Constructivism and International Relations

Emanuel Adler. Handbook of International Relations. Editor: Walter Carlsnaes, Thomas Risse, Beth A Simmons. Sage Publication. 2002.

In this chapter I will explore where constructivism came from, what brings constructivists together–and thus sets them apart from adherents of other international relations (IR) approaches–what divides constructivists, and where constructivism is and should be going. In particular, I will show that constructivists deal extensively with metaphysics and social theories less for their own sake than because constructivism provides a firm basis for building better IR theories.

In addition, I will argue that despite the divisions among constructivists concerning serious issues, all constructivists (modernist, modernist linguistic and critical)–with the exception, perhaps, of the extreme postmodernist wing of radical constructivism–share two understandings: what Stefano Guzzini (2000: 149) summarized as the social construction of knowledge and the construction of social reality. In combination, these understandings are constructivism’s common ground, the view that the material world does not come classified, and that, therefore, the objects of our knowledge are not independent of our interpretations and our language. This means that different collective meanings are attached to the material world twice, as social reality and as scientific knowledge. In other words, knowledge is both a resource that people use in their day-to-day life for the construction of social reality, and the theories, concepts, meanings and symbols that scientists use to interpret social reality.

This dichotomous description is offered for analytical purposes only. For reflexive knowledge or interpretation of the world, when imposed on material reality, becomes knowledge for the world–the power to change the world in accordance with collective understandings and, concurrently, with human motives and intentional acts. The above analysis means not only that there is no perfect correlation between objects ‘out there’ in nature and our classifications of nature, but also that social facts, which are the objects of our study, emerge from the interaction between knowledge and the material world, neither of which is invariant.

Unlike positivism and materialism, which take the world as it is, constructivism sees the world as a project under construction, as becoming rather than being. Unlike idealism and post-structuralism and postmodernism, which take the world only as it can be imagined or talked about, constructivism accepts that not all statements have the same epistemic value and that there is consequently some foundation for knowledge.

I start by tracing four constructivist IR approaches to their philosophical and sociological roots and suggest a synthesis between pragmatism and realism. The next section provides a brief historical account of the evolution of IR constructivism. In my third section I describe three aspects of IR constructivism: (1) the common ground (in ontology, epistemology and methods), (2) conceptual contributions to IR theory (what I call its ‘added value’) and (3) substantive empirical contributions. The fourth section then introduces the major debates within constructivism. Finally, I propose an agenda for helping constructivism become more firmly established in IR. In particular, I emphasize the need to focus constructivist debates on methodological issues.

The Philosophical and Sociological Foundations of Constructivism

Constructivism, which reached the shores of IR in the 1980s, describes the dynamic, contingent and culturally based condition of the social world. It has major implications for an understanding of knowledge, including scientific knowledge, and how to achieve it. Constructivism thus has the potential to transform the understanding of social reality in the social sciences. It stresses the reciprocal relationship between nature and human knowledge and suggests a view of the social sciences that is contingent, partly indeterminate, nominalist, and to some extent externally validated (Kuhn, 1970). With the exception of its radical postmodern wing, however, constructivism does not challenge science, rationalism and modernity; it merely makes science more compatible with the constructivist understanding of social reality.

Let us begin by putting to rest the widely held assumption that constructivism is yet another IR ‘ism,’ paradigm, or fashion, which, highlighting the role of norms in IR and offering a more optimistic approach to IR, has recently joined the ranks of ‘realism’ (neorealism) and ‘liberalism’ (neoliberalism). Constructivism is in fact a three-layered understanding–involving metaphysics, social theory and IR theory and research strategies–of social reality and social science and of their dynamic mutually constitutive effects.

First, constructivism is a metaphysical stance about the reality that scholars seek to know and about the knowledge with which they seek to interpret reality. This position has been applied not only to IR but also to the social sciences in general (for example, sociology, psychology and education), to mathematics and, via the philosophy of science and the sociology of knowledge, to the natural sciences. Thus from an IR perspective in which paradigms are associated with broad world-views of international political life (such as realism, liberalism and Marxism), constructivism is more like a paradigm of paradigms.

Second, building on the metaphysical position, constructivism is a social theory about the role of knowledge and knowledgeable agents in the constitution of social reality. It is as social theory that, for example, we should understand the role of inter-subjectivity and social context, the co-constitution of agent and structure, and the rule-governed nature of society.

Finally, constructivism is an IR theoretical and empirical perspective that, building on the other two layers, maintains that IR theory and research should be based on sound social ontological and epistemological foundations. IR constructivism has led to new and important questions, for example, about the role of identities, norms and causal understandings in the constitution of national interests, about institutionalization and international governance, and about the social construction of new territorial and non-territorial transnational regions.

Debates within IR constructivism take place on all three levels–metaphysics, social theory and IR theory. IR constructivists have often inadvertently ‘jumped around’ the three levels, without specifying whether the points they are making are about metaphysics, social theory, or IR. This may be one of the reasons for the misunderstanding and confusion that exist outside the constructivist camp and for the charges that constructivists do only ‘meta-theory.’ Constructivists, however, are the first group of political scientists to have grounded IR theory on an explicit metaphysics and social theory. Not only does this grounding promote more realistic social assumptions; in the wake of the flood of recent empirical constructivist work, it also disposes of the charge that IR constructivists are meta-theorists. Constructivists could not have reached level three (IR constructivist theory and research) without levels two and one (social theory and metaphysics). Indeed, constructivists could not have approached non-constructivists without letting the latter know that the constructivist picture of the social world (and the way to attain knowledge of the world) is not at all similar to theirs. In fact, the argument of non-constructivists–that IR does not need to be grounded on metaphysics and social theory, or that metaphysical and social theory assumptions should remain unspoken–is a social-construction move par excellence.

Because constructivism in the social sciences builds on centuries of intellectual developments in philosophy, sociology and social theory, it is not easy to speculate about its origins. This is not the place for an intellectual history of constructivism. To illustrate the roots of the debates within IR constructivism, however, I will present four currents of thought that have affected IR constructivism: neo-Kantian ‘objective hermeneutics,’ linguistic ‘subjective hermeneutics,’ critical theory and pragmatist philosophy of science. I will then describe four IR constructivist approaches–modernist, modernist linguistic, radical and critical–which rely, directly or indirectly, on one or more of the above currents of thought–and a strategy for bridging between them.

Constructivism can be traced back to Immanuel Kant–whom Ian Hacking describes as ‘the great pioneer of constructivism’ (1999: 41)–and to nineteenth- and early twentieth-century ‘neo-Kantians.’ Kant believed that although knowledge can tell us something about objective reality, it must nevertheless be ‘restricted to the realm of phenomena, or that which appears to consciousness’ (Delanty, 1997: 45). Neo-Kantians took Kant’s insight–that to know means imposing the a priori forms of our minds on the structures of nature–and carried it from nature to culture. For example, in the late nineteenth and early twentieth centuries, Wilhelm Dilthey (1989) and Edmund Husserl (1962) set the human sciences apart from the natural sciences. Max Weber (1978) called for an autonomous social science, based on the understanding of meaning (‘verstehen’) and the explanation of motivations that lead to actions. More recently (building on Alfred Schutz (1962)), Peter Berger and Thomas Luckmann (1966)–who like Georg Simmel (1955) stressed the intersubjective nature of everyday knowledge, and the interpreted nature of social reality–coined and developed the concept of ‘the social construction of reality.’ Neo-Kantianism, in sum, is an objective approach to hermeneutics. Working within the realm of reason, it stresses the need to understand consciousness. Because it believes in the possibility of attaining empirical knowledge without the mediation of language, it aims at explaining society. Neo-Kantianism generally follows a ‘particularizing positivist strategy that reconstructs historical processes and narratives, rather than Carl Hempel’s (1965) covering-law type positivism, which aims at prediction. Neo-Kantianism, which I will hereafter call the ‘weak programme’ of constructivism in the social sciences, looms large in modernist versions of IR constructivism.

Constructivism’s ‘strong programme’ in the social sciences is based on a turn from consciousness to language and from objective to subjective hermeneutics. Led by Martin Heidegger (1962) and Ludwig Wittgenstein (1953), this current of thought directly challenged positivism and argued that social facts are constituted by the structures of language and that, accordingly, consciousness can be studied only as mediated by language. To be sure, as with John Austin’s (1962) student, John Searle (1995)–who adopted the notion of the construction of social reality while rejecting the social construction of nature–not every turn to linguistics was a relativist turn. In general, however, the turn to linguistics radicalized the anti-positivist movement by taking science as being at best ‘forever constrained by its social context’ (Delanty, 1997: 53) and at worst as a discourse that cannot attain objective knowledge or criticize society. Peter Winch (1958), who argued that social action is ‘“rule following” within a concrete form of life’ (Delanty, 1997: 55), brought the radical linguistic logic to the social sciences. Arguing that the objects of science are socially constructed in laboratories, Bruno Latour and Steve Woolgar (1986) carried this linguistic logic from society to nature itself. Post-structuralists like Jacques Derrida (1978) and postmodernists like Michel Foucault (1980), however, challenged reason, science and modernity, thereby bringing radical linguistics to its ultimate relativist destination. We can trace two IR constructivist approaches–a ‘modernist linguistic’ approach and a ‘radical’ approach–to constructivism’s strong programme in the social sciences.

