Analysing discourse as a causal mechanism

Introduction

This article seeks a way that discourse can be studied as a causal mechanism in the generation of the macro-social events studied by International Relations (IR) scholars. I do this because of dissatisfaction not only with how discourse is often studied in IR, but also because of the avoidance of discourse analysis in the study of empirical puzzles which clearly involve some significant impact being attributed to a particular discourse or discourses. Of course, to the extent that every social event or act involves some level of communication, anything can be analysed for the discourses that surround and give meaning to it. To do so for most IR scholars means to adopt a ‘discourse perspective’ — specified by the most popular discourse-analytic sub-field in IR, poststructuralist discourse theory (PDT) (Torfing, 2005: 3). Within PDT it is contended that to study discourse one must avoid any pretence to claims of having found some relatively vital causal relationship within a phenomenon, or any meaningful role for extra-discursive ‘reality’. It is this aspect of PDT that I wish to challenge, with the hope of opening up the study of discourse to scholars not willing to adopt a discourse perspective on the social world. To do so I propose a foundation, and some tools, for which discourse might be studied as but one causal thing among myriad possible others.

It may be questioned whether opening the study of discourse to scholars unwilling to adopt a PDT perspective on the social world is necessary, or whether it is even feasible. Maybe causal social science is adequate without integrating discourse within explanatory narratives? Or maybe attempting to do so means integrating that which cannot be integrated, a world of causal ‘things’ and a world of the non-causal but instead constitutive?

As far as the question of need is concerned, this article presents, as an illustrative example, the IR literature on the nature of late-modern Western war. Here PDT’s dominance seems to either short-circuit the adequate study of discourse, or causes those who adopt a poststructuralist or postmodernist perspective to grant certain discourses a questionable degree of influence. I particularly concentrate on a discourse of humanitarianism said by many to strikingly affect the ‘Western way of war’ since the end of the Cold War. It is clear that this discourse is an important element in theorizations of the present and future manifestations of war for the West. There are, however, vastly different positions on the direction of its effect without any forthright attempt to garner evidence that would go some way towards reconciliation. In short, some further analysis of the discourse is needed. But the nature of existing theoretical frameworks — where significant play is given to extra-discursive causal elements such as technology, system structure or post-industrial social conditions — means that in order to resolve disagreements by building upon existing knowledge, a PDT perspective is inadequate.

To show how this is so, I begin this article with a discussion of the foundational tenets of PDT. I contend that these give rise to methodological choices which could not offer up the kind of evidence needed to address empirical puzzles involving competing views on the impact of a discourse or discourses, and which also involve the operation of extra-discursive causal ‘things’. PDT is in short too ambiguous on the status of the extra-discursive. While PDT does not deny the reality of a world outside of discourse, there is an important block thrown up against integrating it meaningfully into analysis. With an intervention assisted by the philosophical position of critical realism (CR) — which it should be here noted is fundamentally useful for the way it emphasizes ontological reflection over epistemology — it is contended that poststructuralism’s philosophical ontologyis anti-realist and therefore allows researchers to do no more than offer internally rich studies unconcerned with external validity. While there is no sure-fire philosophical argument against seeing our relationship to the social world in a way that such studies are not the pinnacle of social analysis (Jackson, 2011; Montiero and Ruby, 2009), I am making the analytical bet (see Kratochwil, 2007) that a CR-influenced view of discourse can lead to productive new substantive insights by opening up discourse analysis to scholars who want to view it as a causal part of their broader explanation.

After then presenting the Western war literature to illustrate how the dominance of PDT may be contributing to detrimental absences in our understanding of certain phenomena, I turn to CR to develop a view of discourse as a certain kind of causal mechanism among myriad others. Here it is clear that the nature of discourse as a possibly causal thing does presuppose a turn away from studying causality through ‘variable testing’. Discourse, defined as a ‘cohesive ensemble of ideas, concepts, and categorizations’ (Epstein, 2008: 2) that affords a ‘way of speaking which gives meaning to experiences from a particular perspective’ (Phillips and Jørgensen, 2002: 66–67), is rightly posited by PDT as too tightly interwoven with every aspect of social life to allow for its proposition as a variable which independently pushes some separate ‘dependent variable’ (Hansen, 2006: 26).

Fortunately, a general dissatisfaction with the status of causal social science as variable testing is growing in IR. ¹ Some CR-influenced theorists offer a fundamental re-imagining of causality that can account for discourse as a causal aspect of the social world (Kurki, 2008; Lebow, 2009; Wight, 2004a). The key to this is a view of causality as involving the dispositional properties or tendencies of entities that, when activated within a system, generate events — and are then deemed causal mechanisms. And as Colin Wight argues, as ‘one of the processes in a concrete system that makes it what it is … semiotic systems have mechanisms, but they may also themselves be considered some of the mechanisms that make the social what it is’ (2004a: 284). Here one shows causality as well as one can, and does so in ways shaped by the nature of the causal mechanism in question.

Finally, the article concludes with an introduction to critical discourse analysis (CDA), the most promising discourse-analytic perspective for the study of discourse as a causal mechanism — and a method as yet under-utilized in IR (though see Jackson, 2005). CDA contains a wide-ranging field of scholars with one important over arching commitment: the study of discourse as something which stands in dialectical relation to other extra-discursive forms of social practice. To provide an introduction to how a CDA perspective might help IR scholars, I tap into the work of one of CDA’s most prominent theorists, Norman Fairclough, because he has already begun to sync his work with CR philosophical insights (Chouliaraki and Fairclough, 1999; Fairclough, 2000, 2003, 2005). The article thus concludes by emphasising how CDA tools, if grounded in a CR view of the social world, would contrast to a PDT framework by allowing scholars to creatively trace the dialectical relationship between discourse and prominent extra-discursive mechanisms over time in order to develop contingent generalisations about what effect discourse is having on a phenomenon of interest. It is hoped that through this outline, bolstered as it is by an understanding of the deep methodological reasons one might choose to turn to CDA, IR scholars will more readily attempt the discourse-analytic work necessary when their empirical interests intersect with some contended place for important discursive influence.

