12. Realism. For Real this Time: Scientific Realism is not a Compromise between Positivism and Interpretivism

Jorge Rivas

12.1. Introduction

Within the fields of international relations (IR) and political science, philosophical realism (from here, ‘realism’) and scientific realism (SR) have been understood primarily through the highly influential works of Alexander Wendt. Given all the attention Wendt has received,1 scientific realists in IR might be happy with the well-known introduction to SR that he has provided to such an extensive audience. But this coverage has come with a great deal of discomfort for many scientific realists because of Wendt’s representation of SR as a compromise, or a ‘Via Media’, between positivism and interpretivism (2006: 182). In other words, while Wendt’s overall project of making ‘the study of ideas in international politics … scientifically respectable’ (2000: 165) has been tremendously successful, it has produced the impression that the content and implications of SR are primarily a compromise between positivism and interpretivism, which results in an unfortunate misunderstanding. While such a compromise may be tenable, and, for some, attractive, this chapter will argue that it should not be seen as the primary understanding of SR. By doing so, it will present alternative understandings on SR’s crucial positions regarding ontology and epistemology, as well as the implications of these positions for scientific realist understandings of the relationship between structure and agency, thus highlighting the real advantages of SR and its difference from positivism and interpretivism. The focus of this chapter is not on challenging Wendt’s work but rather on clarifying the subject of this book (i.e. SR). However, the importance of Wendt’s presentation of SR to many of us working from a realist perspective comes precisely from the tremendous influence Wendt has had in the field of IR, and thus it requires the kind of direct examination found in this chapter.

This chapter’s goal is to emphasise that SR is not a middle-ground position between positivism and interpretivism, and that the latter two positions actually share much more in common with each other than either does with SR. Thus, by associating SR with a Wendtian Via Media, the general picture of SR held by the IR and political science communities has been, in our view, very unfortunate. By declaring himself a positivist while claiming to be using realism, Wendt bundled together the two philosophical approaches (despite the fact that they are so diametrically opposed) and thus undermined the potential of SR as being seen to have an alternative set of solutions to the challenges, and indeed, the quagmire of the positivist/post-positivist debates in IR. Unfortunately, this is not just a matter of labels. The key elements of any social science approach are the philosophical positions on (1) ontology, (2) epistemology, (3) the structure–agency question, and (4) the relationship between the material and the ideational.

In a nutshell, Wendt’s Via Media explicitly claimed to draw on SR to combine the subjective ontology of interpretivism (i.e. ‘anarchy is what states make of it’; ‘states are people too’, etc.) with the objective epistemology of positivism (e.g. ‘The epistemological issue is whether we can have objective knowledge of these structures’ (Wendt, 1995: 75). This chapter will explain how a more coherent realism in philosophy entails the inverse: an objective (realist) ontology and a subjective (relativist) epistemology, a combination which is wholly different from either positivism or interpretivism, and not a compromise between the two. Consequently, rather than adopt the understanding of structure and agency as ontologically distinct, these positions led Wendt to apply an ‘instantiationist’ approach that conflates agents and social structures by claiming that structures are entirely dependent on the understandings and practices of the agents who instantiate them (à la Anthony Giddens’ structuration theory). The result is that the real emergent existence, properties and causal powers of social structures are denied; social structures are confused with both their genesis and their effects. Finally, these philosophical positions end up determining his substantive constructivist IR theory (in privileging an ideationalist account of social structures). Material structures are relegated to being inconsequential because they only have causal force through the interpretations and practices of actors. Wendt accepts that material structures exist, but not that they can have any significant real causal force (‘rump materialism’). In sum, rather than base his constructivist IR theory on the empirical evidence in favour of the causal weight of an ideationalist account of social structures, he bases it on his (idealist) foundational philosophical positions. Thus, in IR, these positions are now associated with SR.2

12.2. Realism as a Via Media?

One of the central goals of Wendt’s work is to argue that the social world can be studied and explained in a scientific manner, and that this is still possible even if ideational factors (meanings, beliefs, rules, roles, ideologies, culture – in other words, ideas) play a central role in the social world. This included not only making the case for the possibility of the scientific study of the ideas, but also defending a more general scientific approach to social science from the postmodern and interpretivist challenges to the credibility, and indeed the very possibility, of social science. Wendt’s goal was not to dispute the constructivist, interpretivist, hermeneutical or postmodernist approaches to the social world, but rather to show that even these typically anti-science approaches can also be grounded on scientific foundations. In order to do this, Wendt sought to marshal a realist philosophy of science to sweep away the ‘unnecessary philosophical barriers’ that have been built over time, both by traditional positivist hesitations and reluctance toward the study of ideas, and by post-positivist objections to any ‘scientific’ study of the social world. As a whole, this is a laudable project, one that all scientific realists in social science would likely support at the philosophical level, even if some would not agree with his emphasis on ideas at the substantive level. What has made many realists uncomfortable is the way in which he went about presenting and deploying SR in the service of this project.

In order to retain scientific legitimacy within a field whose dominant understandings of philosophy of science had not yet discarded or moved on from the mistaken conflation of science (in general) with positivism (a particular approach to science), Wendt sought to construct a path between the alleged positivist/interpretivist conflict over whether social phenomena can be studied in a ‘scientific’ (i.e. ‘positivist’) way (2006: 182). As Wendt rightly points out, ‘after their “Third Debate” most IR scholars today seem to think the idea of such a Via Media is incoherent; one must be either a positivist or an interpretivist’ (2006: 182). More precisely, Wendt describes his philosoph ical goals in Social Theory of International Politics (STIP) in the following way: ‘I tried to do something that, in a justly classic paper, Friedrich Kratochwil and John Ruggie (1986) in effect said could not be done: find a via media between positivism and interpretivism by combining the epistemology of the one with the ontology of the other’ (2006: 182). More recently, Wendt appears to agree with Guzzini and Leander’s claim that his Via Media might be better described as a synthesis of the two rather than a middle way (2006: 182). Of course, what is called the ‘Third Debate’ in IR created this dichotomy; scholars do tend to think one’s philosophical position with regard to the social world must either be positivist and scientific, or interpretivist and antiscientific. Under these conditions, providing a synthesis or compromise of these two approaches may indeed be a very appealing project, but philosophically to call this realism can be very misleading. Of course, a great deal of confusion continues to exist about what is meant by SR, and more importantly, why does it matter? This section will show that, in our opinion, the benefits of SR stem precisely from the fact that it is wholly different from both, and therefore to present it as a combination of the two undermines both its potential and its appeal.

In IR, the dominance of positivism has been under attack since at least the 1980s from what have been called interpretivist, hermeneutical, reflectivist, perspectivist and post-structuralist critiques, and which often are grouped together under the heading of ‘post-positivist’. From the point of view of these interpretivist or post-positivist critiques, the positivist approach to social science is philosophically inappropriate for studying social phenomena. However, these post-positivist critiques are seen, by defenders of positivism, as producing not just ontological relativism, but also as leading to the use of methods and practices in social science which lack rigour, are unscientific, and are possibly even incoherent (Patomäki and Wight, 2000: 13; see also Isaac, 1987 and Gunnell, 1995). As Patomäki and Wight point out, a common resolution to this dichotomy in IR has been to try to build a middle-ground position. Wendt’s Via Media epitomises this kind of compromise. But, as Patomäki and Wight argue, there are two problems with these compromise positions:

1. Staking out a ‘middle ground’ between two problematic approaches does not resolve, but rather reproduces, the problems inherent in both; and 
2. Positivism and interpretivism actually entail many common assumptions and arguments (which realists see as being at the root of their problems).

