about the relation of ontological argument toresearch have not been improved upon by those of later realists, his ontology ofthe social world has been superseded by that put forward by Margaret Archer inRealist Social Theory: A Morphogenetic Approach (1995). Archer revises andelaborates upon Bhaskar’s framework in two ways. First, she argues that he doesnot make a clear enough separation between the emergent powers of structuresand those of agents. Second, Archer proposes that an additional ontologicalelement, cultural emergent properties, should be incorporated into analysis.These are differentiated from structure because the powers of structures emanatefrom their ‘primary dependence upon material resources’, whereas the powers ofculture emanate from the constraining force of the logical relations betweenbeliefs (Archer, 1995: 175, original emphasis).

Despite its attempts to improve upon Bhaskar’s ontology, Archer’s accountruns into at least one important difficulty on its own terms. Archer’s ontologyrelies upon there being a strict division between culture and structure, but herdefinitions of these concepts do not allow us to clearly identify which entities fallinto each category. This becomes apparent, for example, when Archer discusses


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 179



one important type of social structure, the ‘role’ (1995: 186–9). Within Archer’sframework, it seems correct to state that social roles involve structural elements;for example, fulfilling the role of office clerk requires access to resources such asfiles, computers, and so on. But this does not demonstrate that a role should beunderstood as primarily structural in character. After all, Archer elsewhereacknowledges that a role has necessary and internal connections with rules as tohow that role is to be undertaken, which in her terms must be cultural (1995:275–6). Because of this cultural aspect, roles with similar structural elements mayhave quite different characteristics. For example, managers will call on the samekinds of structural elements in their roles as office clerks, e.g. computers, filingsystems, and so on, but this does not mean the roles are fundamentally the sameas one another; indeed, they are quite different.

One possible response to this would be to reclassify roles as cultural in Archer’ssense, preserving her basic ontological distinctions, but reassigning some of theless fundamental features. However, her account of roles makes it clear thatstructural elements are of some relevance in grasping the character of a role, soto define a role as purely cultural would be just as unsatisfactory as defining it aspurely structural. Although Archer claims to have established a clear-cut distinc-tion between structure and culture, the fact that entities like roles cannot bestraightforwardly placed into either category throws her division into question.11

I would like to suggest that ontological arguments such as those put forwardby Archer and Bhaskar tend to be troubled by numerous problems of this kind.12

For my purposes here, however, such problems are intended to serve as a second-ary illustration of the difficulties arising from a focus on ontology. My morefundamental concern is whether a convincing rationale is provided for the claimthat philosophical argument can establish the ontology of the social world.Within the realist approach, it seems to me that Bhaskar’s argument is still theclassic defence of this position. To my mind, this defence is inadequate, asBhaskar fails to provide an effective surrogate for the successful scientific theor-izing that provides the basis of a critical realist analysis in natural science. I wouldargue, then, that Bhaskar’s initial, cautious assessment of the possibility of estab-lishing an ontology of the social world was actually correct. In the absence ofsuccessful empirical theorizing from which to derive ontological conclusions,there is little justification for claims about the ontology of the social world.Convincing ontological arguments will only emerge (provisionally) after theestablishment of (provisionally) successful theories.

Ontology and Research in the Social Sciences

Perhaps the main reason that realists feel that ontological issues are pressing, andworth debating, is because they feel that ontology can make a genuine contri-bution to research in the social sciences. It is to this issue that I would like nowto turn. In particular, I would like to consider the contention that ontologicalargument should regulate research, ruling in and out explanatory concepts on the


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 180



basis of their ontological presuppositions. Critical realists often see this as animportant feature of their approach, and I shall consider the version of thisargument put forward by Archer.13 Discussing the role of ontology in bothcritical realism and other approaches, Archer claims that social ontologies ‘governthose concepts which are deemed admissible in explanation as in description’ (1995:20, original emphasis). She goes on to state: ‘Because the ontology containsjudgements about the ‘ultimate constituents’ (and non-constituents) of socialreality, it thus governs what sorts of concepts may properly be countenanced forany purpose whatsoever’ (1995: 21).