Poststructuralist discourse theory

As Louise Phillips and Marianne W. Jørgensen describe the impetus for the discourse-analytic movement in the social sciences, ‘underlying the word “discourse” is the general idea that language is structured according to different patterns that people’s utterances follow when they take part in different domains of social life’ (2002: 1). Through discourse analysis these patterns can be studied for their impact on events and the general development of the social world. PDT does this by adapting for social analysis certain tenets of the wider philosophical position of poststructuralism.

PDT’s fundamental premise is that while there is of course a ‘real world’ of objects independent of our knowledge, it is only through meaning-making that these objects become real to us (Laclau and Mouffe, 2001). Ernesto Laclau and Chantel Mouffe (2001: 105) call this articulatory nature of the social world discourse, and they contend that all social phenomena are encompassed by it (Phillips and Jørgensen, 2002: 33). Thus discourse is ‘coterminous with the social’ (Torfing, 2005: 8), and the social is structured in the same way as that which is assumed to be the most elemental system of social interaction, and progenitor of the meaning-making capacity of discourse — language (Phillips and Jørgensen, 2002: 35; Zito, 1984: 8–9).

A particular understanding of language allows PDT theorists to emphasize how there is no real ‘transcendental centre [i.e. referent] that structures the entire structure’ (Torfing, 2005: 8). They draw on the structuralist linguist Ferdinand de Saussure’s (1986) influential thesis that the relation between linguistic signs and their supposed referents in the world is arbitrary. The meaning of signs in language, Saussure contended, is only determined by their relation to other signs — and this relation-making is a social process between signifier (words) and signified (concepts), rather than something determined by nature (Sayer, 2000: 36). At the social level, signs gain their meaning within wider discourse structures, or ‘discursive totalities’ (Laclau and Mouffe, 2001: 107). Thus as a particularly important node in social construction, discourse creates meaning and its traces create reality (Epstein, 2008: 7).

It is important to emphasize the ontological nature of these claims, as this has significant influence on the methodological choices of poststructuralist discourse research in IR. Discourse ‘is not equivalent to “ideas”; discourse incorporates material as well as ideational factors’ (Hansen, 2006: 17). In fact, analysts claim to be ‘moving beyond the metaphysical oppositions of idealism and materialism’ (Campbell, 1998: 25; Daly, 2008). It is not ‘as if’ the world is discursive, or even that we should only study discourse because it is the most reliable source of knowledge. All the factors researchers take into account — biological, psychological, institutional or whatever — are discursive objects first and foremost. This means that integrating into our analyses any foundational or objective reality ‘in its naked existence’ (Laclau and Mouffe, 1987: 85) — that is, apart from the social interpretation of it — ‘is not only unnecessary methodologically but it amounts to collapsing back the very distinction (between the social and the natural world) that had opened up the space for constructivist approaches in the first place’ (Epstein, 2008: 7).

Social analysis here is about mapping discursive structures to show how they produce objects, and especially subjects — and critique takes the form of deconstructing discursive totalities away from their ‘hegemonic’ status to open up what is perceived as a ‘closure’ (Torfing, 2005: 20). Subjects, or more accurately their identities, are developed against an ‘other’. In deconstruction, the ahistorical or fixed constructions of objects or identities are exposed as contingent and relationally structured, and counter-discourses are proposed as possible ways a situation could have been (Campbell, 1998: 20–24). Analysts may also seek to find some ‘blank spot’ or contradiction for which a particular discourse shows its weakness, or conduct a ‘genealogy’ which ‘traces the formation of a concept’ (Hansen, 2006: 212).

Also following from the poststructuralist ontology is a rejection of causality. As Lene Hansen writes:

for discursive causality to be considered an actual casual effect, one needs to separate two variables and to observe each independently of the other. This, however, is precluded by poststructuralism’s insistence on the ontological significance of discursive practice: identities are produced, and reproduced, through … discourse, and there is thus no identity existing prior to and independently of … [action]. (Hansen, 2006: 26)

Analytically, this also means that there is ‘no extra- or non-discursive realm of explanations from which one might construct competing explanations’ (Hansen, 2006: 25). Analysts are not concerned with the contentions of causal social science in any meaningful sense, but with asking questions that the mainstream would not — questions primarily about how discursive representations of the social constitute the identities of actors (Torfing, 2005: 22).

While poststructuralists neither deny the material world nor claim that it has no place within their analyses, ² many critics do contend that an extra-discursive dimension to social processes and practices is detrimentally absented within PDT analysis. Though Hansen (2006) argues that practice simply is discourse, others define practice as the extra-discursive ‘competent performances’ that ‘engage with the relationship between agency and the social and natural environments’ (Adler and Pouliot, 2011: 2). We may begin to ask how plausibly this part, as seen by others, is incorporated.

Sympathetic critics such as Mark Laffey (2000) and Iver B. Neumann (2002) worry that poststructuralists have not integrated a sufficiently wide array of factors to constitute convincing or even socially responsible analyses. Based on a critique of David Campbell’s (1992) work on the US foreign policy Laffey argues that poststructuralist accounts cannot help but ‘participate in, rather than overturn, “established modes of thought and action …” [and as] explanations … are undermined by the account of the social they presuppose’ (2000: 430). Given that Campbell claims that poststructuralist analysis should help one ‘consider the manifest political consequences of adopting one mode of representation over another’ (Campbell, 1992: 4), Laffey finds the representation he establishes absents internationalized capitalist forces in favour of a global social order driven by new representations of ‘flux and uncertainty’, and thus actually reinforces the capitalist system by eliding its possible influence (2000: 434–440).

We can extend this critique by positing that even if these capitalist forces are at some point representationalized in discourse, Campbell must only cover a limited number of discursive practices because many do not have presence in the utterances pertinent to his cases. For instance, we can easily find, as is Campbell’s (1998) central concern in National Deconstruction, lack of a multicultural ethos in the Western plans for Bosnian partition in the mid-1990s. Those policy documents have as their subject actually warring parties. So as much as they may constitute these parties through their representation, there is also a practically efficacious aspect to conflict that biases these documents towards speaking of the parties as such — as different groups. But rarely in such policy documents will one find meaningful reference to certain underlying economic or political forces. If an official is trying to look as though she is doing all she can for a humanitarian goal, she of course does not speak about such seemingly petty things as market stability or the Vietnam syndrome.