Critical theory, which resulted from an attempt by the so-called Frankfurt School to turn a Marxist critique of political economy into a critique of ideology (Adorno, 1976; Horkheimer, 1972), falls between the weak and the strong programmes of constructivism in the social sciences. Jürgen Habermas (1978, 1984, 1987, 1996) combined objective hermeneutics and language philosophy in order to extend critical theory into (a) a powerful critique of instrumental rationality and (b) a social theory of ‘communicative action’ and ‘deliberative democracy.’ Habermas’s critique of instrumental rationality showed that the social sciences should abandon the cognitive interest in control, which is characteristic of instrumental rationality, in favour of a cognitive interest in emancipation. In turn, his social theory explained how emancipatory interests become reconstructed in both theory and practice and especially how deliberative democratic processes help people free themselves from distorted communication. ‘Critical’ IR constructivism builds on Habermas’s blend of insights from the philosophy of language with beliefs in the distinction between the natural and social sciences, the possibility of explanation in the social sciences and human progress.

Another current of thought that bridges the weak and strong programmes of constructivism in the social sciences is pragmatism. Dismissing the Cartesian notion that we must choose between objectivism and relativism, pragmatism (Dewey, 1977; James, 1975; Peirce, 1966) suggests that we need to adjust our ideas about truth as experience unfolds (Smith, 1996: 23). More specifically, it underscores the role of choice, deliberation, judgement and interpretation by communities of scientists, who immerse themselves in a type of rational persuasion that must aspire, but cannot always be assimilated, to models of deductive proof or inductive generalization (R. Bernstein, 1985). For example, Thomas Kuhn’s (1970) pragmatist philosophy of science played a large role in the development of the four IR constructivist approaches mentioned above. Although not a pragmatist, Karl Popper (1982), who stressed the role of background expectations in the development of scientific theories, also contributed to the development of IR constructivism.

The above philosophical and sociological approaches imprinted the various strands of IR constructivism, which I will describe in brief. A modernist type of constructivism in IR (John Ruggie (1998a: 35) called it ‘neo-classical’) results from the combination of objective hermeneutics with a ‘conservative’ cognitive interest in understanding and explaining social reality. Thus, for example, IR modernist constructivists, such as Emanuel Adler and Michael Barnett (1998), Jeffrey Checkel (2001), Martha Finnemore (1996), Peter Katzenstein (1996), Thomas Risse-Kappen (1995), John Ruggie (1998a), and Alexander Wendt (1999), uncover the causal social mechanisms and constitutive social relations that make IR more intelligible.

Modernist linguistic (or ‘rules’) constructivism results from the combination of subjective hermeneutics with a ‘conservative’ cognitive interest in explaining and understanding social reality. Modernist linguistic or rule-oriented constructivists, such as Friedrich Kratochwil (1989) and Nicholas Onuf (1989), believe that, because of the primacy of epistemology, understanding social reality means uncovering the processes by which social facts are constituted by language and rules. They are interested in explaining how social rules (including legal rules) and what Austin and Searle have called ‘speech acts’ ‘make the process by which people and society constitute each other continuous and reciprocal’ (Onuf, 1998a: 59). Other proponents of modernist linguistic constructivism, such as Karen Litfin (1994), Neta Crawford (1999) and Jutta Weldes (1999), though emphasizing discourse and its power to construct social reality, nevertheless conduct empirical historical and interpretive research with the aim of understanding the emergence of social reality.

Radical constructivism in IR, which often embraces postmodern and post-structuralist perspectives, results from a combination of a radical turn to language (and thus to subjective hermeneutics) with a dissident emancipatory or deconstructionist attitude toward knowledge in general. As such it lies at the extreme edge of the strong programme of constructivism in the social sciences. In general, radical constructivists do not question the existence of material reality; sometimes they even conduct empirical research (Der Derian, 1987; Doty, 1996; Gill, 1988; Weber, 1995). Because, however, they believe that material reality cannot be truly represented, they are agnostic about material reality and prefer to concentrate on discourse, narratives and texts (Ashley, 1987; Campbell, 1992; Der Derian, 1990; Peterson, 1992; Walker, 1993). What drives many radical constructivists outside the constructivist ‘common ground’ is neither their emancipatory or deconstructionist cognitive interest, nor their insistence on uncovering power structures disguised as truth and their pessimistic view about the social world. Rather, it is their view that no statements can be more valid than others, that nothing can be done to assess the validity of normative and epistemic claims, and that science is accordingly just one more hegemonic discourse.

Critical constructivism in IR results from the combination of objective hermeneutics (mainly the approach of Habermas and his followers) with a dissident interest in the emancipatory effects of knowledge. Critical constructivists, such as Andrew Linklater (1998) and Robert Cox (1986) (who follows Antonio Gramsci (1971)), share the view that striving for a better understanding of the mechanisms on which social and political orders are based is also a reflexive move aimed at the emancipation of society. In general, critical constructivists follow a pragmatist approach, to which I now return. Pragmatic realism, a term I borrow from Hilary Putnam (1990, 1998) to designate a combination of modernist pragmatism and scientific realist philosophy (especially a ‘critical realist’ view of the construction of social reality), may provide a way to consolidate the common ground within IR constructivism. Pragmatic realism says that although representations of the natural and social world are always made from a point of view and are thus interpretations, there none the less exists a material reality outside human interpretations; social facts emerge from the attachment of collective meaning to a previously existing material reality. It follows that rules that evoke reasons for action, individuals’ reasoning processes, and collective understandings within dialogical communities–all of which are part of a pragmatist interpretation of social reality–may also be interpreted as being part of the social mechanisms that scientific realists believe help explain social reality. These mechanisms, and the structures on which they are based, involve the intersubjective ‘stuff ’ that makes material reality meaningful; they do not exist outside human practices. Hence pragmatic realism does not postulate the sameness of the natural world and the social world.

The Evolution of Constructivism in International Relations

One often reads (e.g., Price and Reus-Smit, 1997) that IR constructivism was a result of IR theory’s ‘third debate’ (Lapid, 1989) and that the end of the Cold War made it popular (Checkel, 1998; Hopf, 1998). This is true if one looks at the immediate conditions of its acceptance and growing influence, including the partial disenchantment with materialist and positivist views of social science and, in the wake of the end of the Cold War, by the dismal record of prediction in IR. More generally, the IR discipline has also responded to some earthshaking changes, such as the decline of sovereignty, the growing social and economic importance of knowledge, globalization, the Internet, and changes in the natural environment. These and other changes have been bringing home the post-positivist message that ‘science is not independent of its object but constructs it’ (Delanty, 1997: 8). Despite all this, it was in fact a century or more of interpretative sociological scholarship that penetrated IR at least a decade before the end of the Cold War that made IR constructivism possible.

Another common narrative, which, though it gives due credit to some pioneers of constructivism, is narrow and suffers from a short memory, is that in the beginning there was Onuf (1989), who coined the concept of constructivism in IR; then there was Wendt (1992b)–and the rest is history. To refine this narrative, we should add some synergetic links between people, trends and research programmes that made IR constructivism possible. Because I have done no research on the matter and have only a few pages in which to tell the story, however, I will not pretend to write a full history of IR constructivism here.

Some of the credit for the development of IR constructivism should go to the radical constructivists who, in the late 1970s, shocked the IR community by building their arguments around Foucault (1980) and Derrida (1978). Dialectically, they opened a space for the development of less radical strands of constructivism, which I have identified as the ‘middle ground’ (Adler 1997). Particularly influential were works by Ashley (1987) on power, practice and international community, by James Der Derian (1987) on diplomacy, by David Campbell (1992) on US foreign policy, by Andrew Linklater (1990) on moral community, and by R.B.J. Walker (1993) on sovereignty. Also influential were the neo-Gramscian critical theory of Robert Cox (1986) and Ann Tickner’s (1992) feminist theory. It was mainly their work that Yosef Lapid (1989) had in mind when he wrote his powerful article on the ‘third debate’ in IR theory.

IR constructivism, however, has older and deeper roots. Karl Deutsch et al. (1957) and Ernst Haas (1958) anticipated modernist constructivism. In the 1950s Deutsch promoted a research programme on security communities, which dealt with peaceful transnational collective identities. Deutsch himself was not a constructivist–constructivism had yet to make its way from sociology to political science–and favoured a positivist epistemology. His sociological approach, however, which emphasized social transactions and social communication, had an indelible influence on later developments in constructivism. For example, Hayward Alker (1996), who studied with Deutsch, became a leading methodologist working within the constructivist tradition. And Peter Katzenstein (1996), who also was a student of Deutsch’s, edited a trailblazing book on culture and national security. Many of its chapters were written by Katzenstein’s students, who also became leading and widely published constructivists, including Audie Klotz (1995), Richard Price (1995), Christian Reus-Smit (1999), and Nina Tannenwald (1999). Also in this book, Iain Johnston (1995, 1996) and Elizabeth Kier (1996, 1997) introduced a distinctive perspective on strategic culture. More recently, Adler and Barnett (1998) put a constructivist spin on Deutsch’s security community concept.