It is true that Archer also sees ontology as being reciprocally regulated by thatwhich is empirically discovered (although I am unaware of any work of herswhich revises any element of her expansive ontology based on the outcome ofresearch). However, I would argue that the general thrust of Archer’s approach isto suggest that ontological arguments set the terms for (‘govern’) empiricalresearch rather than vice versa. This reading is backed up by the priority she givesto the elaboration of ontological arguments, undertaken as a separate activity toempirical research. It is also reinforced by Archer’s suggestion that in order toconduct social research, one must first settle ontological questions. In a discussionabout the legitimacy of permitting different ontological orientations Archerargues the following:

. . . whilst agreeing that both Harré and Bhaskar are both [sic] attempting to get atthe ‘fundamental generative structures and generative mechanisms of social life’, sincethe former entertains only mentalistic contenders and the latter does not, then one ofthem has to be (fundamentally) wrong. As a realist sociologist, I have to judge whichin order to advance any concrete explanatory proposition. (1998: 199)

I would suggest that the regulatory role given to ontology here can be critiquedfrom three angles. In the first place, it seems to me that much successful naturalscientific research has been conducted without the explicit regulation of explana-tory concepts. I acknowledge that this is a broad claim that can only be stronglyjustified through empirical evidence. We can have, however, greater confidencein the more restricted claim that natural science achieved great success withoutontological regulation derived from critical realist concepts. It is not clear, then,why we should see this kind of regulation as necessary in the social sciences.Perhaps the idea that there is something very special about the subject matter ofthe social sciences leads realists to believe that ontological clarification must laythe groundwork for empirical research.14 Even if there is something ontologicallydistinctive about the social world, however, this does not entail that abstractphilosophical argument can establish what it is prior to empirical research.

Second, Archer’s demand that ontological argument settles certain mattersprior to research seems to unnecessarily foreclose upon the options forresearchers. Surely a researcher with a genuinely investigative spirit, with genuinecuriosity, might actually entertain both of the ontological possibilities shementions (as well as others) and see which allowed the best empirical account ofthe subject matter in question to be produced? If we are to conceive of social


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 181



scientific investigation as involving exploration and discovery, rather than fillingin the details of preset frameworks, we must allow researchers some creativityrather than restricting their options using philosophical argument (for furtherdiscussion, see Holmwood, 1996). Of course, investigation is not entirely fluid,and past successes provide precedents for how to achieve success in the futurewhich it would be foolish for researchers to ignore. Nevertheless, it is intrinsic tothe possibility of progress in investigation that past successes may need to be over-turned. Even where ontological claims have a strong basis in empirically success-ful theories, they may be challenged by the development of newer, more powerfultheories, with different ontological presuppositions. Where they do not have thisbasis, there are fewer still grounds to allow ontology a regulative role.

The third point that I would like to make is that the idea of regulation byontology implies that researchers should begin by making decisions about theirontological beliefs before moving on to substantive theorizing. This is related tothe assumption, present in much realist work, that, as all theories have ontologicalpresuppositions, the only reasonable course is to make these presuppositionsexplicit and engage in philosophical debate about them. Any other course wouldinvolve ‘head in the sand’-style avoidance of the issue. I see this assumption asmistaken. Even if we agree that all theories contain some ontological presuppo-sitions, this does not entail that it is therefore productive to extract these anddebate their validity at a philosophical level. Indeed, as the key supportingevidence in favour of an ontology is its derivation from an empirical theory thatis widely held to be successful, until such theories are generated in social sciencethere is little to be gained from engaging in ontological debate. Further, as I havesuggested above, it is not the case that the ontological arguments made by Archerhave resulted in a consistent set of ontological categories. As I argued in relationto the notion of ‘roles’, it may, in fact, be hard to place the entities being analysedin the research consistently into one ontological category or another. This under-mines the idea that the use of such categories is adding to our understanding ofthe domain being researched.