While Laffey is writing from a historical materialist perspective, Neumann (2002) generally seeks to maintain a poststructuralist stance on the social world while improving analysts’ ability to account for practice in all its facets. To do so he turns to the ethnographic concern with culture, presenting it as a mediating stratum between discourse and practice (Neumann, 2002: 630). In his empirical example, culture becomes synonymous with the ‘stories’ we tell about how practice ought to play out (Neumann, 2002: 648–649). So in Neumann’s telling, a discourse or discourses exist, and actors work against or with these to create conceptual proposals or stories about future practice. Once practice becomes initiated, the cycle turns back and stories are told about it which then influence discourse. He writes that to study this we ‘must and should choose other points of departure and include other kinds of material than the study of texts’ (Neumann, 2002: 651).

While this ‘practice turn’ is portrayed by Neumann (2002) as a ‘compliment’ to poststructuralist work, it actually reflects the need to break with poststructuralist ontology when seeking to explain the effects of discourse. For other design-stage ‘fixes’ to poststructuralist analysis — both Hansen (2006: 218) and Jennifer Milliken (1999) offer cogent attempts to systematize and make more rigorous poststructuralist discourse-analytic methods — emphasize how fundamental a break Neumann is suggesting, as they actually contradict his concerns.

Hansen emphasizes the differing analytical goals possible within PDT. Her own empirical analysis diverges significantly from Campbell’s findings on Bosnia (Hansen, 2006: 218), suggesting that competing explanations can be juxtaposed in search of more precise knowledge and a fix for the sense that conclusions are ‘given in theory’ (Laffey, 2000: 430). But as Hansen (2006: 218) admits, within a poststructuralist frame it is impossible to ‘determine who is right’. There can only be an internal concern with ‘the relationship between analytical and methodological decisions and the analysis produced’, because there is no way to ‘measure’ the causal impact of one discourse over another, and no way to step outside of discourse (Hansen, 2006: 218). There is thus no explanatory reason to include Neumann’s ‘other kinds of materials’.

Milliken (1999: 227–228) argues that it is a mistake for PDT analysis to remain abnormal, as part of the ‘anti-scientism’ and ‘dissident’ nature of discourse theory, to the degree theorists actively avoid considering best methods and practices. To arrive at these best practices she determines the maxims that all poststructuralist analyses are committed to. Her first maxim — that discourses are systems of signification — is presented in the Saussurian sense of being detached from any relation to a referent (Milliken, 1999: 229). With this in mind, scholars are logically prevented from analytically stepping outside of discourse in order to explain phenomena (Wight, 2006: 135), as Neumann seems to demand. The second and third maxims — that discourses are productive of objects being defined by the discourse, and are constantly open to re-articulation — are portrayed in ways that militate against seeing discourse as something which interacts with other social objects. The commitment to seeing discourses as in need of constant re-articulation, as in constant flux, is principally true but empirically an interesting matter of degree (Chouliaraki, 2002), something Milliken does not emphasize, thus Neumann’s concern over ‘which … narratives will make a difference, of how this may or may not happen, and with what effects’ (2002: 651).

The real problem from a CR perspective is that these methodological considerations are actually shaped from epistemological reflection on what can be known rather than ontological reflection on how the nature of objects would demand certain methodological moves. Wight (2004b: 205) argues, in a critique of deconstruction that is applicable to PDT, that ‘the real distinction between the Real of critical realism and that of deconstruction lies in the a priori manner … deconstruction denies the possibility of coming to know the real’. He claims that deconstructionists are ‘drawing ontological conclusions from epistemological arguments’ (Wight, 2004b). Roy Bhaskar (2008: 5) identifies this ‘epistemic fallacy’ as the root of damaging philosophical and methodological errors. Wight specifically notes Jacques Derrida’s (1988: 148) claim — one that fits well within PDT — that ‘all reality has the structure of a differential trace’ as an example of this epistemic fallacy at work; assuming that because we cannot know the world except through our interpretive experience, it is also entirely structured as the language we use to interpret it. As one advocate admits, poststructuralists ‘simultaneously accept the existence of referents and insist that whatever reality is ascribed to them, this ascription is the product of a political articulation’ (Wilmont, 2005: 763, latter emphasis added).

CR would further note the epistemic fallacy at work in the denial of causation. Analysts claim to be interested in showing the impact that certain discourses have — a ‘focus … on the political consequences and effects of particular representations and how they came to be’ (Campbell, 1998: 5). Then why not propose discourse as a causal element of the social world, however difficult this may be to show? Analytical difficulty undergirds ontological specification. Because discourse is in constant entwinement with action, talk of causation is elided and we at best are left with rich analyses incommensurable with each other outside a comparison of analytical choices, or at worst with a discursive representation that others would with good reason find problematic.

Epistemological trepidation pushes PDT, despite acknowledgement of a real world, into anti-realism — defined as acknowledgement of a real and knowledge-independent world but rejection of any meaningful independence of it from our minds (Miller, 2010). Language might as well be all that is real. If we know at the very least that something exists without our knowledge of it, is it wholly inappropriate to begin positing how some of these things are when they may be causally efficacious to social events? If it is inappropriate or even unnecessary there is no place for explanations involving the discourse causes of social phenomena as related to other kinds of things, nor the establishment of representations that can be judged more or less right against competing explanations. This is not to say that we can hope to test whether something like geography or technological superiority is more causal than a particular discourse, but that at the very least we should be able to determine if a discourse has a directional effect of constraint or enablement on certain actions within a phenomenon given other influences like geography, technology, or whatever.

But is this epistemological trepidation really a problem with substantive effects on our knowledge of IR phenomena? What are the situations in which PDT is inadequate, and what do they tell us about the way discourse might need to be reconceptualized? In the following section I offer the illustrative example of a literature which contains a discourse-centric empirical puzzle that remains under-analysed precisely because of a lack of conceptual resources by which discourse can be causally connected to a wider realm of extra-discursive (and also discursive) practices that form the bulk of the explanation. With the discussion above it should become clear that a PDT analysis would not help, except with certain methods, to address this puzzle: the puzzle’s force is predicated on there being two diametrically opposed explanations relative to discourse; and a failure to analyse discourse in depth flows directly from an inability to explicitly present discourse as causal in relation to other things.