Although Raymond Duvall did not study with Deutsch, he collaborated with Bruce Russett, who did (Russett and Duvall, 1976). Duvall became the mentor of, among others, Wendt (1999), Michael Barnett (1998), Roxanne Doty (1996), and Jutta Weldes (1999). In a seminal 1987 article, Wendt brought Giddens’s (1979, 1984) structuration theory and scientific realism to the attention of IR scholars; David Dessler (1989) followed suit shortly thereafter. Wendt then wrote a series of very important articles (1992b, 1994) and a book (1999); these established him as one of the leading constructivist scholars. I am not saying that Duvall and Wendt owe their constructivist perspective to Deutsch. It is noteworthy, however, that the substantive part of Wendt’s theory deals with security community-like collective identity formation. It is also noteworthy that in the early 1960s Onuf, one of the most influential early constructivists, studied with Deutsch at Yale. According to Onuf, Deutsch ‘got him thinking’ about constitutive and regulative legal action (personal communication). Onuf’s 1989 book, where he first referred to the interpretive turn in IR as ‘constructivism,’ along with Kratochwil’s 1989 book on rules, norms and decisions, became a beacon for modernist linguistic and rule-oriented constructivist research. It was no coincidence that both books promoted a legal theoretical approach; Kratochwil briefly studied with Onuf at Georgetown and they afterwards maintained a dialogue that lasted fifteen years.

In the early 1980s Ernst Haas (1983) suggested a powerful sociological of international co-operation based on learning, that is, on the introduction to politics of scientific consensual understandings. Borrowing the concept of ‘episteme’ from Foucault, Ruggie (1975), who studied with Haas, further developed this programme, which P. Haas (1992) and Adler and P. Haas (1992) turned into an agent-oriented constructivist research programme on ‘epistemic communities.’ Adler (1991) also used E. Haas’s ideas to develop the concept of ‘cognitive evolution,’ a constructivist interpretation of collective social learning, which involves the innovation, selection and international diffusion and institutionalization of collective understandings. In 1986, Ruggie joined forces with Kratochwil, who came to constructivism via insights from international law and language-based ‘speech-act theory’; together they wrote a seminal article on international regimes from a constructivist perspective (Kratochwil and Ruggie, 1986). Robert Keohane (1988) picked up the gauntlet thrown down by these two scholars, whom he called ‘reflectivists,’ and challenged them and other ‘reflectivists’ to develop empirical research along positivist lines. This call stimulated a second generation of constructivists to engage in empirical research, although generally not along positivist lines. Ruggie and Kratochwil later wrote a series of important articles that helped establish the modernist type (Ruggie, 1993b, 1998a) and linguistic type (Kratochwil, 1993, 1996) of IR constructivism.

In addition to Kratochwil, other German scholars were prominent in the development of constructivism, mainly by initiating an important debate between instrumental rationalists (e.g., Keck, 1997) and Habermas-inspired communicative rationalists (Müller, 1994). For example, Risse (2000) not only did important work on communicative rationality, but also became a ‘conveyer belt’ of ideas between German and American constructivist scholarship.

The English School, which interprets IR as being social and historical, and which stresses the existence of an international society that is driven by norms and identity (Buzan, 1993; Hurrell, 1993), played a role in promoting constructivist ideas (see Jepperson, Wendt and Katzenstein, 1996). In fact, some theorists of the English School (e.g., Dunne, 1999) argue that the differences between ‘middle ground’ constructivism and the English School are small (but see Finnemore, 1996; Waever, 1999). Moreover, the work of English School scholars (e.g., Jackson, 1990; Linklater, 1998), especially those of the latest generation (Dunne, 1995; Epp, 1998; Rengger, 1999) has sometimes gone further than modernist constructivism in stressing discourse and the critical aspects of knowledge.

We cannot talk about the English School’s influence on IR constructivism, however, without also referring to the collaboration between Buzan and Waever (1997). In addition to Buzan and Waever’s ‘Copenhagen School,’ many other Scandinavians have had a strong impact on the evolution of constructivism in IR. Walter Carlsnaes (1992), for example, was one of the first scholars to build on critical realism; he was followed by Heikki Patomäki (1996). Iver Neumann (1999), in turn, conducted important studies of collective identity formation. Scandinavians (e.g., Carlsnaes, 1992, 1994) also played an important role in early agent–structure debates and helped establish important IR journals, such as the European Journal of International Relations, that became a forum for the publication of constructivist ideas. Lately, Scandinavian scholars, such as K. Erik Jorgensen, have also played a role in bringing constructivism to studies of European integration (Christiansen et al., 1999).

Back in the United States, Martha Finnemore (1996) brought John Meyer’s (1980) ‘sociological institutionalism’ to IR; her stress on the diffusion of Western norms to the Third World reinforced constructivist arguments about the constitutive effect of cultures. Constructivists (Barnett and Finnemore, 1999; Finnemore and Sikkink, 1998) also used other forms of sociological institutionalism (March and Olsen, 1998; Powell and DiMaggio, 1991) to explain the generation, diffusion and institutionalization of culture. In addition, a sociological turn toward social movements and networks also made inroads into IR constructivism, especially the idea of ‘transnational advocacy networks’ (Keck and Sikkink, 1998).

IR Constructivism’s Common Ground, ‘ Added Value’ and Substantive Empirical Contributions

IR Constructivism’s Common Ground: Ontology, Epistemology and Methods

All strands of constructivism converge on an ontology that depicts the social world as inter-subjectively and collectively meaningful structures and processes. In this world, ‘material resources only acquire meaning for human action through the structure of shared knowledge in which they are embedded’ (Wendt, 1995: 73). Several crucial implications follow from this. First, the social world is made of intersubjective understandings, subjective knowledge and material objects (Popper, 1982; Searle, 1995). The world that constructivists see, therefore, is neither better nor worse than the world seen by neorealists and neoliberals. But it is a world that is broader, more contingent, more unexpected, more surprising and endowed with more possibilities. Second, social facts, which are facts only by human agreement and which account for the majority of the facts studied in IR, differ from rocks and flowers, because, unlike the latter, their existence depends on human consciousness and language. In other words, social facts depend, by way of collective understanding and discourse, on the attachment of collective knowledge to physical reality (Searle, 1995). For example, when we classify and refer to some people as ‘self’ and to other people as ‘the other,’ a notion of what is in ‘our’ interest, as opposed to the ‘other’s’ interest, emerges. Third, although individuals carry knowledge, ideas and meanings in their heads–where else would they be?–they also know, think and feel only in the context of and with reference to collective or intersubjective understandings, including rules and language. In other words, it is from this context or background that people borrow the epistemic, normative and ideological understandings, rules and discourses that ‘make individuals into agents by enabling them to act upon the world in which they find themselves’ (Gould, 1998: 81). Fourth, constructivists (except for radical constructivists) all consider the mutual constitution of agents and structures to be part of constructivism’s ontology.

Again with the exception of the radicals, constructivists share, albeit only partially, an epistemology that makes interpretation an intrinsic part of social science and that stresses contingent generalizations. Contingent generalizations do not freeze understanding or bring it to closure; rather, they open up our understanding of the social world. Moreover, constructivists of all types are not interested in how things are but in how they became what they are. In addition, most constructivists agree with the premise that the validation for knowledge is only partly internal. In other words, constructivists argue that even were it possible to grasp social reality’s minimalist foundations and thereby inch toward truth, in practice theories are far from being true pictures of the world.

This consensus notwithstanding, there are also wide epistemological disagreements among constructivists. For example, some modernist constructivists follow scientific realism (Carlsnaes, 1992; Wendt, 1999) and look in the workings of social mechanisms for causal and/or constitutive explanations of social phenomena. Other modernist constructivists (Barnett, 1998; Reus-Smit, 1999; Ruggie, 1998a) establish causality by means of abduction or ‘a process of successive interrogative reasoning between explanans and explanandum’ (Ruggie, 1998b: 880), thereby teasing out tentative explanations from thickly described narratives. Still other modernist constructivists embrace a ‘particularizing positivist strategy’ (Katzenstein, 1996; Sikkink, 1993), or, like Checkel (2001), take constructivism to be consistent with positivism’s generalizing or covering-law strategy.