Sometimes these philosophical discussions can sound quite arcane, but the argument is relatively straightforward: both positivism and interpretivism entail anti-realist ontological positions and base these on their anthropocentric (empiricist and interpretivist, respectively) epistemologies. Thus, both ultimately share a subjective or relativist ontology (Patomäki and Wight, 2000: 215). Essentially, for both positivists and interpretivists, all discussion and postulation about the nature and existence of the world must depend on how it is that we know it. Wendt acknowledges this (1999: 52–3), but by creating a synthesis of the two, he reproduces it. Both approaches also share another key position that philosophical realists fundamentally reject: an acceptance of the positivist representation of what ‘science’ is3 (including the empiricist notion of reality, the Humean ‘constant conjunction’ understanding of causation, the deductive–nomological model of explanation, the Hempelian ‘covering law’ model of scientific laws, an instrumentalist approach to theoretical terms, the reductionist view of entities and structures, and the symmetry of explanation and prediction). This understanding of science is then used by positivists to endorse its application in the social sciences (i.e. naturalism) and used by post-positivists to either reject applying ‘scientific’ methods to social phenomena (i.e. anti-naturalism) or to challenge the project of ‘science’ altogether (i.e. anti-scientism) (Patomäki and Wight, 2000: 216–17; see also Hollis, 1996: 303–4; Isaac, 1987). It is on these crucial issues that realism is so radically different. As Wendt rightly notes, SR is premised on ‘the following three principles: (1) the world is independent of the mind and language of individual observers; (2) mature scientific theories typically refer to this world; (3) even when it is not directly observable’ (Wendt, 1999: 51). Taken to their logical conclusions, the implications of these positions produce a rejection of epistemological objectivism, ontological relativism and all of the above-mentioned aspects of the positivist model of science (Sayer, 1992; Collier, 1994; Lawson, 1997). Thus, we think realism is not a middle ground, compromise or synthesis between these two (quite similar) positions, but rather a radical break which provides an altogether different set of solutions to the problems associated with the Third Debate in IR.

However, as Wendt has noted, the political science and IR debates over the appropriate philosophy of science have been dominated by positivism and the interpretivist reaction to positivism, and have largely ignored realism as an alternative to both (Wendt, 1999; Patomäki and Wight, 2000; see Isaac, 1987 for a similar argument). While various forms of realism have become central in the philosophy of natural science since the ‘overthrow’ of positivism in the 1960s and 1970s (for a classic outline of this overthrow see Suppe, 1977), there appears to be little awareness of this shift in the social sciences in general (Gunnell, 1995: 925) and in IR in particular where positivism and science continue to be conflated and where ‘post-positivist’ alternatives are largely reactions against positivism. As the editor of a recent compilation on philosophy of methods in the social sciences has noted, ‘Despite repeated attempts by social theorists and researchers to drive a stake through the heart of the vampire, the disciplines continue to experience a positivistic haunting’ (Steinmetz, 2005: 3).4 The unfortunate fact remains that the philosophical conflicts in IR, particularly those deriving from the Third Debate, are mired in this dispute between positivism and post-positivism, and to progress we need a better understanding of philosophical alternatives. In fact, despite the extensive debate over positivism, there even has been much confusion over what positivism really means (see Wight, 2002: 26 for many examples of this confusion), which brings us back to the confusion created by Wendt’s Via Media. While claiming to be a scientific realist, Wendt declared, ‘I am a “positivist” ’ (1999: 39), and did so, as Chernoff notes, just when positivism was losing credibility in IR and in the social sciences ‘particularly outside the realms of statistical modeling and rational choice theory’ (Chernoff, 2002: 194). This is all the more confusing because STIP contains an excellent exposition and summary of the problems inherent in the central elements of positivism: empiricism, reductionism, deductivism and instrumentalism (Wendt, 1999: 47–91).

To be fair, Wendt has indicated he sees this as a ‘red herring … since I think it is pretty clear that unless otherwise noted my use of the term “positivism” has always been with a “small p”, which is to say equivalent to naturalism or science’ (2008, personal communication) . However, we hold that the distinction is crucial. As Wight correctly points out, ‘This is an impossible position to hold. One cannot be both a scientific realist and a positivist; the two accounts of science are diametrically opposed on some very fundamental issues … Positivism, in this sense, has lost all meaning’ (2002: 36). As Milan Brglez notes, ‘This might come as a surprise since [Wendt] has been able to see and argue that both positivists and post-positivists within IR share a latent anti-realist philosophical ontology. It is simply inconsistent to accept any kind of positivism after one makes and uses such arguments for developing his ontology’ (2001: 347). There is not the space needed to engage in a discussion of these issues here, but as Chernoff has correctly pointed out, ‘For those familiar with the history of the debates over positivism in the philosophical literature, this is an idiosyncratic definition’ (2002: 194). More importantly, it is also problematic because the model of science entailed by positivism is radically different from that entailed by SR. It is one of the goals of this book to show how a better understanding of SR in IR can overcome the obstacles that the continuing positivist ‘haunting’ of our notions of science creates toward better theorising and the conduct of research in the social sciences (for examples of these obstacles, see Wendt, 1987; 1999; Dessler, 1989; Patomäki and Wight, 2000 and Wight, 2006a). And while interpretivists are right to reject the application of the positivist model of science to the study of the social world (since it is fundamentally flawed even in the study of the natural world), Wendt’s conflation of the two has hindered realists in their project of overcoming the interpretivist objections to social science altogether. Or, as Guzzini and Leander put it in a different context, ‘The real issue at stake is not positivism versus post-positivism, but what exactly this non-positivist “social science” is all about’ (2006: 80). As I will argue, despite recognising the fundamental flaws of positivism, Wendt still held on to a great deal of the positivist model of science, and thus created other problematic outcomes.5

12.3. Philosophy inverted: realism with a subjective ontology and an objective epistemology?

This brings us to the crux of the problem: What exactly is the nature of Wendt’s compromise between positivism and interpretivism? Essentially, as Wendt puts it, ‘Epistemologically I have sided with positivists … on ontology – which is to my mind the more important issue – I will side … with post-positivists’ (1999: 90). Or, as Krasner puts it, Wendt has tried to show, ‘In John Searle’s felicitous formulation it is possible to have a subjective ontology but an objective epistemology’ (2000: 131). Thus, to be clear, Wendt’s Via Media adopts:

 

• a subjective ontology (i.e. the dependence of social structure on ideas; anarchy is what states make of it; states are people too, etc.) and 
• an objective epistemology (i.e. ‘The epistemological issue is whether we can have objective knowledge of these structures’) (Wendt, 1995: 75).

In fact, these two positions are quite remarkable because they are an inversion of what we think a coherent SR implies:

• an objective (realist) ontology (i.e. social structures can be independent of interpretations) and
• a subjective (fallibilist and relativist) epistemology (we can never have an objective knowledge of these structures).

Indeed, as is argued by critical realists, philosophical realism can be summarised as a combination of ontological realism and epistemological relativism (producing judgemental rationalism) (Bhaskar, 1975; 1979; 1986; 1989; Porpora, 1989; Collier, 1994; Archer, 1995; Joseph, 2002; Patomäki and Wight, 2000). In contrast, Wendt’s subjective ontology is relativist and indeed is much closer to the empiricism (and even phenomenalism) of positivism and the idealism of interpretivism. Figure 12.1 illustrates the extent of the difference between Wendt’s Via Media and what we think SR entails.