Humanitarian discourse and the Western way of war

Because of the Western coalitionary wars of the post-Cold War era, indeed ‘postnational’ in Ulrich Beck’s (2005) view, a research programme has developed that views certain wars as the product of often global, or at least broadly Western, social relations (for early examples, see Roxborough, 1994; Shaw, 1988: ch. 1). Studies of these ‘Western wars’ concentrate on how unlikely major-power war is after the Cold War because of nuclear weapons, economic integration and other factors (Moaz and Gat, 2001; Mueller, 1989); and the nature of Western warfare is concluded to be something of a technologically advanced spectator-sport for Western societies (Gray, 1997; McInnis, 2002; Van Creveld, 1991). Theorizations also feature concern with the moral discourses that both enable and constrain war. War transformation issues even bleed over significantly into international ethics (see Banta, 2011; Burke, 2004, 2005; Orend, 2006). Prominent in numerous theorizations, along with a continued emphasis on the influence of globalization and technology, is a striking concern with the role of humanitarianism in shaping warfare (for landmark studies, see Coker, 2001; Ignatieff, 2000). Beginning with the ‘war on terror’ especially, a striking divergence in opinion developed over what exactly humanitarianism was doing to shape Western war — whether it was constraining or enabling the West’s warring tendencies.

IR scholars contend that a surge in humanitarian security discourse in the 1990s was the result of a unique combination of post-Cold War liberal dominance and the recognition that instability within states was one of the greatest threats to international peace and security (Barnett, 2008; Murphy, 1996; Weiss, 2007). Many scholars also believe that the human rights logic embedded within such security endeavours works counter to the logic of war (Farrell, 2005; Risse-Kappen et al., 1999; Thomas, 2001). And yet, as Western powers sought to conduct war after 9/11, the rhetoric of humanitarianism seemed to play a crucial role both in its initial legitimization, and in the combat and peace operations that followed in Afghanistan and Iraq (Freedman, 2005; Heinze, 2006). This led to surprise and even outrage in some corners (Roth, 2004). These wars were ‘Old War’, as Mary Kaldor (2005) put it — replete with marshal motivations of state’s interest aimed towards the total defeat of another enemy-state regime. They should not, according to some, have contained a high degree of humanitarian legitimation, thought incompatible with a ‘war system’ (Lawler, 2002: 154). How do we theorize the Western way of war given this puzzle?

One perspective, holding that humanitarian discourse on balance constrains the West from fuller manifestations of war, emphasizes how war is globalized and shaped by concerns for global markets, institutions and sentiments more than some sense of pure military victory (Rasmussen, 2006: 11). And globally conscious societies means that Western war-makers are forced to address the dictates of burgeoning humanitarianism. So, for instance, humanitarian organizations and rhetoric must be ‘annexed’ to make acceptable the deaths of civilians that might otherwise be interpreted as the result of callous risk avoidance (Shaw, 2005: 91–92). Or, Western states symbolically close off war by declaring the end to ‘major-combat operations’ because of concerns that the ‘post-war reconstruction’ period will result in higher rates of civilian casualties — and thus adverse public attention (Shaw, 2005: 77). But, these theorists conclude, ‘humane warfare can only be a palliative measure to reduce the risks of getting it horribly wrong’ (Coker, 2009: 130).

As Martin Shaw argues, the over-riding risk to be avoided for war practitioners is soldier deaths, and so no matter the effect sought through humanitarian appropriation, it will often be revealed that ‘[c]ivilians’ risks were proportional not to the risks to soldiers … but to the political risks of negative coverage’ (2005: 135; see also Crawford, 2007; Smith, 2008). This, he believes, causes Western societies, because we live in an ‘age of human rights’ (Shaw, 2002: 357), to demand ever greater civilian protections when soldier deaths approach zero (2005: 136) — what Michael Ignatieff calls the ‘basic equality of moral risk’ (2000: 161). Driving further constraints on war is thus ‘not so much legal norms as the perception that all the individual lives mattered … [such that even] the world’s most powerful state could be brought to account’ (Shaw, 2005: 138).

This perspective, it is argued by critics, contains the logical implication that anything that allows major war to occur must be of a ‘second order reality’, and will eventually be called to account by more fundamental material realities and their concomitant normative discourses (Malešević, 2008: 107). Humanitarianism aligns with a contended likely future world order that has adapted to globalization through institutional mechanisms concerned with the global citizen (Beck, 1997; Held, 2003). And it inherently goes against the grain of some salient practices or by-products of modern war, such as civilian casualties, the destruction of infrastructure and the increased media coverage that makes them more noticeable (Kaldor, 2006). In all, we see that humanitarianism is theorized as an important constraining piece within a much broader set of social relations.

A counter-perspective, of which many adherents are more in line with a PDT outlook on the social, emphasizes humanitarianism as a force multiplier within the new Western way of war. Western war is made attractive largely because of humanitarianism. Technological war capabilities are, through the help of electronic media, ‘deployed with a new ethical imperative for global democratic reform … and humanitarian intervention’ (Der Derian, 2000: 772; 2009). There is a symbiotic relationship between humanitarianism and other determinants, rather than their more dialectical relation in the above perspective (see also Dillon and Reid, 2009; Jabri, 2007).

These theorists make a connection between the cosmopolitanism motivating the previous position and policy makers’ more self-interested motives. Cosmopolitan scholars ‘are providing … a legitimizing discourse in which neo-conservatism can situate itself’ by constructing a humanitarian discourse within the security realm that serves to de-politicize the choice of war (Dexter, 2008: 56; see also Douzinas, 2007: chs 6–7; Žižeck, 2005). The symbiotic relationship between humanitarianism and Western war is contended to be especially evident in its use to legitimize even the means of war (see Hancock, 2010). Humanitarianism is argued to ‘produce’ civilian deaths as ‘accidents’ rather than violations of humanitarian norms (Beier, 2003; Owens, 2003).