As such, I would argue that researchers need not begin from ontological reflec-tion before engaging in research. Even if the theories that they are developinghave ontological assumptions, the point of these theories is to achieve substan-tive success, and this should be the focus of theorizing. Given the lack of justifi-cation for ontological arguments in the social sciences, and the internal problemspresent in ontological claims, the insistence that researchers engage in extensiveontological reflection before undertaking their research is unwarranted.

In order to illustrate these points about the problems with focusing on onto-logical argument instead of substantive research questions, I would like tocontrast two discussions of unemployment and the underclass, one of which issubstantively oriented and the other of which is ontologically oriented.

The substantively oriented discussion is Robert MacDonald’s article ‘FiddlyJobs, Undeclared Working and the Something for Nothing Society’ (1994).MacDonald’s study of the activities and views of individuals who do ‘fiddly jobs’,i.e. engage in undeclared working while claiming benefits, is motivated by the


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 182



desire to offer evidence for or against various claims about such individuals,including the idea that they are part of a ‘dependency culture’ and that theirparticipation in welfare fraud shows that they are lacking moral principles.Bringing together data from a large number of interviews, MacDonald reachessome interesting and surprising conclusions, of which I shall mention just twohere. First, those involved in fiddly work shared a fairly clear sense of the kindsof benefit fraud that were legitimate and those that were illegitimate. Combin-ing undeclared work with benefit claims was understood to be morally justifiablewhere the work was short-term, motivated by family need and done for a smallamount of cash to top up benefit payments. By contrast, combining a full-time,properly waged job with claiming the benefit was understood to be immoral, andan abuse of the system. Thus, participation in minor benefit fraud was not anindication of an amoral orientation on the part of individuals, but rather of amoral orientation towards meeting the needs of the family (MacDonald, 1994:519–20). Second, MacDonald argues that those formally unemployed indi-viduals who engaged in fiddly work were not consigning themselves to partici-pation in a ‘dependency culture’. Instead, he suggests that such individuals werewell placed when full-time, legitimate employment did become available, becauseby engaging in fiddly work they avoided social isolation and did not becomeresigned to their status as unemployed (1994: 526).

As a counterpoint to MacDonald’s approach we can consider Justin Cruick-shank’s discussion of the underclass, entitled ‘Underlabouring and Unemploy-ment: Notes For Developing a Critical Realist Approach to the Agency of theChronically Unemployed’ (2003). In this piece, Cruickshank does not engage inresearch into unemployment and the underclass, but instead argues that researchin this area should be conducted by starting from realist ontological concepts,and then working towards the development of a ‘domain-specific meta-theory’which calls on these concepts but fills them out by engaging with the empiricalparticularities of the area of research. In order to do so, Cruickshank calls onArcher’s elaborated realist social ontology of cultural and structural emergentproperties and he attempts to show which of these categories each empirical itemfits into. For example, he identifies the moral views of the fiddly job seekers as aculturally emergent property, one that operates in the wider context of the struc-turally emergent property of neo-liberalism and de-industrialization (Cruick-shank, 2003: 122–4). Cruickshank also uses realist ontological principles tocritique other approaches to unemployment and the underclass, including thatof MacDonald. According to Cruickshank, MacDonald’s work ‘adheres to animplicit methodological individualism’, in that MacDonald focuses on the beliefsand actions of individuals, implying that social reality is simply the sum of these(Cruickshank, 1994: 121). Cruickshank suggests that MacDonald, therefore,fails to include or analyze contextual factors that shape the situation of indi-viduals.

What might then be said about the comparative merits of these twoapproaches to unemployment and the underclass? To my mind, MacDonald’ssubstantively oriented approach is the more insightful and valuable. His work fills


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 183



in gaps in our knowledge about those who work undeclared on benefits. It alsoprovides evidence which challenges one politically dominant understanding ofsuch individuals, suggesting that these individuals are not lacking in moralbeliefs, but have particular, intelligible ideas about which kinds of behaviour arelegitimate and illegitimate. Thus, MacDonald’s work pushes forward socialscientific understandings of unemployment and contributes to wider politicaldebates. In comparison, the contribution made by Cruickshank’s approach is, Iwould argue, rather limited.