Modernist linguistic constructivists (Kratochwil, 1989; Onuf, 1989) and critical constructivists (Cox, 1986; Williams and Krause, 1997) reject the natural-science concept of causation and argue that ‘to ask for a reason for action is to try and find the rule that led to the action’ (Smith, 2000: 159). Consequently, their approach to the social world is based on consensus within a community of research practitioners; to arrive at ‘truth’ they use argumentative procedures, abduction, narrative analysis and practical reasoning. Finally, a majority of radical constructivists embrace postmodernist pragmatism and study how the world is ‘talked into existence’ by means of signs, discourse and narratives.

Constructivists use a large variety of methods: positivist, post-positivist, quantitative, qualitative, and combinations thereof. The conventional qualitative methods most used by constructivists include case studies (Klotz, 1995); process tracing, including process tracing of ideas and their institutionalization in practice (Sikkink, 1993); counterfactuals (Checkel, 2001); and the comparative method (Reus-Smit, 1999). A combination of quantitative and qualitative empirical methods–what Alker (1996) calls ‘emancipatory empiricism’–has also been used to promote a critical approach. Some constructivists have followed the conventional path of grounding research on one concept, such as ‘epistemic communities’ (Adler and Haas, 1992; P. Haas, 1992); others have used formal methods, such as agent-based models (Cederman, 1997); while some have used statistics (e.g. Adler, 1987). Interpretive methods applied with great success have included genealogy (Bartelson, 1995; Price, 1995), ethnography (Zabusky, 1995), semiotics (Bially, 1998), discourse analysis (Milliken, 1999), narrative analysis (Barnett, 1998) and a combination of cognitive mapping and symbolic analysis (Johnston, 1995). All of these methods have proven useful in identifying background intersubjective meanings and social structures and the agents involved in social processes. They also have helped identify the reason why some discourses and practices become established, but not others, and the minimal foundations that validate some statements rather than others.

Several things stand out in this diversity, uniting many constructivists and setting them on a collision course with positivism, however. First, there is the notion that the quest for explaining causal processes requires the interpretive practice of uncovering intersubjective meanings. Second, constructivists generally draw descriptive inferences by means of traditional quantitative and qualitative methods and draw causal or constitutive inferences by means of historical narratives. Wendt (1999: 86), for example, argues that constitutive theories are explanatory and not merely descriptive (but see King et al., 1994). Constructivists generally believe that the barriers to true knowledge are posed not only by poor or defective methods, but also by the nature of social reality, which is at least partly indeterminate and contingent. Constructivist explanations, therefore, usually include reconstructed narratives that–because the manner in which social facts become established in the social world is relevant to the way in which they exert their influence (Adler, 1997: 339)–are as much about partly indeterminate processes as they are about partly determinate outcomes. The use of narratives and other interpretive methods, however, does not mean that all statements or all variables have the same weight; rather, such methods are used to uncover the validity of statements (Morrow, 1994) and to reveal social structures, social mechanisms and empirical regularities.

Constructivism’s ‘Added Value’ and Substantive Empirical Contributions

By added value, I mean substantial improvements in the understanding of some of the conceptual building blocks of IR theory, especially knowledge, change, social communication, rationality, language and power.

(1) Constructivism considers intersubjective knowledge and ideas to have constitutive effects on social reality and its evolution. When drawn upon by individuals, the rules, norms and cause–effect understandings that make material objects meaningful become the source of people’s reasons, interests and intentional acts; when institutionalized, they become the source of international practices. Constructivism’s added value, therefore, is that it helps explain why people converge around specific norms, identities and cause–effect understandings, and thus where interests come from (Adler, 1987, 1991; Finnemore, 1996). Moreover, it puts to rest the naïve notion that either material objects or ‘ideas’–but not both–constitute interests. Instead, constructivism advances the notion that interests are ideas; that is, they are ontologically intersubjective but epistemologically objective interpretations about, and for, the material world (Weldes, 1996). This means that interests cannot be mechanically deduced from international anarchy and the distribution of material resources. As Wendt (1999) has shown, international anarchy may be consistent with a state of permanent war, a state of calculated partial cooperation, and a state of more or less permanent peace.

(2) Contrary to the argument that constructivism is agnostic about change in world politics (Hopf, 1998: 180), it may be only a slight exaggeration to say that if constructivism is about anything, it is about change. For rather than using history as a descriptive method, constructivism has history ‘built in’ as part of theories. Historicity, therefore, shows up as part of the contexts that make possible social reality, the path-dependent processes involving structural and agent change, and the mechanisms involved in the explanation of change.

Constructivism’s added value, therefore, is to take change less as the alteration in the positions of material things than as the emergence of new constitutive rules (Ruggie, 1998b), the evolution and transformation of new social structures (Dessler, 1989; Koslowski and Kratochwil, 1995), and the agent-related origins of social processes. In reference to agency, constructivism has generated theoretical and empirical studies about, for example, policy entrepreneurs (Checkel, 1998), epistemic communities (Adler and Haas, 1992), and transnational advocacy networks (Keck and Sikkink, 1998). Regarding the mechanisms of change, some constructivists emphasize collective learning, cognitive evolution, epistemic change and the ‘life cycles of norms,’ all of which involve the institutionalization of people’s novel knowledge, practices and discourses (Adler, 1991; Finnemore and Sikkink, 1998; Haas, 1990; Ruggie, 1993b). Critical constructivists (Cox, 1986; Linklater, 1998), instead, take change not as something to be passively observed and explained, but as something that may occur as a result of reflexive analysis. Thus critical constructivism points to potential alternatives to prevailing structures (Hopf, 1998: 180).

(3) Social communication is another important added value of contemporary constructivism. Not only do collective understandings diffuse across time and place by means of it; it also enables agents to fix the meanings of material reality (Luhmann, 1989: 17). When fixing meanings, agents select from ‘a “horizon” of possibilities’ (Mingers, 1995: 157). In so doing they contribute to the institutionalization of practices and consequently to the unintentional survival of social structures. We also may find added value in constructivist theories that build on Habermas’s (1984, 1987) theory of ‘communicative action.’ The main idea behind this theory is that social actors do not bargain to achieve the utilities they expect–as rational choice theory maintains. Rather, they engage in a discourse that helps demonstrate the validity of their arguments; this discourse in turn promotes collective understandings (Risse, 2000). So, rather than studying instrumental bargaining and choice, constructivism focuses on the effects of social communication on social relations; for example, how debate and persuasion help promote shared understandings. Recently, some rationalists have started to address social communication issues such as socialization, pointing to the rhetorical and thus instrumental nature of agents’ actions (e.g., Schimmelfennig, 2000). By contributing to a better understanding of the micro-foundations of social construction, they are adding to the pool of knowledge from which some constructivists also draw.

(4) The relationship among acting, communication and rationality is critical for constructivists. Contrary to common belief, constructivists consider rationality and reason to be of critical importance for their explanations. Constructivists, however, cannot accept the notion that rationality means only instrumental rationality. Thus they advance the notion of practical or communicative rationality, which, though sometimes calculating and choice-related, is also based on practical reason, is sensitive and contingent to historical, social and normative contexts, and emphasizes the communicative and persuasion logic in social theory. Practical rationality is one of constructivism’s most important recent contributions to IR theory and research. When scholars emphasize the role of norms, the logic that stands in contradistinction to rational choice is that of ‘appropriateness’ (Finnemore, 1996; March and Olsen, 1998). Accordingly, agents do not choose between the most efficient alternative, but ‘follow rules that associate particular identities to particular situations, approaching individual opportunities for action by assessing similarities between current identities and choice dilemmas and more general concepts of self and situations’ (March and Olsen, 1998: 951). When scholars stress social communication, Habermas’s notion of ‘communicative rationality’ suggests itself.

Regardless of the route we follow to characterize practical rationality, however, what stands out is the significance of intersubjective understandings; the capacity for rational thought and behaviour is above all a background capacity (Bourdieu, 1977; Searle, 1998). Rationality lies less in choosing instrumentally on the basis of true theories than in behaving in ways that stand to reason, given people’s background expectations and dispositions. It follows, then, that, because instrumental action is prompted by expectations and intentions, which are drawn from previously constituted social structures, constructivism subsumes rational choice under its more general principles. In other words, rational actors live and act in a socially constructed world and instrumental action takes place as a backdrop, not only to the knowledge that individuals share qua individuals, but also to all institutionalized knowledge (such as norms). Although very few rationalists accept this argument, some of them (Bates et al., 1998a; Fearon and Laitin, 2000, Zürn, 1998) have become more sensitive to the effect of discourses, narratives, identities and norms on rational choice. This demonstrates not only that constructivism has already left its mark on the discipline, but also that rationalists and constructivists are now in a position to cooperate on IR theoretical work.