What is probably most surprising to those still mired in the positivist/interpretivist dichotomy is that scientific realists argue that positivism’s objectivist epistemology is actually connected to a subjective ontology that is remarkably similar to that of interpretivism, and that realism is therefore the only approach that entails a truly realist ontology. Positivism is fundamentally premised on empiricism. The empiricist ontological position is that reality consists merely of the empirical world (i.e. the observable world, the objects of experience, or more fundamentally, the observation data themselves). Empiricists came to this ontological position from the empiricist epistemological position that all knowledge claims must be grounded in empirical observation. Empiricists therefore argued that any talk of reality which is not directly derived from observation is anti-scientific. This continued to be the basic ontological position of some branches of positivist philosophy of social science as late as the 1970s (as typified by behaviouralism). It is important to understand how empiricists arrived at this position to understand fully its implications. Early empiricists, such as Hume, argued that all we can know to be real is observation data (i.e. the sense data themselves, not even that to which sense data refer). From this, they derived the ‘true empiricist’ position that only observation data are real. This was a truly subjective ontological position and seemed to defy common sense. However, if this empiricism was combined with an objectivist epistemology (i.e. that observation produces an objective understanding or account of the object of study), then they could claim to posit and believe in the objective existence of a real world. But realists argue that this objective existence still hinges on it being observable, and therefore its existence is not really independent of our ability to know it. At most, their form of ‘realism’ is what is called ‘empirical realism’, i.e. that the real is limited to the empirical. By definition, observation data are not independent of observation, and therefore cannot be objective. One must distinguish between the actual and the empirical. Thus, even ‘empirical realism’ is ontologically subjective and remains fundamentally anti-realist (Hollis, 1996: 303 also makes the point that the empiricism of positivism is actually a form of anti-realism).

The basic characteristic that distinguishes philosophical realism from other philosophies of science is its realist ontology. Realism means an acceptance of the existence of a reality regardless of human observation or knowledge of it. A coherent philosophical realism should encompass all that this implies. In contrast to the objectivist epistemology of positivism, it is important to emphasise that a consistent ontological realism inherently implies fallibilist and subjectivist epistemological positions. Many critiques of SR incorrectly state that SR implies scientific theories must be true. This is false. Realists argue that observation of reality can never be truly objective (as [1] all observation is theory-laden, and [2] different observations of the same object of study can produce different empirical results), and therefore that epistemology must inherently be subjective (incorrect descriptions of SR as having an objectivist epistemology unfortunately have been made in IR: e.g. Wendt, 1999 and Smith, 1996). Realists argue that positivism’s objectivism combined with its empiricism actually produces a clear-cut subjective ontology as their ontology depends on their observations but their observations are believed to be objective. For positivists, ‘reality’ in the end boils down to experience.

Interpretivists have made their subjectivist epistemological critique of positivism’s objectivist epistemology a central component of their reaction against the dominance of positivism in social science; realists would agree with this critique. However, interpretivists have also argued that because we cannot know the world outside of our knowledge of it (a subjective epistemology) then we cannot posit that it exists beyond our interpretation of it (a subjective, or relativist, ontology). Statements made in constructivist IR, such as ‘the material world does not come classified, and that, therefore, the objects of our knowledge are not independent of our interpretations and our language’ (Adler, 2002: 95), regularly reinforce this view. It is this move from their epistemological position to their ontological position with which realists disagree (again, the ‘epistemic fallacy’). Realists agree with a subjectivist epistemology but argue that it does not imply a rejection of a realist ontology. Rather, they claim that the interpretivists’ relativist ontology is actually just an altered form of empiricism and therefore much more like positivism. Thus, realists argue that positivists and interpretivists share much in common and are much more analogous to each other than either is to realism. As Patomäki and Wight put it, ‘From an ontologically orientated perspective both the positivists and the post-positivists share a common metaphysical structure … What can be considered real always bears the mark, or insignia, of some human attribute; in effect, an anthropocentric philosophy’ (2000: 217).

Wendt rightly points out that all observation is theory-laden, but he ended up endorsing the objectivist epistemology of positivism. This is why statements by Wendt about adopting an objective epistemology can appear to us to be so confusing. I believe that the reason Wendt adopted an objectivist epistemology is that he accepted the positivist position that confuses a ‘scientific’ understanding of something as requiring an ‘objective’ understanding of it. His presentation engaged in the same conflation of ontological objectivity with epistemological objectivity in which positivists and interpretivists engage. This should not be surprising as he is producing a Via Media between the two. His presentation of these issues therefore produced confusion over the relationship between these approaches and realism. Thus, as can be seen more clearly in the four-quadrant diagram in Figure 12.2, Wendt’s Via Media is a compromise between positivism and interpretivism but to us seems far from SR.

These positions have important implications for understandings of social structures, both for the social and for the structures. With regards to natural phenomena, many self-described ‘positivists’ today would recognise the actual world (as opposed to the empirical world) as real, but when it comes to social phenomena most positivists still hold to empiricism; they continue to deny that entities such as the state, the economy, class or community are real. These social entities are still understood in an empiricist manner (i.e. since we cannot observe them, they are not real). Interpretivism also commits to the ontology of true empiricism with regards to social phenomena, arguing that our subjective interpretations of the social world are all that could be real. Often positivists or interpretivists claim that material reality can be distinguished from the empirical observation of it (in other words, that it is actual), but that social reality cannot (i.e. that social reality is inherently subjective and has no external reality beyond human consciousness or cognition). The social is thus seen as inherently relativist.

With regard to structures, while positivists may distinguish between the actual and the empirical (in other words, between events and perceptions of those events), they reject the distinction between the actual and the structural, causal or generative (in other words, between events and the often unobservable underlying structures which cause or generate those events). This form of empiricism does not accept that there are hidden, unobservable, unknown or unrecognised mechanisms really generating actual events. In fact, the Humean model of causation, which underlies positivism as an approach in philosophy, explicitly rejects this possibility. The interpretivist reaction to positivism also denies the separate reality of the generative or structural domain. As Wendt points out, ‘postmodernists agree with empiricists that we should eschew the search for unobservable deep structures, and focus instead on surface phenomena’ (1999: 62). Thus, whether or not there is recognition of the difference between perception and event, empiricism denies the existence of an unobservable, generative or structural level of reality (Collier, 1994; Sayer, 1992).

The next section turns to what these positions imply for social theory and how they relate to Wendt’s approach. I will use one of the central theoretical issues in social theory – the relationship between agents and structure – to provide a discussion of both the social science implications of a subjective ontology and of the vestiges of positivist reductionism to be found in Wendt’s social theory.