In the first place, Cruickshank’s critique of MacDonald’s approach asimplicitly invoking methodological individualism is quite misguided, asMacDonald makes a number of references to the wider social context whichshapes fiddly job-seeking.15 Even if MacDonald’s ontological presuppositions hadbeen correctly identified, however, I would argue that focusing on these distractsattention away from the more fundamental issues of whether the substantiveclaims are justified, and whether they are illuminating about their subject matter.Unless ontological arguments can be demonstrated to have a ‘pay off ’ in relationto substantive understanding, they are largely irrelevant to research.

Perhaps more importantly, I would argue that Cruickshank’s reframing ofempirical contributions such as MacDonald’s into realist terminology addslittle or nothing to our understanding of unemployment and the underclass.He simply takes these contributions out of their context and places them intoan ontological framework that has not been justified empirically and has itsown difficulties. As I mentioned above, Cruickshank calls on Archer’s distinc-tion between cultural and structural emergent properties in order to classifyvarious elements of the unemployment debate, suggesting that neo-liberalismand de-industrialism can be understood as structural emergent properties andpatriarchy can be understood as a cultural emergent property (Cruickshank,2003: 122–4). One initial puzzle here is why neo-liberalism is characterized as astructural emergent property and patriarchy as a cultural one. Does patriarchyhave nothing to do with the issues of resource distribution and utilization thatArcher, and, following her, Cruickshank classify as structural?16 Similar problemsemerge in the placement of ‘de-industrializing capitalism’ as structural rather thancultural. Capitalism might appear to be primarily linked to material resources,but it is surely the case that an important role in the transformation of capitalismhas been played by complexes of belief that Cruickshank would identify ascultural. One example of these beliefs would be ideas about why old-style indus-tries are no longer productive and profitable within a British context. We can alsonote that items which Cruickshank would classify as cultural are also a funda-mental aspect of production processes themselves, both in ‘old-fashioned’ capi-talism and ‘knowledge-driven’ capitalism; neither Model-Ts nor microchips canbe produced without conceptualized procedures and skills. Thus, we can see thatthe problems with Archer’s definitions of culture and structure feed through intoCruickshank’s use of these distinctions, giving rise to incoherencies in his classifi-cation of items in his research domain.17 Summarizing the issue, I would arguethat MacDonald’s focus on the substantive is more productive than Cruickshank’s


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 184



focus on the ontological. Whereas MacDonald’s approach reveals to us featuresof the social world that were previously unknown to social scientists, Cruick-shank’s approach translates existing claims into an empirically unjustified philo-sophical framework that contains contradictions of its own.18

Having critiqued the idea that realist ontological arguments should be used toregulate research, I should emphasize that this does not mean that realist onto-logical ideas are completely irrelevant to research. Rather, it means that theyshould be taken not as legislating the framework within which research must beconducted, but instead taken as possible inspiration for ways of thinking aboutresearch questions, which may turn out to be more or less productive. It couldbe argued, for example, that the realist invocation of the notion of social struc-ture makes a potentially useful parallel with successful explanations in physicsand chemistry.19 On the other hand, it may be that, for example, it will be moreproductive for social scientists to take inspiration from the ontological andexplanatory theories of evolutionary biology rather than physics and chemistry.Whatever the case, the most productive uses of realist ideas will, in my view, bethose that are most closely linked to actual research, and focus strongly on theproduction of valid explanations. For example, Ray Pawson’s work on middle-range realism is inspired by critical realism, but Pawson does not insist thatinvestigators must adopt an elaborate ontological framework in order to conducttheir research (Pawson, 2000). Rather, Pawson takes a few simple realist conceptsand argues in detail that two particular pieces of research would have producedstronger explanations if they had been thought through using those concepts.Such an approach is valuable in that (i) it does not argue for realist concepts philo-sophically but attempts to show their value by showing how their deploymentresults in more successful empirical explanations; and (ii) it does not attempt toimpose on researchers a wide-ranging conceptual structure, but picks out a fewuseful guiding principles. These modest uses of realism are preferable to attemptsto restrict and regulate research that are not justified by a demonstration of theirempirical success. Crucially, it is the empirical success of the theory which willgive support or otherwise to the ontological claims that stimulated it, rather thanphilosophical argument disconnected from this success.