With near-diametrically opposed conclusions on humanitarianism and war, highlighted is doubt about whether scholars have got ten this important issue right. As it stands, though, it is difficult to cull evidence of discourse-in-action from existing works that would allow for some determination between the two perspectives. Both make claims implying that the most crucial test period is humanitarian discourse’s influence during the height of conflict, where soldier and civilian casualties, the influence of new technologies, and other exigencies would most forcefully activate the various causal determinants of the way of Western war. Yet in the first perspective, humanitarianism is defended based on logical arguments for its conceptual fit with new realities theorized outside of a war context. Kaldor, for instance, acknowledges the humanitarian rhetoric of the wars, but refuses to accept its possible influence on grounds of conceptual friction with a reality where ‘if the concern is humanitarian, it cannot be authorized unilaterally by a government that represents a particular group of citizens, it requires some multilateral authority’ (2008: 29). Instead of investigating humanitarianism in war, as a discourse, its operative character is derived from prior cosmopolitan theorizations of globalization. The functional whole of globalization dictates how war practices and the social forces of their generation are perceived.

The ultra-critical view of humanitarianism holds that the proposed universality of human rights creates non-acceptance of the Other, which drives extreme differentiations and vilifications that lead to the perceived necessity to change others through force and expansion (Baker, 2010; Jahn, 2007a: 90–94; 2007b). Human rights can thus stand ‘as a moral structure above the law or as a substitute for the legal structure itself’ (Hardt and Negri, 2004: 27). Given the concrete effects contended, one would assume that evidence would have to go beyond noting some conceptual quirk, or even the correlation between the rhetoric of a largely academic, official or activist discourse of humanitarianism and the presence of war. But this is never done (Chandler, 2009). To make the leap from a correlation to causation without considering possible counter-explanations, especially when the contention is so counterintuitive, seems unpersuasive (Walzer, 2004: 252). As Shaw (2008: 379) notes of Vivienne Jabri’s (2007) work, the argument that humanitarianism is a legitimating discourse of war is gleaned from an abstract reading of elite and scholarly discourse rather than any meaningful engagement with the practices of war.

What is clear is that neither a critical-explanatory account of Western war nor a more postmodern perspective adequately analyses a discourse which does so much work in their theorizations. For the former, it may simply be that given a proclivity to understand the world in terms of material structures and processes such as globalization and technological advancement, there is no room for a discourse perspective. How would one study discourse alongside the causal structures that form the backbone of the explanatory framework if the dominant discourse-analytic perspective presupposes that we cannot understand these structures as such? For the latter, it seems that the worst fears of poststructuralism’s critics are at work. Because causality and external validity are chimeras, scholars seem to simply choose a discourse present in elite utterances and grant it a great deal of power. There is an all too simple movement from media or official texts to contending ‘an untested constitutive nature and a perlocutionary force upon the social issue’ (Molina, 2009: 194). The fundamental premises of CR, tapped in the next section, can help develop a view of discourse that fits within an explanatory social science that can delineate between competing truth claims based on empirical evidence of the relations between the discursive and the extra-discursive.

Critical realism: Discourse as a causal mechanism

While there are numerous varieties of philosophical realism (Harré, 1986; Putnam, 1982), the critical realism begun by the work of Bhaskar (1989, 1998, 2008) is most concerned with the social sciences and their conceptually imbued subject matter. If we accept realism, we must acknowledge that it is the make-up of objects ‘that determines their cognitive possibilities for us’ (Bhaskar, 1998: 25). Thus social science practice can either implicitly or explicitly begin with ontology. Explicit ontological reflection, however fallible (Cruickshank, 2004), offers the possibility of a more cogent foundation by forcing one to place their presuppositions out front. CR seeks to ‘underlabour’ in this endeavour by determining the necessary though very basic ontology of both the natural and social worlds from transcendental question and argument — that is, asking what must be true for X to be possible (Collier, 1994: 20). Bhaskar centres his transcendental argument on the success of scientific experiments, and — especially for the social sciences — the possibility of social practice. ³

With broad CR methodological implications extensively detailed elsewhere (see Archer, 1995; Danermark et al., 2002; Sayer, 1992), I will merely outline CR’s implications for the place of discourse within our scientific and social ontologies. At a basic level, CR argues for a world that is ‘structured, differentiated, stratified, and changing’ (Danermark et al., 2002: 5). Bhaskar argues that reality is deeply layered, ⁴ so that lying beneath appearances are structures which act as mechanisms in the generation of events (1998: 9–12). Thus what is real is not exhausted by what is experienced or readily apprehensible; magnetic fields or norms are unseen, but we deduce their reality by first posing them as theoretical entities and then seeking to observe evidence for them in the world of appearances. Ontological reflection on this deeply structured social reality, and discourse’s place within it, can help us identify how discourse should be studied.

Ontological depth implies the structured, differentiated and relative materiality of social objects in the world. There must first of all be structure to reality for it to have depth which we can uncover and make use of. Differentiation and materiality are dependent on a commitment to emergence in social phenomena, the latter meaning that there are objects that ‘arise from and depend on some more basic phenomena yet are simultaneously autonomous from that base’ (Bedau and Humphreys, 2008: 1). Emergent objects contain powers and are able to have effects not contained within their more basic building blocks. And ‘materiality’ here merely denotes some relatively enduring thing that exists beyond the conception of individual agents (see Bhaskar, 1989: ch. 7; Jäger, 2001). Thus though we may speak of a material world in the sense of rocks, rain and the like, CR identifies ‘emergent materiality’ as anything that has ‘an ontologically objective and socially consequential existence, whether or not any actors are aware of them’ (Porpora, 1993: 222). This would include discourse, at least for individual actors at specific moments in time. Meaning-making results in emergent shared conceptual structures and ideas that individuals must work with as they act (Fairclough et al., 2004).

Also derived from scientific and social practice is the contention that events in the world truly are caused, but that they are not caused by concatenations of other events (‘constant conjunctions’). Especially in a conceptually imbued open system, similar events over time may be preceded by different combinations of events, and so there must instead be tendencies within social entities that combine to generate events (Kurki, 2008). Bhaskar contends that social scientific generalization can thus only consist of the ‘historically restricted tendencies’ of things which in certain circumstances or for periods of history ‘may never be manifested but which are nevertheless essential to the understanding (and the changing) of the different forms of social life’ (1998: 54). In place of a view of causation as constant conjunctions, which we will remember is the view that poststructuralists explicitly reject, CR proposes the causal mechanism. Defined as ‘that aspect of the inner and environmental structure of a thing by virtue of which the thing has a certain power’ (Demetriou, 2009: 444), a mechanism can be any real entity — whether an institution, an agent’s psychological or biological condition, or a discourse — that is ‘the operative or motive part, process, or factor in a concrete system that produces a result’ (Wight, 2004a: 288).