12.4. Approaches to the agent–structure relationship

Although the agent–structure debate in IR had been touched on6 prior to Wendt’s influential 1987 article, Wendt deserves credit for bringing an explicit awareness and discussion of the agent–structure problem to the forefront of social theory in IR. In this and later works, Wendt rightly tried to overcome the problems inherent in the two historically dominant theoretical approaches to the agent–structure relationship: individualism and holism. In lieu of these two approaches, Wendt tried to employ what he labelled a ‘realist’ approach. However, a careful examination shows that, because of his subjective (relativist rather than realist) ontology, Wendt’s approach – following Giddens’ structurationism – actually fails to ontologically distinguish structures from agents in that, in his conception, structures are ultimately reduced to and entirely dependent on the understandings and practices of the agents whose behaviour ‘instantiates’ them. In short, Wendt conflates structures with both their causes and their consequences. In the process, the ontological reality and the emergent properties and causal powers of structures end up being denied (see Porpora, 1989; Wight, 2006a; Archer, 1995 and Bhaskar, 1979 for similar discussions; for other discussions in IR, see Dessler, 1989, 1991; Hollis and Smith, 1990, 1992, 1994; Patomäki, 1991 and Carlsnaes, 1992). Since the following realist critiques of structurationist arguments are both controversial and not widely understood in IR and political science, I begin with a very brief overview of individualism and holism and then turn to a discussion of the structurationist approach which conflates the two, and of which Wendt’s approach is a version. The last section contrasts this with a deeper ontologically realist understanding which clearly distinguishes and differentiates structures from agents. I finish by arguing that Wendt’s metaphysics ended up determining his substantive social theory, thus highlighting what, for us, are the serious repercussions of his Via Media.

‘Social structure’ is a foundational concept in social science. However, it is understood and deployed in radically different ways. These differences are tremendously important because particular understandings can rule out other ways of conceptualising, theorising and therefore explaining what happens in the social world. Two broad approaches to social structure have historically predominated in social science. The first approach, which has variously been called individualism, reductionism or microfoundationalism, is most closely associated with methodological individualism, Weberian voluntarism and the microfoundationalist approaches associated with neoclassical economics and rational choice theory. In this conception, ‘social structures’ are simply particular configurations of aggregate individual behaviour that exhibit some temporal pattern or stability (Porpora, 1989: 339–40). Social structures do not actually exist, but rather they are solely the behaviour of agents that produce the patterns that we then call social structures. This approach views the use of social structural terms in social science, such as ‘the state’, ‘society’, ‘community’, ‘class’, ‘capitalism’ or ‘the international system’ as simply abstractions or theoretical constructs posited to understand aggregated individual behaviour. Being just abstractions, social structures cannot have any independent causal force and are therefore epiphenomenal. Social structures do not cause actions or behaviour; they are simply a product of behaviour (Porpora, 1989: 341; see also Wendt, 1999; Wight, 2006a). This approach is rooted in the empiricism, instrumentalism and reductionism of positivism, and is therefore regularly touted by its adherents as the only ‘scientific’ conception of social structures. The empiricist position on the unreality of unobservables, as well as the instrumentalist position on the unreality of theoretical terms, is regularly used to argue that unobservables such as ‘the state’ or ‘class’ cannot actually exist (e.g. Arrow, 1951; Downs, 1957; Olson, 1965; Monroe, 1991 and Levi, 1997 provide very good discussions of these issues). In short, individuals are all we can observe, so individuals are all that are real. The reductionist position claims that science explains things solely by explaining their component elements. These empiricist and reductionist claims have been used to argue that any social theory, to be ‘scientific’, must be built on ‘solid microfoundations’ (i.e. individuals).

The second approach, often called holist or structuralist, has its roots in the Durkheimian tradition in social science, especially in structural sociology, as well as with Parsonian structural-functionalism, but can also be found in continental structuralism, Wallersteinian world-systems theory, some versions of macrostructural historical sociology (an archetypal example is Skocpol, 1979), and some positivist representations or interpretations of ‘structural Marxism’ (Porpora, 1989; Wendt, 1999; Wight, 2006a). In this holistic approach to social phenomena, structures are seen as being real and having real causal effects, whereas the actions of agents are portrayed as solely being a product of these effects. This approach comes in many forms, including quantitative, positivist, ‘covering-law’ versions; historical macrostructural versions; and more hermeneutical, ideational and interpretivist versions. As Porpora puts it, the more positivist and quantitative version divorces social facts and group properties entirely from the subjective and psychological domains and ‘represents social structure as something entirely devoid of the influence of human agency’ (1989: 342). The macrostructural sociology versions explicitly or implicitly reject voluntarism or agency as being significant in macro-outcomes.7 The more hermeneutical and ideationally oriented versions treat discourses and ideas, of which individuals simply become the carriers, as intersubjective structures that constitute and drive human behaviour. While there is not enough space to detail these various approaches, the crucial point is that all versions portray the behaviour of units as being solely a product of the structures in which they are embedded. Agency is replaced by structural determination. Thus, in contrast to the reductionist approach, the benefits of this holist approach are that it recognises the emergent quality of structures (going back at least to Durkheim who was explicit about this) and emphasises the extent to which structures constitute agents and influence their behaviour. But the drawbacks are the tendency to portray agents as the passive carriers of the effects of structures and the privileging of deterministic structural forces over the capacity of agents to act otherwise. What is lost in this approach is the free will, or ‘agency’, of actors (Porpora, 1989).

The third approach arose out of critiques of the first two, and for my purposes here is best represented by Giddens’ ‘structurationist’ approach since Wendt relied heavily on Giddens’ structuration theory (Giddens, 1979, 1981, 1984) to outline his own approach. It is also often found in some versions of post-structuralism and postmodernism where it is claimed that ‘nothing exists outside of discourse’ (Campbell, 2001: 444). In fact, it can be argued that Giddens’ approach is a ‘soft’ form of, and a precursor to, the post-structuralist understanding of social structure (Porpora, 2001: 264). Porpora also associates this approach ‘with ethnomethodologists, some symbolic interactions, and other sociologists who employ a linguistic model to conceptualize social structure’ (1989: 340). This approach aims to avoid the extremes of the first two by taking a kind of compromise position between them in which neither agency nor structure is granted primacy. Giddens essentially ends up arguing that agents and structures cannot be conceived of apart from one another and Wendt adopts this argument. One should emphasise and commend Wendt’s goal of avoiding the problems inherent in solely agentic or solely structural approaches. Unfortunately, from our point of view, he fails to do so fully, and – in another example of how his Via Media derails his other scientific goals – this failure again comes from Wendt’s relativist rather than realist ontological position.

The focus here will be solely on the key aspects of Giddens’ approach that make it conflationist.8 The jumping off point is that Giddens differentiates ‘social system’ from ‘social structure’:

A distinction is made between structure and system. Social systems are composed of patterns of relationships between actors or collectivities reproduced across time and space. Social systems are hence constituted of situated practices. Structures exist in time-space only as moments recursively involved in the production of social systems. Structures have only a virtual existence. (Giddens, 1981: 26, quoted in Porpora, 1989: 345)

For Giddens, patterns of social relationships are not themselves social structures but rather social systems without any causal power that are the product of (i.e. ‘structured by’) social structures. In turn, these social structures are, for Giddens, the rules and resources associated with those social relationships. In short, rules establish social relationships that then have differences in resources. Crucially, as Porpora puts it, ‘Giddens gives analytical priority to rules and in fact denies that the relationships of a social system have any causal properties independent of the rule-following activity of human actors’ (1989: 350). Thus, what others might refer to as ‘social structures’, i.e. patterns of social relationships or distributions of material capabilities, do not themselves have any causal power (Callinicos, 1985; Porpora, 1989; Archer, 1995 and Wight, 2006a all emphasise this point). The attraction to a constructivist such as Wendt is clear: Giddens conceptualises causally effective ‘social structures’ as ‘shared rules’ rather than as the relational or material properties of ‘social systems’.