Conclusion

To conclude this article, I would like to emphasize the links between the moredetailed arguments I have made about critical realism and the broader issues thatI mentioned at the beginning of the article. In the first place, there is the questionof the powers and limitations of philosophical argument. My analysis of realistarguments about the ontology of the natural and social worlds supports the claimthat there are strong limits on the ability of philosophical argument to run aheadof substantive investigation in these areas. Certainly, critical realism provided avery useful clarification of the ontological assumptions present in successfulinquiries within fundamental physics and chemistry. But this analysis is


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 185



dependent on the successful substantive inquiries in question, and its claimscould be revised or overthrown as substantive inquiry in this area, or other areas,develops. In relation to the ontology of the social world, critical realists such asBhaskar attempt to derive strong ontological claims from beliefs about featuresof the social world that are purported to be generally held. This, I have suggested,makes the conclusions of these arguments rather less persuasive than thosederived from successful natural scientific theorizing. In neither the natural orsocial scientific case do the philosophical arguments of critical realism take thelead, in the sense of convincingly establishing claims about some domain priorto research in that domain.

In the second place, there is the issue of what general social theory can do forthe social sciences. Proponents of general social theory often wish to movebeyond the disunity and disagreement prevalent in the social sciences by estab-lishing a single shared framework in which all research should be carried out(Holmwood, 1996). I suggested above that this is true of critical realism, withBhaskar claiming that ontological argument will help us to move beyond theangst-ridden state of the social sciences, and establish a single appropriate frame-work for studying the social world. The idea that critical realism establishes anontological theory which can then be called on to regulate the explanatoryconcepts used in research, promoted by Archer, seems consonant with this goalof unification. However, I argued that such regulation is problematic insofar asit calls on weakly justified philosophical arguments in order to restrict theinvestigative scope of empirical researchers. Even where ontological argumentshave a stronger basis, as in fundamental physics and chemistry, they should notbe used to rule out explanatory concepts inconsistent with them. Rather, thepriority is with successful explanations, which show us which ontological theoriesare viable, rather than vice versa. Given the weak justification for ontologicalarguments in the social sciences, any regulation in this area is of extremelydoubtful value. The paradox of critical realism can then be formulated as follows.Critical realists see ontological arguments as the route to success and unificationin as yet unsuccessful areas of research. Yet a convincing ontology can only be(fallibly) established once research in an area is successful, and there is a degreeof unity in acknowledging this success.20 Thus, although they may make a moremodest contribution, critical realist arguments are no more successful at estab-lishing the centrality of their framework to the social sciences than are thearguments put forward by other general social theorists.

Following Holmwood, then, I would suggest that general social theory cannotprovide social science with a route to success that will bring about a degree ofunity similar to that of the successful natural sciences.21 If success is going to beachieved, it will be through detailed attention to the explanatory problems thatarise from engaging with the subject matter of research. It will have to becompelling explanations that generate agreement among practitioners, ratherthan philosophical argument. Obviously, explanatory work conducted up untilthe present does not seem to have generated extensive agreement among prac-titioners, and there may be ways in which the development of more successful


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 186



explanations could be promoted (such as institutional reform). But it will not bephilosophy disconnected from empirical research that leads the social sciencestowards success.

Acknowledgements

I would like to thank the British Academy for their support during the completion of thisarticle through their Post-Doctoral Fellowship scheme. Thanks also to Ipek Demir, JohnHolmwood, and Gregor McLennan for their comments on this article, and to SharaniOsborn for her editorial work.