For the scientific process of uncovering causal mechanisms, a distinction is made between the intransitive and the transitive. The former are the objects of scientific analysis, while the latter are the always-mediated discourses we use to describe them. Poststructuralists want to contend that all of society is transitive because the world is made real through discourse (Laclau, in Laclau and Bhaskar, 1998: 10). CR contends that discourses being analysed are intransitive enough to be studied as causal objects. Even as discourse is shaped by the words and actions of many agents over time, at any one time it is relatively intransitive to those studying it or even being affected by it. All transitivity means is that our scientific conjectures are always more or less fallible. What ends up happening for PDT is that there are meanings and objects, and they reject anything we can say about the latter. I want to contend that not only should we say something about the latter, but that there are also ‘meaningful objects’. As Amit Ron argues, if society depends upon ‘relatively enduring shared meanings’:

… meanings become (relatively) fixed and thus (relatively) independent from the discursive context of their production. Therefore … in some cases we can treat meaningful objects the same way we treat natural objects. (2010: 156–157)

Anti-realists often contend that this objectification of discourse means an essentialization that fixes it to its referent, and thus requires a strict correspondence theory of truth that is easily shown to be untenable (Potter, 1996). Remember that for PDT all that is important is the relationship between signs, where the referent has no role. CR does not wish to do away with this relationship, but to add the referent in a tripartite schema that can realistically represent the processes of stabilization or change in discourse (Sayer, 2000: 36–38). Furthermore, the referent contains both the object for which words refer, and also other discourses and our ‘practical involvement in the world’ (Sayer, 2000: 38). Instead of an overwrought concern with essentialism, Andrew Sayer contends that ‘we can avoid implying that any object has a single essence … if essences can change or be acquired because it is merely that “which makes it one kind of object or social relation rather than any other kind”’ (2000: 85, 87). The causal status of a discourse is akin to understanding its momentary essence — that aspect of its structure by which it is able to have an effect at some specified moment. The productive ability of discourse is acknowledged, but we avoid what can only be an exaggerated metaphor when it is contended that discourse constructs persons or objects. Instead, we must insist on the more accurate formulation that the discursive effect ‘of treating certain persons as if they were x, can in varying degrees, depending on the situation, succeed in making them x’ (Sayer, 2000: 45).

Given differentiation through emergence and a view of causality as dispositional properties, the possibility is opened for the causal analysis of discourse as we analytically separate it from other entities. Portraying that real separation from other social elements is inherently difficult; most students of meaning turn to ‘constitutive theorizing’, simply investigating the conceptual relations that answer a ‘how possible’ question (Wendt, 1998). CR would demand that we acknowledge this difficulty and make affordances to it (more on this below). As Milja Kurki (2008: 181) writes, ‘constitutive theorizing … is not just about inquiring into conceptual relations (meanings) but about inquiring into how they play themselves out in the social world, giving rise to certain practices and social relations’.

The biggest obstacle to modelling discourse causally is that once we overcome the fear of differentiation, we are still left with the fact that discourse is complexly and interminably in dialectical relation to the extra-discursive world (Fairclough et al., 2004: 27; Laclau and Bhaskar, 1998: 12). While, say, a physically essential causal power, such as a bomb’s ability to explode, is impossible to alter fundamentally (though it can be made latent with the introduction of other social mechanisms contributing to the object’s interpretation and use), discourse is always relatively altered as it acts in social events. Even relative stabilization of a discourse is only accomplished through constant articulations that contribute to its reproduction. So even as a discourse may seem to be remarkably stable, it is impossible to say that such a discourse is asserting more power than in a case where a discourse’s features are changing.

What we can do, though, is identify the key aspects inside and outside of discourse which are in mechanismic (Gerring, 2007) relation with one another; we can measure causation as directionality, as enablement or constraint on agents’ desires and practices. In particular, we would want to see what social elements outside the discourse in question are able to, especially if in a patterned or generalizable way, influence which particular conceptual elements of the discourse and vice versa (Nellhaus, 1998). Strong relational patterns in such explanatory accounts then may enable analysts to produce always fallible but useful contingent generalizations on something like the development of the Western way of war vis-a-vis humanitarian discourse. The next section introduces a branch of discourse analysis which has developed tools to do just this.

The need for critical discourse analysis

CDA emerged as a discourse-analytic programme in the 1980s, mainly with the work of Fairclough, Ruth Wodak (1989) and Teun van Dijk (1980, 1987). While specific CDA approaches vary greatly — from cognitivist orientations (e.g. van Dijk), historical comparativist concentrations (e.g. Wodak) and historical materialist leanings (e.g. Fairclough, 1992) — at least one major CDA theorist, Fairclough, has explicitly attempted to sync his specific approach with the insights of CR. Where other theorists often see CDA simply as a toolkit that aims at specific critical theory problems of inequality (Howarth, 2005: 318; Van Dijk, 1995), Fairclough emphasizes the important meta-theoretical differences marking off CDA from other approaches, and attempts to think through what this may mean for the analysis of discourse. Fairclough’s is also the most elaborated form of CDA (Phillips and Jørgensen, 2002). As such, and though the connection he makes between ontological and epistemological issues in CDA can be adopted by and serve to strengthen any CDA approach, this account of CDA is ardently Faircloughian. But even this adumbrated introduction should serve to demonstrate how bringing CDA to IR might serve to fill the need for discourse analyses based on a CR view of discourse.

CDA sets up a dialectical relationship between discourse and society, contending that both are conditioned by each other. Discourse is but one ‘moment’ of the social among other kinds of social practices (Chouliaraki and Fairclough, 1999: 28). Discourse is still defined, following Foucault, as the ‘domain of statements’ at various levels of abstraction which constitute a way of ‘representing aspects of the world’ (Fairclough, 2003: 215). But instead of these statements encompassing all aspects of social life, discourse internalises but is not reducible to them (Chouliaraki and Fairclough, 1999: 6). Discourse is thus solely representational in nature, while relationally connected to other non-representational elements. And crucially, our embeddedness in the world is taken by CDA in the same direction as CR: if our scientific theories are internal to our object of study, we have a vital duty to get things as right as possible by explaining the role that discourse plays in events (Van Dijk, 2003: 353).