(5) Social communication and practical rationality depend on language, which is the vehicle for the diffusion and institutionalization of ideas, a necessary condition for the persistence in time of institutionalized practices, and a mechanism for the construction of social reality. Constructivism’s added value, therefore, consists in spelling out the role of language in social life. To begin with, language is the medium for the construction of intersubjective meanings. The sense of right, obligation and duty that political actors borrow from social structures depends on language that is oriented toward collective purposes. Second, ‘speech acts’ (e.g., ‘this meeting is adjourned’) have an ‘illocutionary’ dimension (‘doing something by saying something’); hence, not only do they describe a reality, they also construct it (Kratochwil, 1989: 8). Third, and moving toward constructivism’s ‘strong programme,’ discourse–in Foucault’s (1980) celebrated interpretation–is power, in the sense that ‘it makes us understand certain problems in certain ways, and pose questions accordingly. It thereby limits the range of alternative policy options, and it enables us to take on others’ (Diez, 1999: 603). Finally, if we start from the premise that language expressions represent a potential for new constitutions of reality (Derrida, 1978; Diez, 1999: 607), then discourse is also a source of change.

(6) Stressing material capabilities and overlooking the intersubjective dimension of power, traditional IR theoretical perspectives lost touch with some of the main forms of power, such as speech acts (Onuf, 1998a), hegemonic discourses (Cox, 1986), dominant normative interpretations and identities (Checkel, 2001), and moral authority (Hall, 1999). When, for example, someone uses guns and tanks, or makes a threat, it usually means that the ability to impose meanings, status, or functions on physical objects by collective agreement has already failed. As the case of the disintegration of the Soviet Union shows, guns and tanks were of no use when the regime’s legitimacy and the system of collective understandings about identity, status and functions collapsed (Searle, 1995: 92). The imposition of meanings on the material world is one of the ultimate forms of power, and thus is where constructivism’s added value with regard to power lies. The added value that results from interpreting power from a constructivist perspective also includes what Hacking (1999) has called ‘making people’; in other words, labelling people in such a way that they change their identity, status and functions in reaction to the labelling. It also includes the power branded by social groups ‘to provide the authoritative vision of the world’ (Guzzini, 2000: 172), as well as Gramscian hegemonic power, which brings the interests of powerful groups into harmony with weaker groups and incorporates these interests into ‘an ideology expressed in universal terms’ (Cox, 1983: 168). Finally, it lies in the discursive ability to force one meaning of the world onto others (Bially, 1998).

Substantive empirical contributions Contrary to the still-common belief that constructivists avoid empirical research, there is a growing empirical constructivist literature about, for example, norms, identity, sovereignty, institutionalization and international governance, which has already made a substantive mark on the field.

(1) Norms constitute social identities and give national interests their content and meaning. Constructivist research grounds the notion that how people apply norms to classify the world is not irrelevant to the manner in which world politics unfolds. For example, Katzenstein and his associates (Katzenstein, 1996) have persuasively shown that states face security choices, and act upon them, not only in the context of their physical capabilities but also on the basis of normative understandings. Klotz (1995), in turn, has shown that the end of the apartheid regime in South Africa became possible because of the emergence and institutionalization of a global norm of racial equality. Moreover, according to Finnemore (1996), international organizations ‘teach’ or help diffuse norms and thereby help constitute the national interests of states that adopt these norms.

(2) Identity lies at the core of national and transnational interests. Consequently it is crucial for an understanding of international behaviour, practices, institutions and change. It is just as important for an understanding of international conflict and war as for an understanding of international cooperation. Constructivism’s critics argue that though it may be true that identity lies at the core of people’s interests, identities do not change as often as constructivists say they do; hence there is no reason not to assume that interests are fixed (Mearsheimer, 1994). Adler and Barnett (1998), however, have shown that the ‘we feeling’ or identities of national groups may expand across national borders and lead to the development of security communities. Moreover, if identities are fixed, it would be difficult to explain the case of post-war Germany, which came to include Europe as part of its identity (Banchoff, 1999). And while the Middle East seems to be the area where realist thinking would take us the farthest, Barnett (1998: 15) has shown that ‘Arab politics can be understood as a series of dialogues concerning the relationship between identities, norms, [and] regional order.’ Lynch (1999), in turn, has elucidated the notion that changes in Jordan’s foreign policy are foremost changes in Jordan’s identity.

(3) Constructivism has made important contributions to the understanding of sovereignty (e.g., Bartelson, 1995; Biersteker and Weber, 1996; Walker, 1993). For example, constructivists have shown that the components of state sovereignty, such as territory, authority and national identity, are not fixed, but evolve with changing practices (Biersteker and Weber, 1996: 15). Building on Ruggie’s insight (1983, 1993b) about the transient nature of the Westphalian international system, constructivists have also been drawing attention to alternative constitutive norms–for example, human rights (Risse et al., 1999)–around which future systems might develop. Hall (1999) traced the social construction of national sovereignty in recent centuries and demonstrated its differential impact on interests (and thus behaviours), and, more generally, on international order and international systems. Moreover, constructivist analysis of sovereignty has shown how people collectively draw the boundaries between ‘inside’ and ‘outside’ (Walker, 1993) and how these boundaries are ‘produced, reproduced, legitimated, contested, changed and naturalized’ (Thompson, 1994: 13).

(4) Constructivists understand institutions as reified sets of intersubjective constitutive and regulative rules that, in addition to helping coordinate and pattern behaviour and channel it in one direction rather than another (Ruggie, 1998a: 54), also help establish new collective identities and shared interests and practices. For example, based on case studies of ancient Greece, Renaissance Italy, absolutist Europe and the present international system, Reus-Smit (1999) has shown that societies are constituted by ‘deep institutions’ that result from beliefs about the moral purpose of the state, sovereignty, and the norm of procedural justice. Adler and Barnett (1998), Finnemore (1996) and Keck and Sikkink (1998) have shown that socialization, learning and emulation may enable international institutions to establish, articulate and transmit norms across nations, to define what constitutes legitimate behaviour, and to shape the identities of their members. Constructivism, however, has yet to provide more convincing answers to two questions: first, why certain ideas are institutionalized and others are not (Kowert and Legro, 1996; Legro, 2000); second, what keeps international institutions stable.

(5) Empirical constructivist work about new actors on the global scene–such as epistemic communities (Adler, 1992; P. Haas, 1990), NGOs and transnational advocacy networks (Keck and Sikkink, 1998), and moral communities (Linklater, 1998)–is enriching our understanding of international governance. Uncovering previously unrecognized ‘chunks’ of international and transnational social reality that occupy previously unidentified transnational spaces, it suggests diverse ways of organizing the study of international political reality, which transcend IR’s ‘domestic analogy.’

Debates with in Constructivism

The most salient and sustained constructivist debate in IR, which blends ontological, epistemological, and theoretical issues, has revolved around the ‘agent–structure’ problem. Three other Epistemological debates have dealt with: (1) constitutive vs. causal theory, (2) explanatory vs. emancipatory cognitive interests, and (3) modernism vs. postmodernism. In addition, three debates about IR theory deal with: (1) the nature of agency in IR, (2) the role of rationality in the construction of social reality; and (3) liberal constructivism and its discontents. A methodological debate, which is imperative, has yet to begin.

The Agent–Structure Debate

The agent–structure debate focuses on the nature of international reality; more precisely, whether what exists in IR, and the explanation for it, should revolve around actors, structures, or both. Kenneth Waltz’s structural theory of IR (1979) became the debate’s point of entry for early constructivists, especially Wendt (1987). Wendt argued that Waltz’s international structure and system, being creations of states, can only constrain state agency, but cannot generate state agents themselves; this argument meant that Waltz was ‘not at all the structuralist he claimed, but, to the contrary, an ontological individualist’ (Gould, 1998: 84). By contrast, claimed Wendt, Immanuel Wallerstein’s (1974) ‘world systems’ are all structure and no agency, productive of states, which have no productive powers themselves. In order to avoid having to choose between agency and structure and to make it possible to deal with the nature of their relationship, Wendt imported Giddens’s (1984) social structuration theory and Bhaskar’s (1979) critical realist theory to IR. He argued that just ‘as social structures are ontologically dependent upon and therefore constituted by the practices and self-understandings of agents, the causal powers and interests of those agents, in their own turn, are constituted and therefore explained by structures’ (1979: 359). Similarly, Dessler (1989: 452) challenged Waltz’s ‘positional’ model with a ‘transformational’ model of international structure, in which ‘all social action presupposes social structure, and vice versa.’ Wendt’s and Dessler’s work on the agent–structure problem was a crucial moment for constructivism, but also the basis of a new agent–structure debate within constructivism itself.

Reacting to Wendt’s claims, Hollis and Smith (1990: 1) used an epistemologically driven approach to the agent–structure problem, along with a ‘level of analysis’ argument, to suggest that, as far as the social world is concerned, there are always two stories to be told. One story is about ‘explaining’ from the perspective of an outsider or observer, as in the naturalist approach to science. The other story is about ‘understanding’–a hermeneutic inside view that involves getting to ‘the point’ or the meaning of things. This notion led to several interconnected sub-debates, which were framed by a series of articles by Wendt (1991, 1992a, 1998) and Hollis and Smith (1991, 1992, 1994, 1996), and which I distinguish for analytical purposes only.