Notes

1 We might usefully distinguish Parsons, Alexander and Giddens from Habermas inthat the former are primarily sociologists drawing on general philosophical arguments,whereas the latter is primarily a philosopher who, though often concerned with thesocial sciences, also has a wider range of interests. Despite the sociological groundingof Parsons, Alexander, and Giddens, however, none of them attempts to adduceempirical support for their general theoretical programmes. Parsons and Alexanderexplicitly characterize their arguments as independent of issues of empirical support(see Parsons, [1937] 1949: 733; Alexander, 1982: 114–15), and Giddens implicitlydoes so by failing to consider empirical evidence that might have a bearing on hisstructuration approach.

2 John Holmwood has engaged with Andrew Sayer’s work (see Holmwood, 2001) butI will be focusing largely on the arguments of Roy Bhaskar, as the pioneer of the realistapproach, and Margaret Archer, as an important recent exponent.

3 At various points in this article I refer to judgements of the epistemic and/or explana-tory success of theories, and readers might legitimately ask what this amounts to.Obviously, this is a large topic that I cannot seriously address here. Further, in linewith my general views on philosophy, I do not believe that there is a single criterionfor successful explanation that can be established by philosophy. However, existingtheories that are judged to be successful can stand as exemplars of the forms ofexplanatory success achieved so far within science, and their qualities can be identified.Arguably, these qualities include: the ability to coherently account for the relevantfactual claims in a domain; the ability to reduce one level of phenomena to another;and the ability to explain how entities interact with one another.

4 I use the term ‘realist’ here as shorthand for ‘critical realist’. Although there are otherspecies of realist, this article focuses entirely on critical realism.

5 Another dimension of the fallibility of ontological theories relates to problems in thereasoning by which ontological conclusions are reached.

6 The extent to which this epistemic agreement is found in relation to any particularnatural science may vary. As such, the limitations of the realist analyses of socialscience that I will indicate here may apply equally to the analysis of natural sciencesin which there is little agreement regarding their epistemic warrant.

7 This is clearly a very broad claim which Bhaskar does not back up with evidence,and I cannot offer any extended discussion of this here. Nevertheless, studies of


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 187



particular areas in social science, such as James Rule’s analysis of theories of civilviolence, suggest that research in these areas is often non-progressive, with laterapproaches being disconnected from earlier approaches rather than being demonstra-ble improvements upon them (Rule, 1988). Because of the absence of a positive,progressive dynamic, there is extensive disagreement about the epistemic success ofparticular approaches, rather than the wide-ranging (if not total) agreement about theepistemic virtues of particular theories found in many areas of the natural sciences.

8 What I mean by this is that critical realists such as Archer do not dwell on the legit-imacy or otherwise of establishing an ontological framework for the social scienceslargely through philosophical argument disconnected from empirical research.

9 It is worth noting that in The Possibility of Naturalism Bhaskar offers another, slightlyless bold description of his intentions, which is to establish ‘the properties thatsocieties and people must possess if they are to be (or demarcate the sites of ) possibleobjects of knowledge for us’ (1998: 167, emphasis added). This is less bold insofar asit suggests that Bhaskar is not establishing what the social world is actually like, merelywhat it must be like if it is to be known scientifically. Nevertheless, as I will be arguingin relation to Bhaskar’s boldest approach, this modified version expects too much fromphilosophical argument. It seems to me that empirical investigation of some area mayreveal new kinds of objects which can be known in new kinds of ways. Philosophycannot establish in advance of investigation what properties an entity must have inorder to be knowable.

10 This is not to say that all Marxist and Durkheimian approaches deny a role to inten-tional agency, just that this is characteristic of some versions of these lines of thought.

11 This discussion is intended to show that within her own terms Archer’s concepts areunsatisfactory. As such, when I suggest that certain elements of roles should be under-stood as ‘structural’ or ‘cultural’ on Archer’s terms, I am not committing myself tothis description of them.

12 For further discussions of the kinds of incoherence found in general social ontologies,see Holmwood and Stewart (1991) and Pleasants (1999).

13 Not all critical realists would necessarily approve of this regulative stance, however,one prominent objector being William Outhwaite (1987).