The primary difference this makes relative to PDT is that for many research questions, greater attention will be paid to a temporal tracing of pertinent sites of discourse production and the other social practices which precede or follow them. The most oft-heard critique of CDA is that the ambition of explaining the dialectical relationship between discourse and society is rarely convincingly accomplished (Blommaert and Bulcaen, 2000: 460; Hammersley, 1997; Luke, 2004). This may be because Fairclough (2003) emphasizes the daunting need to integrate a full gamut of necessary data, from in-depth linguistic analysis to broad social empirics, in order to grasp discourse’s effects on social practice. Thus, CDA is necessarily interdisciplinary, with the supplementary use of non-linguistic ‘middle-range theories’ chosen based on the subject matter (Molina, 2009). And it emphasizes a more detailed linguistic accounting than is usually undertaken in PDT (Chouliaraki and Fairclough, 1999: 152). Thus Fairclough’s (1995: 2) seminal ‘three-dimensional model’ for what a CDA should examine emphasizes the text(s) (also called the ‘communicative event’), the discursive practices within which this text is embedded (also called the ‘order of discourse’), and the social practices encompassing the order of discourse (also called the ‘social field’). Putting these three elements into play in a way that allows us to reach an explanation of their causal inter relations is a process constantly evolving as new problems and phenomena come into focus (Chouliaraki and Fairclough, 1999: 59).

To study something like the relationship between humanitarian discourse and the social field of Western war we would begin with the ‘order of discourse’ for this social field. Fairclough translates the social functions of language into a tripartite analytic schema for discourse analysis: genre, or ways of acting out discourse such as interviews or news editorials; discourse, or the actual representations of the world in talk or text; and style, or the identity proposed by behaviour associated with discursive enunciation (2003: 26). The starting point for any CDA analysis is to grasp the ‘order’ of all three above elements within a chosen ‘network of social practices’ (Fairclough, 2003: 220). The order of discourse acts as the baseline from which analysis of future discursive practices are judged. Developing a view of this is, as Fairclough (2003: 6) notes, a quite ‘labour-intensive’ qualitative endeavour that may also require some quantitative analysis of keywords and collocations across a large corpus of texts. But it is the necessary first step in beginning a causal analysis of something like humanitarian discourse and Western war.

In CR terms, the role a discourse plays in an order of discourse — how it draws on and re orders genres, styles and other discourses — is its dispositional power as a possible causal mechanism. To analyse this ordering between discourses, CDA draws on the insights of systemic functional linguistics (SFL) (see Halliday and Matthiessen, 2004; Martin, 1992). This is because SFL views ‘language as a strategic, meaning-making resource’ (Eggins, 2004: 2), and offers tools to understand how different levels of meaning are created in a text and may work to correspond upwards into the social level. These tools, too myriad for a discussion here (for guide books on the linguistic analytic tools on offer, see Fairclough, 1995, 2003; Van Leeuwen, 2008; Wodak and Krzyzanowski, 2008; Wodak and Meyer, 2009), offer a way of studying language in its social context and as a product not of formal or innate language structures but functional structures associated with social interaction (Fairclough, 2003: 5–6).

To understand how something like humanitarian discourse transmits its power from the order of discourse into the field of social practice that is war, CDA demands using theories of war to place discourse in its political context. We do this to identify what the realist philosopher Dave Elder-Vass (2011: 10) calls a ‘discourse circle’, or the group of positioned individuals who act and speak in such a way that a discourse becomes ‘endorsed and enforced by a wider social group that makes such behaviour more effective than it would be if simply perceived as the behaviour of certain specific individuals’. It is only through such groups and persons, always structurally embedded, that discourse can eventually affect events.

Here we come up against the difficulty of analytically determining the discursive from the extra-discursive. Differentiation of objects — discourse from discourse and discourse from other extra-discursive social relations — is always a conceptual and fallible endeavour. Our question would drive how we go about this, rather than an epistemological mandate to view only discourse as an analytical object. Humanitarian discourse, as the causal discourse of interest, might be differentiated from other social relations of war identified in the literature, such as the military–industrial complex, the economies of under developed states or the domestic politics of casualties. Absent an established literature, we turn to creative use of analogy and metaphor, broader social theory, and empirical evidence to model the social relations we believe to exist and be especially operative (Patomäki, 2002). So while of course something like the practice of strategic bombing has a discursive aspect, and may play a prominent role in the order of discourse of Western war, this element may need to be conceptualized in its extra-discursive aspect as a relatively regularized ‘competent performance’ described through relatively objective data: the number of bombs dropped, their operative technological infrastructure, the amount of damage done or the institutional procedures for authorization. The analytical goal here is to specify how a discourse is a ‘moment’ of the field of practices under question — ‘either discourse as part of the activity, or discourse in the reflexive construction of the practice, or both’ (Chouliaraki and Fairclough, 1999: 61).

Lilie Chouliaraki and Fairclough (1999: 61) identify four ‘moments’ of practice as a helpful way of analytically dividing the complexity of the social world. Material activity (physical acts such as bombing, building, etc.), social relations and processes (class relations, the military–industrial complex, etc.), and mental phenomena (beliefs, values, etc.) all flank the discursive moment of a field of social practice. Such a framework can help especially at the reflective stage of trying to picture how discourse might be able to have an effect, and thus where to look in our research. In the phraseology of Theo van Leeuwen, discourse ‘recontextualises’ social practice by representing it in particular ways — deleting elements of social practice, rearranging or adding aspects (2008: 17–21) — and thus often ascribing to practices new justifications or understandings that then shape their future iteration (2008: 5–6). But only a prior conceptualization of practices based in existing theory can allow for a fully causal attempt at understanding the important ways that discourse might be doing this relative to a particular social field. In a sense, CDA simply demands a step further than PDT. After understanding how something like humanitarianism is constructed in war, the analyst also seeks to understand through empirical analysis how this particular construction influences the order of discourse and the extra-discursive practices of which it is a part — and/or vice versa — rather than noting the construction and suggestively placing it within some context of an event.