First, there is the question whether, as Hollis and Smith argue, we can explain social phenomena from the ‘outside’ and understand people’s meanings from the ‘inside,’ or whether, as Wendt (1991: 391), Carlsnaes (1994) and Dessler (1999) maintain, we may be able to explain social phenomena, even when studying people’s meanings from the ‘inside.’ What divides these scholars is the question of whether one should start from ontology or from epistemology. Scientific realists, such as Carlsnaes, Dessler and Wendt, first seek to identify social essences and only then to explain and/or understand the social world. They thus argue that it is possible to explain and understand from the inside or from the outside. Hollis and Smith (1996: 111, 113), however, hold that one can make ontological judgements only after deciding ‘what kinds of criteria allows us to judge what kinds of things exist in the social world. … Ontological statements without an epistemological warrant are mere dogma.’ Second, the above controversy directly impinged on Hollis and Smith’s (1990) ‘level of analysis’ argument, according to which, at every level of analysis, one can explain or understand IR by proceeding either from system to unit or from unit to system. In their view, at one level of analysis the international system is the ‘whole’ or structure and the state is the unit or agent. At a lower level of analysis, however, the state is structure and the bureaucracy is the agent. The same is true with the bureaucracy vs. the individual. Wendt (1991) retorted that the level of analysis argument, as originally formulated by David Singer (1961), is best suited for ‘assessing the relative importance of causal factors at different levels of aggregation in explaining the behaviour of a given unit of analysis’ (Wendt, 1991: 387). But in Hollis and Smith’s account, ‘the phenomenon to be explained changes; first it is the behaviour of state actors, then the behaviour of the international system. This is a problem of ontology: of whether the properties or behaviour of units at one level of analysis can be reduced to those of another’ (Wendt, 1991: 388). To avoid confusion, Wendt (1992a: 185) suggested reserving the level-of-analysis discourse ‘for questions about what drives the behaviour of exogenously given actors, and agent-structure talk for questions about what constitutes the properties of those actors in the first place.’ For Hollis and Smith (1992: 188), however, the question of ontology emerges at any level of analysis. Thus a distinction between explaining unit behaviour and identifying its properties is empty (Patomäki, 1996: 107).

The third question that Wendt (1991, 1992a) and Hollis and Smith (1991, 1994, 1996) debated directly followed from the previous question. Arguing that there can be only one kind of international system, that is, Waltz’s, Hollis and Smith strongly objected to Wendt’s portrait of Waltz as an individualist. Wendt, for whom a systemic theory must account for how states are constituted in the course of their interactions, and thus also for how identities and interests are socially constructed, retorted that the fact that Waltz’s system can only constrain behaviour is indicative of an individualist approach. According to Wendt (1992a: 183), therefore, it was Waltz’s de facto individualism that allowed Hollis and Smith ‘to reduce the question of systemic causation to the question of whether the international system conditions the foreign policy behaviour of states’–in other words, ‘to reduce the agent–structure problem to one of levels of analysis.’

If Wendt thought that Hollis and Smith (1990) conflated the agent–structure problem and the levels of analysis problem, Carlsnaes thought that Wendt conflated agent and structure in ways that made it difficult to do empirical research. Carlsnaes’s important contribution was to invoke Margaret Archer’s (1989) early morphogenesis theory in order to introduce the time dimension to the agent–structure debate. He argued, with special reference to foreign policy, that what is needed to explain an action is ‘indicating the reciprocal interplay over time–in terms of developmental patterns or cycles–that exists between structure and action’ (Carlsnaes, 1992: 264). Hollis and Smith (1994: 244) were not persuaded; they retorted that the addition of the time dimension does little to solve the agent–structure ontological problem. On the contrary, it may actually make things worse, because morphogenesis means ‘treating agents and structures as if they take turns affecting the social world’ (Gould, 1998: 92); furthermore, it is not clear how adding the time dimension helps judge rival accounts of agents and structures. Replying to Hollis and Smith, Carlsnaes (1994: 280–4) suggested that a single integrative conceptual framework, which involves ‘intentional,’ ‘dispositional’ and ‘structural’ levels of explanation, may go a long way towards resolving the agent–structure debate. Structures, ‘cognitively mediated by the actors in question rather than affecting policy actions directly,’ may thus be part of causal interpretive explanations.

In recent years, Patomäki (1996), Wight (1999), and Patomäki and Wight (2000) entered the agent–structure debate with a critical realist argument based on the work of Archer (1995) and Bhaskar (1979). For example, Patomäki (1996: 108) took explanation as a mode of interpretation, reasons as causes, and actors and regulative and constitutive rules as involved in the production of reality. Wight (1999), criticizing Wendt’s (1992b) identification of states as agents, argued that only individuals can be agents and suggested a multilayered definition of agency, involving individual attributes, sociocultural systems and roles. This realist position, however, did not sit well with Doty (1997; cf. Suganami, 1999); making a rare contribution to the agent–structure debate from radical constructivism, she advanced the post-structuralist argument that what matters is neither structures nor agents, but the role of discursive practices.

I doubt whether the agent–structure debate will ever be fully resolved. Because of the debate, however, we now have a much better understanding of the metaphysical and social-theory foundations of the relationship between agents and structures. While constructivists have disagreed markedly about agent–structure, there is much more in common in their work than they are aware of or care to acknowledge. The agent–structure debate can thus profit from some ‘consolidation,’ by which I mean concentrating on the consensus already achieved and that still can be achieved, and then turning our efforts to translate the agent–structure metaphysical and social theory positions into theoretical and empirical propositions.

In particular, theoretical and empirical discussions of how social structures act on the subjective level and how ideas held by individuals become institutionalized and taken for granted seem to be a good place to restart the debate. Constructivists have started to pay attention to the micro-foundations of intersubjective phenomena and to the macro-foundations of reasoned acts and are beginning to search for mechanisms that link them. These mechanisms include ideational diffusion and learning (e.g., Adler, 1992; Checkel, 2001; Finnemore and Sikkink, 1998; Wendt, 1999), socialization (e.g., Checkel, 1999; Johnston, 2001; Schimmelfennig, 2000; Wendt, 1999), social communication and persuasion (e.g., Checkel, 1998; Lynch, 1999; Risse, 2000), and institutionalization (Adler and Haas, 1992; Legro, 2000; Ruggie, 1993a).

Searching for the mechanisms that can make social action more intelligible and make the agent–structure problem more manageable, however, will require learning more about the development of communities of shared meaning, discourse and practice, and focusing on the dynamic aspects of agent–structure. Learning processes, for example, occur in people’s heads, but their outcomes exist in the intersubjective world. Thus, only when practices change as a result of a re-conceptualization of reality, can multiple interacting actors and future actors draw upon these understandings and, thus, learn the same or similar lessons over time. And when people’s collective knowledge is institutionalized, it becomes a building block of their reasons and the spring of subsequent social action.

The Epistemological Debates

A debate about causal vs. constitutive explanations, which deals primarily with the kind of knowledge that constitution entails, has been brewing mainly between Wendt (1998, 1999, 2000) and his critics. In Wendt’s view, causal theories ‘answer questions of the form “why?” and, in some cases, “how?”’; whereas constitutive theories ‘account for the properties of things by reference to the structures in virtue of which they exist.’ Thus, for example, the factors that constituted the Cold War–which ‘do not exist apart from a Cold War, nor do they precede it in time’ (Wendt, 1998: 104–6)–are not the same as its causes. Echoing linguistic and critical constructivist approaches, Smith (2000: 157) claims that Wendt’s view of constitution looks very much like a causal theory, ‘or at least … a form of theorizing that leads to, or is prior to, causal explanation.’ It thus differs from the ‘understanding’ hermeneutic-like view that is dominant in the interpretive social sciences. For Smith (2000: 157–8), the problem with Wendt’s rendition of causal theory and constitutive theory is his Cartesian separation between ideational and material forces, according to which ‘at some level material forces are constituted independently of society, and affect society in a causal way’ (Wendt, 1999: 111). This is far from what Smith considers to be a hermeneutic strategy of rule- and language-constituting action.

Some of the differences between Wendt and his critics may be reconciled by pragmatic realism. Contra Smith, we need a realist ontology ‘about what it is that brings about changes in the world outside of the texts we are writing ourselves’ (Patomäki and Wight, 2000: 229). Contra Wendt (1999), however, we need a pragmatic epistemology that, without neglecting human agents, does not separate between the material (causation) and social structure (constitution). Pragmatic realism consolidates the two arguments by taking social constitution as the dynamic collective attachment of meaning, function and value to material reality (Searle, 1995), and causal explanation as the identification of the mechanisms that are involved in the social constitution of institutions and practices. Expectations, for example, are intentional, and are thus part of the causal relation between mind and the world (Searle, 1998: 100–7). On the other hand, expectations are simultaneously drawn upon the background of intersubjective dispositions that constitute but do not determine human reasons.