14 Archer, for example, opens Realist Social Theory with the remark ‘Social reality isunlike any other because of its human constitution’ (1995: 1). This idea of the noveltyof social reality is not confined to critical realism, animating nearly all forms of generalsocial theory from Parsons through Giddens to Habermas (for further discussion, seeHolmwood and Stewart, 1991).

15 MacDonald refers to various features of the social context including: the change inemployment practices towards emphasizing flexible employment which encouragescost-cutting contractors to pay for cheap, undeclared labour (1994: 527), the ways inwhich higher unemployment may actually decrease levels of undeclared work becauseof its effect on opportunities to arrange trades or pass on information about possiblework (1994: 518), and the forms of morality and ideology related to work and thefamily held by those on benefits (1994: 509).

16 This placement of patriarchy as cultural compared to the structural matter of capi-talist economics is paralleled in other realist arguments (see, for example, Sayer, 2000;for a critique, see Holmwood, 2001).

17 To criticize Cruickshank’s realist ontological arguments in this way is not to implythat the concepts used in MacDonald’s substantive research are entirely coherent andadequate. However, I would argue that the most productive way to deal with any


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 188



inadequacies that might be identified in MacDonald’s concepts is to attempt toreformulate them to produce a better account of the subject matter in question, i.e.the underclass and unemployment. Resolving the incoherencies will then improvesocial scientific understanding of the research domain, rather than being orientedtowards general philosophical problems.

18 Despite these difficulties with Cruickshank’s approach, I should note that he doesusefully argue that realist ontological claims may be challenged and thus transformedin the course of research (Cruickshank, 2003: 114–15). My dispute with hisarguments derives from his prioritization of ontology as the starting point for thinkingabout research despite the current absence of empirical evidence to justify realist onto-logical claims.

19 Unfortunately, the utility of the notion of social structure idea for empirical expla-nation is weakened by the insistence that the social world contains agency as well asstructure. This analytical combination tends, in general social theories, to promoteindeterminacy in explanations, rather than clear, testable empirical claims (seeHolmwood, 1996).

20 Bhaskar discusses this paradox in the first chapter of The Possibility of Naturalism, andsees himself as avoiding it by starting transcendental analysis in this area fromcommonly accepted features of social life. I have suggested that this does not offer away out of the difficulty.

21 We should certainly not over-estimate the unity of the successful natural sciences,implying that there is agreement on all matters. Indeed, as Lakatos (1978) hassuggested, a certain amount of disunity is probably explanatorily beneficial, not tomention unavoidable. Nevertheless, in successful sciences there seems to be a greaterdegree of agreement than there is in most areas of social science.

References

Alexander, Jeffrey (1982) Theoretical Logic in Sociology, vol. 1: Positivism, Presuppositions,and Current Controversies. London: Routledge and Kegan Paul.

—— (1985a) ‘Introduction’, in Jeffrey Alexander (ed.) Neo-Functionalism. London: Sage.—— (1985b) ‘Review Essay: Habermas’s New Critical Theory: Its Promise and

Problems’, American Journal of Sociology 91(2): 400–24.Archer, Margaret (1995) Realist Social Theory: The Morphogenetic Approach. Cambridge:

Cambridge University Press.—— (1998) ‘Introduction: Realism in the Social Sciences’, in M.S. Archer, R. Bhaskar,

A. Collier, T. Lawson and A. Norrie (eds) Critical Realism: Essential Readings. London:Routledge.

Benton, Ted (1981) ‘Realism and Social Science: Some Comments on Roy Bhaskar’s “ThePossibility of Naturalism” ’, Radical Philosophy 27: 13–21.

Bhaskar, Roy (1979) The Possibility of Naturalism: A Philosophical Critique of the Contem-porary Human Sciences, 1st edn. Brighton: Harvester Press.

—— (1986) Scientific Realism and Human Emancipation. London: Verso.—— (1989) Reclaiming Reality: A Critical Introduction to Contemporary Philosophy.