Fairclough (2003: 31–32) advises the analysis of ‘genre chains’ as a way of tracing discourse through to this second step. After an initial important communicative event — for war, an elite political speech or document — is analysed for how it tries to reshape or reproduce an order of discourse, media or other political reactions and recontextualisations of this event and the order of discourse are traced in real time so that aspects of some future social event can be attributed to this process of discursive structuring. IR scholars attuned to the macro-events of history might bring to CDA a greater emphasis on such procedures by tracing and comparing between discourse-mechanisms at theoretically connected points in history (i.e. a proposed common ‘effectivity’ of mechanisms on events of interest — see Steinmetz, 2004). The CDA process would be calibrated on episodes of interest (cases), within which we would identify a number of texts besides the communicative event which represent societal responses to it and the social field it addresses. As the capper to analysis, we might analyse the official justifications at the time of some paradigmatic war-event that directly follows our genre chain. Combined with an analysis of the actual war practices for their character in relation to social theories of war, the analyst gains insights into the causal ability of a discourse and its ‘discourse circle’ to alter the order of discourse in one way or another, and connects this to theoretically salient features of extra-discursive practice.

This is only one possibility for a CDA-guided research design addressing the empirical question presented in this article — and one influenced heavily by a Faircloughian CDA approach rather than the numerous others which could also likely be useful — but it is one that PDT would not afford. PDT works often establish the effective nature of a discourse by creatively reading a wide number of prominent texts for the logic within some a priori chosen representation (e.g. Campbell, 1998; Jarvis, 2009). For instance, in Lee Jarvis’s (2009) quite interesting look at how a discourse of temporality helped to construct the War on Terror, his contribution to the stated research question of ‘exploring the functions and significance’ of temporality discourse proceeds through a wide reading of years of texts, only to expose and catalogue the discourse’s various manifestations. Thus Jarvis’s own claims that the Bush administration strategically deployed various temporal tropes as a way of legitimating the war begs the question of just how effective they were, if at all. It is an open question, for instance, whether attempting to construct the war as radically discontinuous with the past — one of Jarvis’s three major forms of temporality discourse — actually enabled the war effort, given that many critics at the time thought invading and protectorising two countries was an all too modern solution to a postmodern problem. Jarvis (2009: 20) refuses to take steps to tackle such issues because proposing to study how discourse affected some ‘extra-discursive reality’, implied by a concrete statement on the relative nature of actual events, is disallowed by PDT.

Likewise, without a dialectical view, theoretical tools not derived from or adapted to the meta-theory of PDT are simply out. As Jonathan Potter (2003: 785) admits, a poststructuralist view of discourse disallows the kind of analytical pluralism presupposed by a view of discourse as but one possible causal mechanism among many other kinds:

the choice of discourse analysis is not like selecting one dessert from an array of different and equally tasty ones. . . . Rather discourse analytic methods have been developed (and still develop) to encompass and address this active use of language. Mixing them with methods that presuppose a very different view of discourse is a recipe for incoherence.

Fortunately, a turn to CR and CDA moves the debate over the mixing of methods from the meta-theoretical level to the methodological task of adapting methods to the differing character of social objects (Wight, 2006: 259).

In contrast to studies of war or other IR phenomena within the PDT tradition, the CDA emphasis on the dialectic between extra-discursive and discursive practices opens the possibility for systematically addressing discourse’s causal effect on the actual extra-discursive practices of events like war. A CDA approach might allow chunks of time during war to be chosen by their bookending with a prominent communicative event and a prominent war event. In-between these bookends one would study genre-chains of related texts for evidence of the ability of the discourse, when deployed by an elite discourse-circle, to actually alter the order of discourse and feed this into the extra-discursive practices which play themselves out — the last assessment accomplished by adopting theoretical standards related to the phenomenon of interest. Repeated a sufficient number of times, such a procedure might allow the analyst to see patterns to the discursive–extra-discursive dialectic that lead to important generalizations able to speak to a scholarly audience interested not only in the nature of the discourse, but in what it might mean for the concrete practice of important IR phenomena.

Conclusion

In this article I have discussed the dominant form of discourse analysis in IR — that based on PDT. I have shown that it is inevitably constrained by anti-realist tendencies that would not allow it to offer real assistance in solving empirical puzzles such as the one used as an example here. Theorists either inclined or not to a discourse theory perspective on the social have failed to adequately analyse humanitarian discourse because of an abrogation of studying it in its inherent dialectical and causal relation to wider sets of practices within war. With help from CR, it was established how we might view discourse as a causal mechanism that could be put into relation with such extra-discursive practices. To actually analyse discourse as viewed this way, the article introduced the discourse analytic programme of CDA. This introduction was couched in terms of CDA’s meta-theoretical differences from PDT, most emphasized in its Faircloughian version, and their effect on research practices.

PDT demonstrates discourse’s ‘productivity’ by showing how a ‘regime of truth’ may be said to make something possible (Milliken, 1999: 236–237); in particular by a rather suggestive movement from assessed ‘subject-positions’ to actions (Epstein, 2008). In relation to CDA this is simply not a means of conducting ‘systematic, empirical studies of language use’ (Phillips and Jørgensen, 2002: 62–63). PDT-inspired empirical studies are nearly always illuminating works that convincingly show some logic within a discourse. It is merely that the ontology of PDT allows for conclusions relating to social events or actions to be derived only from an accounting of how identity must have been constituted in discourse at some moment, rather than emphasis on the tracing of discursive enunciations in real time. For those not wedded to a PDT outlook on the social world, and those who wish to connect discourse more forthrightly to its role in the generation of events, using CDA will allow for the study of discourse as a causal mechanism relationally connected to the others they find pertinent to phenomena of interest.

This is not to say that CDA is a ready-made template for illuminating discourse’s role as a causal mechanism. Indeed, critics of CDA rightly point to a disconnect between the methodological foundation of CDA as something that demands an explanation of the discourse–society dialectic, and the actual success of this in practice. But the foundation is there for this to be accomplished, and IR scholars who build from this foundation by thinking ontologically about the different elements at play in their subject matter, thinking about how these might demand certain analytic strategies, might help develop studies employing CDA that would more successfully show how discourse is influencing extra-discursive practice or vice versa. CDA’s strength is its foundational conception of the role of discourse when compared to PDT, but the challenge is to realize the pay-off of this strength in studies that help add to knowledge of pressing world political problems by understanding the discourse aspects of their generation in relation to knowledge of existing features.