Giddens characterised himself as a philosophical realist in that he claims that the rules that structure social systems are real causal mechanisms with real causal powers (Giddens, 1979, 1984) and it is this ‘realism’ that Wendt wants to deploy (see Wendt, 1987 for a clear exposition of this). However, on closer inspection, Giddens’ realism is actually quite shallow. Very much like the reductionists and methodological individualists, Giddens insists that the social relationships themselves do not have any independent causal powers (Porpora, 1989; Archer, 1995 and Callinicos, 1985 all make this point). This allowed Giddens to be able to claim that structures (i.e. rules) are real while still remaining in adherence to the dominant view of a ‘scientific’ approach to the social world by accepting that they only have causal force through the practices of the agents which instantiate them. In fact, Giddens explicitly rejects – as naive – theories in which structure is seen as something existing external to the agent (Porpora, 1989: 346–8). Of course, this is precisely the anti-realist objection to the realist view of structure. In other words, for Giddens, the problem with a truly ontologically realist portrayal of structure is that ‘“structure” here appears as “external” to human action, as a source of constraint on the free initiative of the independently constituted subject’ (1984: 16). In contrast, rules, norms, ideology and symbolic orders are all ‘internal’ to agents as a whole as intersubjectively shared cultural constructs. As Porpora concludes, ‘Thus, on Giddens’s own rendering, we are talking about the difference between a concept of social structure as an objective reality and a concept of structure as an intersubjective reality’ (1989: 346, emphasis added). Boiled down, the existence of Giddens’ ‘structure’ depends on the knowledge, or at least tacit acknowledgement, of agents. Structures are neither objective nor material but only subjective and ideational (see Wight, 2006a: 137–54, for a detailed discussion of Giddens’ structuration theory).

The fundamental split between a deeper understanding of social structures and Giddens’ instantiationist version hinges on the independent causal power of objective social relationships. In effect, this is the difference between having a (realist) objective ontology and an (interpretivist) subjective ontology. It is here where the interpretivist half (the subjective ontology) of Wendt’s Via Media compromise rears its anti-realist head. Wendt claims his approach is grounded in realism and he groups Bhaskar and Giddens together in his discussion of social structures, but I argue that his inversion of the two realist positions on ontology and epistemology leads him to adopt an instantiationist view of social structures that is ultimately anti-realist. Despite Wendt’s goals, in the end he endorses a conflationist approach to social structure that fails to distinguish between structure and agency. Wendt makes this quite clear: ‘Because of this chapter’s bias toward structure, however, the following point cannot be emphasized too strongly: structure exists, has effects and evolves only because of agents and their practices. All structure, micro and macro, is instantiated only in process’ (1999: 185, emphasis in original). The claim is being made by Giddens and Wendt that structure is being treated as real, but in fact, both are explicit that the reality of structure is dependent on instantiation by the practices of the agents. It reduces structure to an epiphenomenon of practices. As has been shown elsewhere, this conflationist approach is ultimately a more complex version of the microfoundationalism of the individualist approach (Callinicos, 1985; Bhaskar, 1989; Porpora, 1989; Archer, 1995; Wight, 2006a).

Practitioners of both individualism and conflationism find it hard to accept that social structures, separate from the practices of agents, could themselves be real. Our understanding of scientific realism rejects this microfoundationalism because it rejects the empiricism, instrumentalism and strict reductionism that have characterised positivism (Collier, 1994 has a straightforward discussion of this; see Porpora, 1989 and Archer, 1995 for more detailed comparisons of these approaches). Here, the complete break with positivism is crucial. For us, SR instead entails a causal and emergentist understanding of structures. The key is to recognise that in the realist conception of science, it is the structured relationships of an entity that not only make it what it is but which give it its properties and causal powers. In other words, entities are really structures – that is to say, sets of structured relationships – and their properties and capabilities are a function of these relationships. This is just as true of social entities as it is of natural ones. The causal powers of entities are a product of the generative mechanisms constituted by each entity’s internal and external structure.9 What actually happens in the world, in turn, is a product of the interaction of many different structures exerting their own causal effects, capabilities, forces and tendencies (Outhwaite, 1987; Collier, 1988, 1994; Porpora, 1989). According to the realist understandings of causality and of explanation, the role of science is to explain the causal powers of an object of study in terms of these structures, and to explain actual events or state of affairs in terms of conjunctures of the causal powers of these structures.

Emergence, in turn, has important implications not only for the reality of social structures but also for conceptions of agency. Thus, emergence is critical not only to the realist rejection of reductionism but also to the rejection of holism (or ‘structuralism’).10 A key feature of social phenomena that cause them to be emergent from biological phenomena is the human capacity for choice. This produces the reality of ‘agency’ (as the word is used in the social sciences). As such, agency is a property of social actors which prevents them from simply being the carriers of structures as the holist or structuralist approach entails. Agents may be influenced, constrained, affected and even constituted by larger social structures, but their behaviour is not determined by these structures. Agency is also a feature of human behaviour which ensures that social systems are always open systems (as opposed to the closed systems entailed in controlled experiments). This means that human behaviour can never be predicted in a deductive–nomological fashion in the way that positivist methodological individualist approaches to social science attempt to do, such as neoclassical economics and rational choice theory. It is impossible to isolate ‘laws’ of human behaviour, such as one could in a closed system, not just because it is difficult to construct human experiments, but more importantly because humans can always choose to do other than what these ‘laws’ predict (see Archer and Tritter, 2000 for specific arguments about this). Humans can always choose to act in ways other than what ‘rational actor’ models predict.

12.5. An alternative realist conception of the agent–structure relationship

Taking all these realist arguments into account, there is a conception of the agent–structure relationship which is left out of Wendt’s approach. This fourth approach sees social structures as being enduring sets of social relationships. Broadly speaking, these structured relationships could come in many forms: social positions and roles, the distribution of material things (e.g. factors of production, resources, wealth, or military capabilities), or ideational structures (e.g. distributions of norms, values, rules, beliefs, knowledge). In essence, from a realist point of view, any set of relatively enduring and structured relationships that are a product of social activity and have an effect on the social world could be a social structure. The difference between this realist conception of social structure and that of Giddens is that this can include the actual relations of society,11 rather than being limited to the rules and roles that produce these relations (this point is emphasised by Collier, 1994; Porpora, 1989; Archer, 1995 and Joseph, 2002). The structured relationships themselves can have causal powers and properties.12 From a realist understanding of science, their causal powers reside in and derive from the causal mechanisms that are constituted by the structure of these relationships among social positions. These causal mechanisms can then exert a force or produce an effect. Of course, that force or effect can be neutralised, cancelled out, counteracted, diffused, constrained, modified, or in some manner interfered with, by other mechanisms.13 Together, both the effects of social structures and the actions of agents produce the patterns and regularities of the social world (Bhaskar, 1975; 1979; Porpora, 1989; Collier, 1994; Archer, 1995; Carter, 2000; Joseph, 2002; Wight, 2006a).This understanding of social structures can accommodate both the first (individualist) or second (holist) conceptions of social activity and social facts without denying the other. In fact, Wendt correctly makes the point that ‘Any theory of society can be [re-] interpreted in realist terms’ (1999: 51). However, and most crucially, unlike with the third (conflationist) approach, the social structures, once produced, can endure and thus be clearly distinct from and not just instantiated by the agents which may encounter or inhabit them. This position goes far beyond Wendt’s, and therefore entails a real dualism of structure and agency – that is, that structure and agency are truly distinct and irreducible to each other.