London: Verso.—— (1997) A Realist Theory of Science, 2nd edn. London: Verso.—— (1998) The Possibility of Naturalism: A Philosophical Critique of the Contemporary

Human Sciences, 3rd edn. London: Routledge.


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 189



Chalmers, Alan (1988) ‘Is Bhaskar’s Realism Realistic?’, Radical Philosophy 49: 18–23.Cruickshank, Justin (2003) ‘Underlabouring and Unemployment: Notes for Developing

a Critical Realist Approach to the Agency of the Chronically Unemployed’, in J.Cruickshank (ed.) Critical Realism: The Difference It Makes. London: Routledge.

Feyerabend, Paul ([1975] 1978) Against Method: Outline of an Anarchistic Theory ofKnowledge. London: Verso.

Giddens, Anthony (1977) ‘Functionalism: Après la Lutte’, in Studies in Social and PoliticalTheory. London: Hutchison.

—— (1979) Central Problems in Social Theory: Action, Structure and Contradiction inSocial Analysis. London: Macmillan Press.

Habermas, Jürgen (1984) The Theory of Communicative Action, vol. 1: Reason and theRationalization of Society, trans. Thomas McCarthy. Boston: Beacon Press.

Holmwood, John (1996) Founding Sociology? Talcott Parsons and the Idea of GeneralTheory. London: Longman.

—— (2001) ‘Gender and Critical Realism: A Critique of Sayer’, Sociology 35(4):947–65.

Holmwood, John and Stewart, Alexander (1991) Explanation and Social Theory. London:Macmillan.

Lakatos, Imre (1978) The Methodology of Scientific Research Programmes: PhilosophicalPapers, vol. 1, ed. J. Worrall and G. Currie. Cambridge: Cambridge University Press.

MacDonald, Robert (1994) ‘Fiddly Jobs, Undeclared Working and the Something forNothing Society’, Work, Employment and Society 8(4): 507–30.

Outhwaite, William (1987) New Philosophies of Social Science: Realism, Hermeneutics andCritical Theory. London: Macmillan.

Parsons, Talcott ([1937] 1949) The Structure of Social Action: A Study in Social Theory withSpecial Reference to a Group of Recent European Writers. New York: Free Press.

—— (1951) The Social System. London: Routledge and Kegan Paul.Pawson, Ray (2000) ‘Middle-Range Realism’, Archive of European Sociology 41(2):

283–325.Pleasants, Nigel (1999) Wittgenstein and the Idea of a Critical Social Theory: A Critique of

Giddens, Habermas and Bhaskar. London: Routledge.Rorty, Richard (1980) Philosophy and the Mirror of Nature. Oxford: Blackwell.Rule, James (1988) Theories of Civil Violence. Berkeley: University of California Press.Sayer, Andrew (1992) Method in Social Science: A Realist Approach, 2nd edn. London:

Routledge.—— (2000) ‘System, Lifeworld and Gender: Associational versus Counterfactual

Thinking’, Sociology 34(4): 707–25.Shapere, Dudley (1974) Galileo: A Philosophical Study. Chicago: University of Chicago

Press.—— (1984) Reason and the Search for Knowledge: Investigations in the Philosophy of Science.

Dordrecht: D. Reidel.Sober, Elliott (1984) The Nature of Selection: Evolutionary Theory in Philosophical Focus.

Chicago: University of Chicago Press.Wittgenstein, Ludwig ([1953] 1978) Philosophical Investigations. Oxford: Blackwell.

■ Stephen Kemp is currently a Lecturer in Sociology, University of Edinburgh.This article was written during his time as a British Academy Post-Doctoral Fellowat the University of Sussex (2001–4). He is interested in epistemological questions


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 190



about the status of natural and social scientific knowledge, and is working on abook analysing the role of conceptual transformation in the development ofknowledge. Address: Sociology, School of Social and Political Studies, AdamFerguson Building, George Square, University of Edinburgh, Edinburgh, EH8 9LL,UK. [email: [email protected]] ■


04_051762_Kemp (JB/D) 18/3/05 10:22 am Page 191



Critical Realism and the Limits of Philosophy

Documents