The cognitive interest debate within constructivism has pitted modernist constructivists, who believe that explaining social reality is the main goal of social knowledge, and critical theory scholars, who believe that the main goal of social knowledge is emancipation from oppressive structures. Critical theorists say that there can be no explanations of the world as it is, if only because there is no world until we explain it. Thus constructivists should take a normative and ethical stand (Inayatullah and Blaney, 1996) and use theory as a tool for improving the world; for example, advancing democratic transnational community (Linklater, 1998), empowering women (Enloe, 1990; Peterson, 1992; Tickner, 1992), and redefining security (Williams and Krause, 1997: xiv).

Modernist constructivists do not disagree with their critical counterparts about the occasional need to take a critical position against the social world and pursue normative agendas, or about the capacity of knowledge ‘carriers’ to help bring about changes in the social world. But they also believe that the best way to advance normative goals is not to take theory as an instrument for the emancipation of ideologically chosen underclasses but instead to produce systematic knowledge, including about how knowledge and political power interact. Critical constructivists retort that, in their zeal to provide contingent explanations of social reality, modernist constructivists legitimize the existing situation. Thus they really are not constructivists at all, but liberals and positivists in disguise, who stick close to the precepts of rationalist theories (Ashley, 1996; Campbell, 1996; George, 1994).

Price and Reus-Smit (1997), two modernist constructivists with critical leanings, have recently attempted to narrow the rift between the two types of constructivism; they maintain that modernist constructivists share some of the normative concerns held by critical theorists and that, in fundamental ways, modernist constructivists have been advancing a critical agenda. Price and Reus-Smit suggest, therefore, that only through a dialogue between normative arguments and empirically informed accounts can we arrive at better and more ethical practices. Their point is that improving the world requires bringing the two cognitive interests together; without explanation there can be no emancipation.

This is certainly the case with regard to the relationship between social knowledge and political practice. We need theories about the emergence and consolidation of practices that enhance human interests across national borders, including about the manner in which theoretical knowledge intervenes in struggles over meaning and affects these processes (Adler, 1997). Constructivist theory, therefore, can be both ‘critical’ and ‘problem-solving,’ in Robert Cox’s (1986: 208–9) sense.

Finally, an attitude of mutual disengagement and benign neglect (rather than a debate) characterizes the relations between constructivists and post modernists. They differ about (a) the status of material reality, (b) the ontological status of unobservable mechanisms, (c) agency and especially reason, and (d) the notion that a social science, separate from the other sciences, is possible. As a consequence, constructivists have taken postmodernism to lie outside constructivism’s ‘middle ground’ (Adler, 1997). Postmodernists, in turn, tend to regard constructivists as positivists in disguise, aiming to take interpretive action out of postmodernism. Do these differences mean that postmodernists are ‘inside’ constructivism or ‘outside’ it? The question, of course, indicates that we are dealing with a social construct. Until recently, postmodernists have explicitly chosen to remain on the outside. Calling themselves ‘dissidents’ (Ashley and Walker, 1990; George, 1994), they carried over to IR their deep suspicions about anything that looks like discipline and foundation and thus divorced themselves from other streams of constructivism. The ‘middle ground’ thesis, therefore, rather than aiming to exclude postmodernists (Milliken, 1999: 227), echoed the fact that postmodernists explicitly and self-consciously placed themselves beyond what Waever (1996: 169) called ‘the boundary of negativity.’ There is no ‘essential’ reason, however, why constructivists and postmodernists cannot hold a fruitful and constructive debate, aiming to achieve not supremacy of one approach over the other, but mutual learning.

Theoretical Debates

The constructivist debate over the nature of agency in IR is about whether constructivists, following Wendt (1999), should theorize about agency as the attribute of states, or whether, as other constructivists say, they should open constructivism to domestic politics, non-state actors, and the possibility of state transcendence. Wendt (1999, 2000) defends his decision to focus on the state, because his theory is about the inter-state system and states possess a monopoly on the use of force, which they can use on behalf of individuals, domestic organizations and governments. Although, he admits, non-state actors increasingly play critical roles in world politics, these roles are important only through state action. He also believes in the progressive nature of the state. Finally, Wendt holds that states are structures that exhibit macro-level regularities and these, although dependent on individuals’ beliefs, are not explained by them.

Wendt’s state-orientation has been sharply criticized within the constructivist camp. To begin with, his critics have argued that although constructivism is supposed to open structures to different constructions and to different understandings of world politics, Wendt’s theory, like Waltz’s, is conservative and consecrates the existing inter-state system (Inayatullah and Blaney, 1996). Moreover, they say, Wendt’s theory should have assigned equal status to agents and structures, thus permitting the study of the domestic societal manifestations of collective identities (Hall, 1999: 27). Because Wendt brackets the domestic sources of state identity, however, he cannot explain the rise and decline of international societies (Reus-Smit, 1999: 166). Wendt may be able to explain change within systems, but not change between them.

Wendt’s reification of the state has also been faulted as being unhelpful for studying, for example, transnational communities (Koslowski, 1999) and security communities (Cederman, 1997). His critics add that it can be shown empirically that subjects who act in the name of the state are aware that it is not a unitary state (Doty, 2000: 139) and that ontologically, as Bhaskar (1979) has shown, only individuals can express agency (Wight, 1999: 127). Thus, they say, although Wendt may have articulated a ‘via media’ with regard to epistemology, mainly because of his reification of the state and his almost exclusive reliance on social structures for explanation, that middle path eluded him when it came to ontology.

Wendt’s (2000: 175) reply to his critics is simple. ‘It all depends on the question one is asking. Against a book on the states system, therefore, calls to “stop reifying the state!” should be seen really as calls to “change the subject!” I am all for that, but it complements the systemic question, rather than replaces it.’ Wendt is right, as far as his theory is concerned. But, one should ask, why did Wendt settle for such a limiting constructivist theory? As a constructivist, Wendt should have been aware that constructivist theories ought to leave room for new and unexpected structural possibilities. Instead, he offers a theory and a portrait of agency and the state that locks in politics as the study of inter-state relations and ultimately gives up on bringing into the theory the ultimate constructor of worlds–by which I mean the thinking, often reasonable, sometimes surprising, and even at times creative human individual.

Within constructivism, a dialogue that may soon turn into a full debate has been taking place about how to approach rationality. In the background is the increasing realization that constructivism and rationalism are complementary rather than contradictory. Three factors catalyzed the dialogue. First was a debate that took place in the 1990s between rationalists and constructivists in the pages of theZeitschrift für Internationale Bezienhungen (ZIB, Journal of International Relations). A second factor was the lead article by Peter Katzenstein, Robert Keohane and Stephen Krasner (1998: 680) in the fiftieth anniversary issue of International Organization, with its thesis that ‘rationalism and constructivism are generic theoretical orientations that are complementary on some crucial points.’ For example, constructivism may contribute to a better understanding of what rationalists call ‘common knowledge’ and of the role of norms in situations of multiple equilibria. The article also envisions a division of labour whereas constructivists explain where alternatives come from and rationalists explain instrumental choice (Katzenstein et al., 1998: 680-2). The third factor has been the growing trend in the rationalist camp to develop theories of institutional behaviour (Young, 1998) and rational choice-based narratives (Bates, et al., 1998b).

The emerging debate within constructivism is thus whether and how to reconcile rationalism and constructivism, and on whose terms. I can identify four (preliminary and still vague) constructivist responses to the challenge to integrate rationalist and constructivist approaches. The first ‘response’ is no response at all, but an unwillingness to contemplate or even talk about the possibility that rationalism and constructivism may be compatible. This response is common mainly among radical constructivists. Risse (2000) and a small but growing group of constructivists provide a second response, namely, that constructivists should confront the ‘logic of consequentialism’ not only with the normative ‘logic of appropriateness’ but also by adopting Habermas’s (1996) concept of ‘communicative rationality’ and the ‘logic of arguing.’

Checkel (2001) suggests a third response, which is really about scope: in some circumstances, a rational approach is called for; in other circumstances, a constructivist approach is more suitable. Following in the footsteps of Katzenstein, Keohane and Krasner (1998), who take constructivism as a supplement to rational choice, Finnemore and Sikkink (1998: 911) offered a fourth response; namely, a ‘staged analysis’ that ‘could run either way: one could model rational choice as producing social knowledge as easily as one could model social context as a background for rational choice, depending on the empirical question being researched.’ The second, third and fourth responses are consistent with the view that a synthesis between rationalism and constructivism may be possible. A real synthesis, in my view, would integrate rationalism and constructivism into a theory that ultimately transcends both.

Another dialogue, which pits liberal vs. non-liberal constructivist approaches, is just getting under way. In general, constructivist scholars have been critical regarding liberal approaches. This includes postmodernists (Ashley, 1987; Latham, 1995), critical constructivists (e.g., Cox, 1986), and feminist constructivists (Runyan and Peterson, 1991). Some modernist constructivists, however, while distancing themselves from liberal explanations (Moravcsik, 1999) and neoliberal explanations (Keohane, 1984), have nevertheless explicitly followed liberal research agendas, for example, about the democratic peace (e.g., Risse-Kappen, 1995). A possible constructivist debate of the future may thus be about making liberalism more compatible with constructivism without undercutting the latter’s meta-theoretical basis. Another possible debate would follow Onuf’s (1998b) book The Republican Legacy in International Thought, and address constructivism’s Kantian and Weberian roots. Still a third possibility is a debate about the liberal agendas that most modernist constructivists follow. In this generally West-oriented discipline such a debate may engender an increased attention to the Third World, its culture, problems and agendas, which are frequently not liberal and may even be anti-liberal.