The claim about the ontological independence of structures from agents’ practices is conceptually difficult to accept for those working within either the empiricist or interpretivist traditions. And it appears to be true for Wendt as well. This difficulty comes directly from the fact that social structures are ‘activity dependent’ for their creation, so it would seem that they must also be ‘activity dependent’ for their continued existence. As Archer explains, the source of this difficulty comes from the conceptual leap often made from the truism ‘no people: no social structures’ to the belief ‘this particular social structure: because of these particular people here present’ (1995: 141). This difficulty can be overcome if one focuses, as Archer says, on the question ‘specifically whose activities are responsible for what and when?’ (1995: 141, emphasis added). Giddens’ approach avoids this question, by arguing that social structures only become real when instantiated by the current actions of agents. Furthermore, he claims the activities that produce this instantiation are dependent upon the knowledge of contemporary agents about their actions. Thus, Giddens makes the leap to ‘this particular social structure, because of these particular people here present’. Wendt reproduces this leap in STIP. It is certainly true that some social structures may be dependent on current practices for their instantiation, but that is not to say that all social structures are.14 This is a substantive question that must be answered by the empirical study of different structures.

Wendt relied on Bhaskar for his discussion of social structures and overlapped Bhaskar’s approach with that of Giddens’ structurationism (1987, 1992, 1999). But, by emphasising the pre-existence and autonomy of social structures, Bhaskar was definitely avoiding the conflation of structure and agency, as Archer thoroughly demonstrates (1995: 137–61), and which Bhaskar makes clear in his later works (1986; see also Archer, 1995). The crucial characteristic with both of these aspects of social structures which establish their ontological independence from the actions of agents is temporality. In contrast to Giddens’ ‘instantiationism’, Bhaskar’s model has distinct before (the pre-existing social structures), during (the process of interaction between particular agents and those structures), and after (the reproduced or transformed social structures) conceptions of the relationship between structure and agency (1979; see Archer, 1995: 139 for a comparison of the two approaches). Pre-existence reflects that, at any moment in time, society pre-exists the people who inhabit and eventually reproduce, perpetuate or transform it. At any moment in time, social positions and social roles, such as president, general, CEO, teacher, priest or spouse pre-exist the individuals who inhabit those roles. Similarly, the distributions of material factors such as the means of production or the means of destruction pre-exist their use, reproduction or transformation by the agents who interact with them. Religious beliefs pre-exist their being believed by an individual. Of course the interaction between agents and these structures can change (or reproduce) them (this is discussed below), but that does not mean the structures did not exist prior to the interaction. Humans produce new distributions of wealth out of old ones. They produce new ideas out of old ones. States produce a new balance of power out of an old one. Norms, values, rules and many other intersubjective structures pre-exist their being reinforced, reformulated or rejected (Bhaskar, 1979, 1986). In summary, pre-existence refers to the fact that social structures are not necessarily the creation of contemporary actors. Autonomy refers to the fact that their current effects are not necessarily dependent upon current acts of human instantiation (Archer, 1995: 138).

Thus, there are really four conceptions of what social structures are and how they relate to agents. While these are variously described in the literature (Bhaskar, 1979, 1986; Archer, 1995), I will call them:

1. Reductionism: i.e. ‘agents-only’ (e.g. methodological individualism, Weberian voluntarism, microfoundationalism);
2. Holism: i.e. ‘structures-only’ (e.g. Durkheimian structuralism, continental structuralism, structural-functionalism);
3. Conflationism: i.e. ‘no difference’ (e.g. Giddensian structurationism, some post-structuralism, some ethnomethodology, Wendtian constructivism);
4. Dualism: i.e. ‘both are real, distinct, and causal’. This approach is truly realist and emergentist (this approach can encompass causal mechanisms and explanations from any of the previous three, but some explicit examples include network theorists, Marxists and Bourdieu).

 

12.6. Implications for some key elements of social science

This last section turns to some important implications of the difference between Wendt’s presentation of SR and the approach we support here. I start with the point that Wendt explicitly denies the objective existence of social structures separate from process: ‘… structure has no existence or causal powers apart from process’ (1992: 394–5). A brief examination of Archer’s Morphogenetic model of the structure–agency relationship and social change will help us compare Wendt’s approach with an approach that has a deeper realist structural ontology (1995: 135–61). This model is designed to call attention to the temporality and relatively enduring property of social structures by emphasising the before, during, and after phases in the relationship between structures and agents. The before phase is the existing configuration of structural relations which pre-exist and are therefore autonomous from and irreducible to the social agents who confront them. In the during phase, agents engage in social behaviour within the confines of these structural conditions, including making choices and forming and pursuing their subjective interests/preferences. The social interactions produce an after phase in which they have left unaltered, reproduced, modified or transformed the pre-existing configuration of structural relations. The interactions may also have reconstituted the actors themselves. This leads to the beginning of another cycle. In other words, some relationships simply pre-exist the agents (or their practices) who enter into them. The distribution of material resources, for example, is unlikely to be solely a product of current practices. Indeed, whether someone is born into a rich or poor country (or family) is not a product of that person’s ‘practices’. Whether a country today is a ‘First World’ or ‘Third World’ country is not solely a product of current practices or current rules. Social relations can have deep histories, enduring qualities and emergent causal powers, which are not reducible to current agents or their practices. However, agents, through their practices, may (intentionally and unintentionally) reproduce, modify or transform the autonomous and antecedent structural relations they encountered in the before phase (1995: 135–61).

There are also tremendous implications regarding other concepts fundamental to social science, such as ‘interests’. Wendt has denied the objective existence of interests (e.g. ‘without ideas there are no interests’ (1999: 139; see also 1992: 398)). This claim is frequently made in constructivist and post-structuralist IR and it is commonly deployed against both neo-realist and Marxist theories. This may be applicable to ‘perceived interests’ or ‘subjective interests’ but the denial of the reality of ‘objective interests’, which is typical of philosophically anti-realist approaches, is surprising coming from an explicitly ‘realist’ author who claims to recognise the objective existence of social structures. Again, it is Wendt’s subjective ontology that leads him to conflate two very different things in social science: subjective ‘preferences’ and objective ‘interests’. From a realist point of view, the objective existence of structured social relations and their structured properties would mean that the positions in these relationships may in fact entail objectively structured interests that are independent of the particular actors who occupy these positions and/or how they themselves perceive them. A position in a relationship may have certain objective interests that are independent of whether the agent (who inhabits or has found him/herself in that position) holds, recognises or is even aware of those interests (Bhaskar, 1989; Collier, 1994; Carter, 2000). A widely recognised example is the Marxist understanding of class relations. Being positions in sets of structured relationships, rather than being ‘things’ in themselves, classes can have properties or characteristics inherent in their positions without these properties or characteristics being ‘essential’ to the actual individuals in the relationships and regardless of their subjective interests. A person who does not have access and control over any means of production may have particular objective interests that are very different from those of someone who does have this access and control, regardless of who these people are and whether they recognise these interests (Collier, 1988).