A Blueprint for Action

In this concluding section I offer a blueprint for action within the constructivist camp. I am less interested in chartering an agenda for specific IR theory and empirical research than in emphasizing those issues that, if constructivism is to become firmly established in IR, need to be taken into account and dealt with.

(1) Methodology is the major missing link in constructivist theory and research (but see Alker, 1996; Dessler, 1999; Friedman and Starr, 1997). It would be only a slight exaggeration to say that whether constructivism ultimately has a profound effect on the IR discipline will depend on the development of a coherent constructivist methodological base that suggests a practical alternative to imitating the physical sciences. The explanatory endeavour, I believe, should revolve around the historical reconstruction of social facts (Dessler, 1999), on the basis of interpretive narratives, practices and discourses. A coherent constructivist methodological approach also means approaching research less as a predictive enterprise than as an effort to explain how past and present events, practices and interests became possible and why they occurred in time and space the way they did.

This does not mean, however, that constructivism should not be ‘forward reasoning.’ As eloquently proposed by Bernstein et al. (2000), an alternative to deductive-nomological methodology may be offered by an evolutionary approach, in which we read into narratives the ‘so-called lessons of the past’ (p. 50) and, by building scenarios, learn about the likely but not certain path that reality may take. This is appealing for constructivists, because it makes it possible to deal both practically and empirically with what Hacking calls ‘looping effects’ (Hacking, 1999) and with the reflexivity of human knowledge, including how world events affect our knowledge, which then affects us and the world.

(2) It is time for constructivists to tone down epistemological and ontological debates and concentrate more directly on building constructivist IR theory. This does not means neglecting meta-theory and social theory, but concentrating on them only when they are necessary for building substantive IR constructivist theories. In spite of the differences and debates within constructivism, there exists enough common ground among most types to permit dealing with meta-theoretical issues on the basis of IR theoretical and empirical questions.

(3) We should redouble our efforts to develop the micro-foundations of constructivist theory. In recent years, perhaps because most constructivists have relied on structural/functional sociological theories (Sterling-Folker, 2000), there has been a call to ‘go micro’ (Checkel, 1998). For example, Wendt (1999) suggests relying on cognitive theory; Schimmelfennig (2000), on rational choice; Checkel (2001), on rational choice and social psychology, mainly social learning; and Johnston (2001), on social psychology, mainly status theory. I believe that all these paths should be explored further and assessed on their merits. At the same time, the micro-foundations of constructivism and thus constructivist theories of action should not be disconnected from the social structures that individual agents draw upon for their intentional acts. To put this another way, we do not need a structural theory over here and a micro theory over there. Rather, we need a mechanism or, if possible, a theory whose micro-foundations are clearly understood by us. The point is that we first need to know what type of overall social mechanism or theory of institutionalized international behaviour this micro theory is part of.

(4) Alex Wendt’s (1994) ‘corporate identity’ black box should be opened up. In fact, this has started to happen, as constructivist scholars increasingly pay attention to the domestic determinants of change (Reus-Smit, 1999; Risse et al., 1999) and to the domestic impact of international norms (Checkel, 1997; Cortell and Davis, 2000; Risse et al. 1999). This agenda should be broadened and deepened, however. First, the domestic arena, which to liberals (Moravcsik, 1998) is the mechanical bureaucratic backdrop of a rational choice theory of inter-state bargaining, to constructivists is the place where national preferences are born, and international practices are produced, reproduced and transformed. To constructivists, therefore, the socio-cognitive domestic ‘sources of state preferences’ (Moravcsik, 1998) are part of the mechanisms that connect agents and structures in dynamic ways. These mechanisms may help explain individuals’ socially constructed reasons in domestic political settings, which come to constitute social structures at later times and in different places. They also may help explain the domestic selection processes that partly determine what and whose collective understandings, including norms and causal beliefs, become the ‘national interest.’

Second, Robert Putnam’s concept of ‘two-level-games’–according to which national leaders play strategic material power games, albeit domestically and internationally constrained (R. Putnam, 1988)–lends itself to a constructivist reading of conceptual games. In conceptual games, not only leaders of states, but also other state and non-state actors bargain about who gets to impose meanings on material reality and thus to socially construct the situation in their own image. A research programme on conceptual games will help constructivists explain how power affects the domestic sources of corporate identity. Third, constructivists should pay more attention to the role of domestic practices in the constitution of regional, international and even global social structures (Lumsdaine, 1993; Nadelmann, 1990). Finally, constructivists must bring the individual back in–for example, studying how individuals purposefully use social capital and carefully chosen words to legitimize or delegitimize opponents and sometimes entire populations. In fact, this kind of social construction, especially in the information age, is a large part of what politics is all about.

(5) Constructivists will need to face the problem of blending normative theory with explanatory theory. Here, I believe, we have much to learn from the English School, whose ‘international society approach’ is simultaneously historical, normative and systematic/analytical. One of the English School’s key points is that normative and critical goals are part of the frame of systematic problems that scholars set themselves to explain.

(6) Constructivists should stress the practical and political consequences of their approach. First, constructivism can help sensitize practitioners to the sometimes crucial role that learning plays in international politics. For example, constructivism can contribute to a better understanding of the role of intersubjective meanings and identities in conflict prevention and resolution. Second, constructivism can make practitioners more sensitive to the functions, practices and discourses of transnational epistemic communities and transnational advocacy networks. Third, constructivism holds the potential to change how practitioners think about security, for example, the idea that people’s identities lie at the root of conflict and peace. Fourth, constructivism can be conducive to a better understanding of the design of international organizations, stressing that design may be a function not only of efficiency (Koremenos et al., 2001) but also of the way the designers intersubjectively interpret the organizations’ purposes and functions. Fifth, constructivism may help produce a new understanding of strategy and strategic relations. For example, insights about ‘communicative rationality’ may be used to reformulate our understanding of conventional and non-conventional strategy. Finally, a critical understanding of the role of socially constructed knowledge in the practical day-to-day construction of social reality–for example, via epistemic communities–may help advance practical and politically oriented understandings of sustainable development, globalization, arms control and human rights.

(7) There are several directions of interdisciplinary research that, if pursued, may help constructivists better to understand how social construction processes work. I am referring mainly to a cross-fertilization between constructivism, on the one hand, and international law (Dezalay and Garth, 1996), social psychology (Johnston, 1995), evolutionary theory (Adler, 1991; Bernstein et al., 2000), cognitive psychology (Wendt, 1999) and complexity theory (Alker, 1999; Hoffmann, 1999), on the other. Because I do not have enough space to refer in detail to these and other areas of collaboration, I will say a few words about the last two areas only.

Cognitive science can help develop constructivism’s micro-foundations. This does not mean reducing constructivism’s micro-foundations to information processing and AI (artificial intelligence), let alone neurons (Lycan, 1999). Cognitive psychology and its connections to analytical philosophy, however, may lead to a better understanding of reasoning processes, human dispositions, expectations, intentional acts, and the relationship between mind and language (Searle, 1995, 1998).

Complexity theory raises more difficult issues, if only because it is a relatively new field that still lacks a scientific consensus. Again, I do not suggest reducing social issues to the new kid on the physical and natural sciences’ block. Complexity theory, however, has developed some insights that can be particularly useful to constructivists, for example, about interactions that are interconnected in unanticipated and non-linear ways. Non-linearity is consistent with self-fulfilling and self-defeating prophecies and ‘looping effects.’ The idea that small and basically unpredictable changes can become large social transformations may help shed light on how to study amplification processes, such as the social diffusion of norms. Complexity theory may also illuminate the agent–structure problem. For example, the critique that structuration theory is circular is based on a linear approach to change. From a non-linear perspective, however, the possibility that one variable is ‘involved in causing change in another, while the other variable is similarly involved in causing change in the first’ (Brown, 1995: 7) makes perfect sense.

Constructivism has come a long way. It has become an alternative way of doing IR theory and research and has made a substantial contribution to the IR discipline. Constructivism’s common ground, however–the social construction of knowledge and the construction of social reality–which is often concealed by theoretical differences and debates, can and should be articulated more explicitly, not only through dialogue, but also by convergence around shared academic practices. Moreover, IR constructivism is at a preliminary stage only; much work still remains before it becomes a normal and taken-for-granted way of doing IR theory and research. For this to happen, however, constructivism must manage to bridge, first, metaphysics and social theory on the one hand, and IR theory research and methodology on the other, and then bridge IR theory and practice.