Finally, the fundamental differences between Wendt’s approach and ours can also be seen with regard to the effects of social structures. Wendt claims that material structures only have effects on the practices and patterns in social life because of the meanings agents give them (e.g. ‘The distribution of capabilities only has the effects on international politics that it does because of the desiring and believing state agents who give it meaning’ (1999: 185). However, from what we argue would be a more consistent realist approach, the practices and patterns of international politics may also be generated directly by objectively existing social structures (such as the distribution of material capabilities). A major feature of SR that distinguishes it from the anti-realism of both positivism and interpretivism is the recognition of this intransitive nature of social structures. The nature of reality is transfactual in that the way things appear to actors is not necessarily the way they really are (Collier, 1994). Indeed, some social relations may be completely unrecognised but still intransitively exist. Let us return to class relations as one example of generative structures in the social world. Just as, say, ‘the presidency’ is a position in a set of structured social relations which pre-exists, and is not solely dependent on the particular individual who inhabits that position, the same could be true, for example, of structured positions such as ‘capitalist’ and ‘worker’. Class relations can be antecedent to and autonomous of the individuals who inhabit their positions. However, unlike the presidency, these social relations are not solely a product of roles and rules and therefore are not necessarily directly understood or even recognised by their inhabitants.15 Structured class relations could exist independently of agents’ consciousness (Collier, 1988). Being intransitive, these positions do not depend for their existence on the meaning attached to them by their inhabitants or indeed even on the knowledge of their existence.

As such, these particular relationships could generate social patterns, such as inequality of income, class conflict or imperialist wars, irrespective of whether the individuals involved give meaning to or even acknowledge their existence (Collier, 1988). In fact, some generative structures, such as class relations, may only be identifiable and understood through their effects. This would not deny their existence, however. The existence of their effects would demonstrate the reality of the generative structures even if the actors do not know of them, similar to the way falling out of a window demonstrates the reality of gravity despite the unobservable nature of gravity and regardless of the victim’s knowledge or understanding of gravity (Carter, 2000). Thus, the distribution of material capabilities could indeed produce effects that are independent of the meanings actors give them. Whether these effects are large or significant would be an empirical question. Thus, Wendt’s philosophical critique of Waltzian neo-realism is unfounded. It may indeed be the case, as neo-realism asserts, that the distribution of capabilities may produce effects on system structure and state behaviour independent of how states give them meaning. Whether these theories are correct in their assertions is a wholly separate empirical matter. But Wendt should not dismiss them a priori by way of philosophical arguments.

More importantly, however, many social theories would be more coherent if their explanatory mechanisms were reformulated in this realist manner. For example, in IR, Waltzian neo-realism, if reformulated to reject its microfoundationalist ontological reductionism, would be less internally contradictory. Unfortunately, because Waltz was working within a positivist model of science, he adopted an ontologically reductionist approach to structure, claiming that agents are all that really exist. Waltz’s approach was thus internally contradictory in that he claims that structure explains outcomes yet structure does not really exist. As a result, the coherence of his structuralism has been undermined. Yet, while Wendt (1987, 1999), Ashley (1984) and others claim that this means his approach is not truly ‘structural’, Wight (2006a) is correct in pointing out that this critique is confusing Waltz’s ontology (which is individualist) with his explanatory mechanism (which is structural), and that in fact Waltz’s approach is structural in the sense that it is the system structure that explains outcomes in his theory. Thus, while Wendt claimed his approach was more structural than Waltz’s, I would point out that, whereas Waltz’s approach denies structures their reality while granting structures causal power, ironically it is Wendt’s approach which denies their causal power independent of agents (by conflating the two) while claiming to grant them an ontological reality. Both Wendt’s and Waltz’s approaches would be more internally consistent and convincing if they were reformulated in this realist way: structures are both causally effective and real.

However, by having employed neo-realism and Marxism as examples, it is important not to give the impression that it is necessary for these social structures to be material relationships in order to have enduring, emergent properties and causal powers (see e.g. Patomäki, 2002 for a focus on ideational structures). In fact, I will use one of Wendt’s own theories to provide an example of an enduring and causal ideational structure that does not require practices for its continued existence nor for its causal power. As part of his alternative to Waltz’s theory of international system structure, Wendt replaces Waltz’s materialist understanding with ‘an idealist view of structure as a “distribution of knowledge”’ (1999: 16). Wendt argues that the selfhelp/power-politics version of international anarchy is not inherent to the structure of anarchy as Waltz claims, but rather could originally have been a product of predatory state practices engaged in the distant past. Crucially, Wendt argues that it does not matter whether any state today identifies or behaves in a predatory manner. Wendt correctly points out that, despite this, the resultant knowledge structure today confronts states as a real thing. According to Wendt, the mere knowledge of past states having behaved in such a predatory manner is enough to cause states to behave ‘as if’ other states are predatory, even if they are not, because of the uncertainty that the structure of anarchy creates about state intentions (1992). If examined carefully, it is clear that in this argument this ‘knowledge structure’ was created by past (predatory) practices, but (1) clearly continues to exist after those practices may have long ended, (2) continues to exert a causal force on current practices, and (3) current practices do not have to instantiate the knowledge structure itself. Thus, in Wendt’s own telling, even ideational structures do not depend on current practices for their continued existence nor for their causal power.

12.7. Conclusion

To conclude, I turn to the way in which Wendt’s philosophy and ontology determined his substantive theory. Many authors have been confused by the ‘rump materialism’ found in Wendt’s version of constructivism, arguing that Wendt appears to want to have it both ways. For example, Smith argues:

The problem is that Wendt seems to alter his view of the relationship between the material and the ideational. Sometimes the former has ‘some intrinsic causal powers’, at other times the ideational constitutes the material. Thus he wants to ‘defend a “rump” materialism which opposes the more radical constructivist view that brute material forces have no independent effects on international politics.’ … Yet in other places, Wendt argues the opposite: ‘the effects of anarchy and material structure depend on what states want’; ‘it is only because of their interaction with ideas that material forces have the effects they do’; ‘Ultimately it is our ambitions, fears, and hopes – the things we want material forces for – that drive social evolution, not the material forces as such’… I think that these quotes reveal a confusing and ambiguous picture of the relationship between the material and the ideational … (2000: 154)

I think this confusion arises from the lack of a clear division between Wendt’s philosophical approach (the material world exists and possibly constrains the social world) and his substantive theory (the independent effects of the material world end up being negligible outside of how they are constituted by ideas). Were these two (philosophical vs substantive) positions clearly separated, Smith’s criticisms would be invalid. Seen this way, these two positions are perfectly compatible. But Wendt is, to a great extent, responsible for this confusion because his philosophical arguments adopt an explicitly interpretivist (subjective) ontology. This subjective ontology (rather than substantive evidence) then does much of the actual work of showing that the material world has very little independent effect. Indeed, Krasner pointed out that Wendt’s STIP is primarily a set of theoretical arguments with very little empirical evidence (2000: 135). Wendt’s (subjectivist) ontology has determined his (constructivist) social theory which in turn shaped his (‘anarchy is what states make of it’) IR theory. In other words, his idealist philosophy of science privileged his constructivist substantive theory. Most notably, his idealist variation of SR essentially excludes or rules out materialist approaches such as neo-realism or Marxism. This shaping of substantive theory by philosophical approach is precisely what he says social scientists should not do (1999: 51, 91). This justifiably produces the confusion between what should be separate philosophical and substantive arguments, and unfortunately, as Smith’s comments illustrate, makes SR appear to be confusing and incoherent. Because of the tremendous influence of Wendt’s work, this has had a negative impact on how the fields of political science and IR view SR. The aim of this chapter was to present an alternative understanding of SR, highlighting the central differences and crucial implications for ontology, epistemology, the structure–agent relationship, and the emergent, enduring and causal properties of both material and ideational social structures.

Notes

1. Indeed, it would be hard to overemphasise the widespread impact of Wendt’s work in IR. Wendt’s Social Theory of International Politics (STIP), where his philosophical vision was most thoroughly explicated, has been remarkably influential, winning the ISA’s ‘Best Book of the Decade Award’, and being called one of the ‘key books in the field’ by the editors of Review of International Studies (Footnote 1: 123 of (2007), also pp. 26, 123–4). Alker went so far as to suggest that Wendt has become what Carr, Morgenthau, Kaplan and Waltz each were in the earlier evolution of our discipline (2000: 141). Robert Keohane, for example, stated, ‘Wendt’s book is virtually certain to become a classic work on international relations theory, standard on graduate reading lists’ (2000: 125). Alker similarly wrote that STIP ‘is so impressive an achievement that it has a good chance to become a standard text of the mainline, American-oriented, professional International Relations literature’ (2000: 141). Steve Smith has also said that ‘Wendt’s book is likely to be as influential as Waltz’s ... The book is likely to become the standard account for those working within the social constructivist literature of International Politics’ (Smith, 2000: 151). Guzzini and Leander have similarly stated that ‘Wendt’s work has rapidly become one of the main reference points of the theoretical debate in International Relations (IR)’ (2006: xvii). It is through STIP and Wendt’s other works that IR scholars are most familiar with SR. Of course, one product of this success is that Wendt’s work on the subject of SR is widely considered authoritative in IR. Stephen Krasner, for example, declared Wendt’s ‘discussion of SR ought to be required reading for any student of international relations, or political science for that matter’ (2000: 131). As Guzzini and Leander have noted, STIP ‘was acclaimed and contested even before it was published in 1999’ and has since ‘been the issue of special panels at many conferences, of a forum in the Review of International Studies … and in Cooperation and Conflict … to name just two. It became the object of an interminable series of book reviews, including a paragraph in The Economist (3–9 March 2001: 89), an event perhaps unique for any theoretical treatise in IR’ (2006: xvii).Others, such as Fred Chernoff, while disagreeing with some parts of Wendt’s presentation of the philosophical issues involved, note that, ‘The recent increase in interest in scientific realist foundations for international relations theory [has been] spearheaded by Wendt in various works, most fully articulated in his Social Theory of International Politics’ (2002: 189).

2. As a final note, this chapter is not necessarily an engagement with Wendt today, as he has stated elsewhere that he is ‘already reconsidering [STIP’s] argument from the ground up’ (2006: 184), but rather with regard to his arguments in STIP because of their widespread influence and impact.

3. Wight makes this point for IR (2002: 25); see Gunnell (1995: 925) for political science; see Taylor (1987) for an example of this conflation in political theory; see Bhaskar (1986) and Collier (1994) for a fuller exposition and critique of this in the philosophy of science.

4. See also Smith (1996: 11) for the claim that the importance of positivism comes not from self-labelling but that it ‘has determined what could be studied because it has determined what kinds of things existed in international relations’.

5. Having made these points, I also want to emphasise the many areas where Wendt’s presentation is accurate, incisive and informative (a few of which I will come back to in the course of discussing the other issues in this chapter), including: his excellent critiques of empiricism and instrumentalism, and of the deductive–nomological (D-N) model of explanation; his brief yet highly explanatory summary of abductive inference to the best explanation (IBE); his descriptions of many of the primary and relevant aspects of philosophical realism; his recognition that realism implies a fallibilist approach to knowledge; his claims that positivism has long been abandoned in the natural sciences and the implication that it should be in the social sciences as well; and his argument that ideas and other socially constructed phenomena are nonetheless real.

6. Primarily through the related but different issue of level of analysis; see Wight (2006a: 104–19), for a clear differentiation of these two issues. For the structure– agency debates in IR, see Ashley (1984), Wendt (1987, 1999), Dessler (1989), Hollis and Smith (1990, 1991, 1992, 1994), Patomäki (2002), Carlsnaes (1992), Doty (1997) and Wight (2006a).

7. Skocpol, for example, identified the role given to ‘deliberate effort’ on the part of actors as the primary flaw of non-structuralist approaches, and instead adopts an explicitly non-voluntarist, structural perspective (Skocpol, 1979: 13–15) that focuses only on structurally determined ‘objective relationships and conflicts among variously situated groups and nations, rather than the interests, outlooks, or ideologies of particular actors in revolutions’ (Skocpol, 1979: 291).

8. Again, there simply is not enough space for a detailed discussion of Giddens’ structuration theory, but the interested reader could return to Wendt’s excellent exposition. See also Callinicos (1985), Porpora (1989), Archer (1995) and Wight (2006a: 137–54) for outstanding critiques. 9. Hausman (1989), Fleetwood (2002), Collier (1994), Sayer (1992), Outhwaite (1987), Lawson (1997) and Kurki (2002) all provide good descriptions of the realist critique of, and alternative to, the positivist conception of cause and explanation.

10. See the chapters by David Leon and by Jonathan Joseph for detailed discussions of emergence; see also Collier (1994) and Archer (1995).

11. ‘The poor’ or the ‘less developed countries’, for example, do not inhabit these positions relative to ‘the rich’ or the ‘developed countries’ solely because of the rules and roles surrounding poverty or lack of development. The position in the relationship is itself likely to also be a product of its direct relationships to other material structures. And in contrast to what Giddens and Wendt claim, these relationships to the material world may be a cause of actor behaviour in that they could constrain or shape the behaviour independent of any meanings, norms, identities or rules attached to their associated ‘roles’. Indeed, these material relationships may help to constitute the very identities of the actors and the associated rules themselves, rather than simply be a product of these identities and rules. This is not to say that what produces the positions in social relationships are only material. They could be material, ideational or both (a subject I will return to below). It is an empirical matter to determine the causal weight of these material or ideational factors. One cannot determine this a priori. Materialist theories, such as neo-realism or Marxism, may a posteriori claim that the material world matters more (similarly, idealist theories could claim the reverse), but this should be based on the weight of the empirical evidence drawn from research on the object of study, not on the basis of philosophical positions. These are substantive claims about the way in which the social world works, are open to empirical disconfirmation and cannot be established through philosophical arguments.

12. Porpora’s (1989) and Archer’s (1995) explanations are probably the most wellknown for this understanding of social structures. For an application of realist understanding of structures to ‘hegemony’ in IR, see Joseph (2002).

13. Thus, the mechanisms which generate the ‘tendency of the rate of profit to decline’ in capitalism, or the ‘law of comparative advantage’ in trade, or the ‘thesis of declining terms of trade’ for countries in the periphery, may indeed exist, but in particular contexts the effects they generate may be counteracted by other mechanisms. The fact that an actual outcome is not a Humean constant conjunction does not, in itself, disprove the ‘law’ because the law refers to the effects of the causal mechanism, not their empirical instantiation through event regularities.

14. An example that stands out in IR could be the global distribution of wealth. 15. The example of class relations is used here for explanatory purposes, but the

arguments can apply equally to other systems of social relations (Porpora, 1989). Carter (2000), Realism and Racism, is an excellent illustration of how this would apply to social positions based on race; see also Wight (2003). Similar arguments could be made about social positions based on, for example, nationality, ethnicity, gender, citizenship/immigration status or family structure. In international politics, relationships may be based, for example, on military capabilities; on positions in the world economy; on membership in international organisations such as the UN Security Council, IMF, World Bank and OECD; or, as Wendt argues, on Hobbesian, Lockean and Kantian cultures and identities.