Discourse Analysis Iver B. Neumann As a 22-year-old, I held a stray job fetching bundles of fur from minks, seals, polar foxes, blue foxes, and red foxes to show the buyers at Oslo Fur Auctions. The work altered the way I saw the world in other realms. Most striking was how one of my main interests at the time, looking at women, acquired a new dimension. Where I had previously focused on shape and movement, attire now became an important factor. My new interest in fur coats changed the way I sifted what I saw. Psychological experiments confirm my personal experience. For instance, children shown a cup with the handle turned away none the less drew a cup with a handle, because cups, by definition, are supposed to have handles. That is, the children drew the model, not simply what they 'saw' as a result of light waves hitting their optical nerves. People sort and combine sensory impressions of the world through categories (or models or principles). Language, as a social system with its own relational logic, produces reality for humans by mediating these sense data. These examples highlight that perception is mediated by aesthetics, sexuality, morals, or other modes (Bauman 1992). In order not to forget that these meanings are socially reproduced, discourse analysts call them representations - literally re-presented. (I will concentrate on the preconditions for and jobs undertaken by representations; see Dunn's chapter in this book for more detailed discussion of analyzing the components of representations.) Representations that are put forward time and again become a set of statements and practices through which £ertain language becomes institutionalized and 'normalized' over time. They may be differently marked in terms of how influential they are. In the United States during the Cold War, 'dove' and 'hawk' representations of the Soviet Union were both institutionalized, but so was the (even less changing) representation put forward by the American Communist 61 62 Discourse Analysis Iver B. Neumann 63 Party. When people who mouth the same representations organize, they make up a position in the discourse. Like representations, positions may be dominant or marginalized in various degrees. ^Demonstratinginstitutionalized djscoursj:jari_often simply be done bj^rqvingJ:hat_metaphors regularly appear in the same texts. In my study of European discourse on Russia, for example, I found a representation, which stressed that Russian females had been raped by Mongol and Tatar males for centuries, and that this had fostered a particularly wild and barbarous people ('scratch the Russian, and the Tatar will emerge'). This representation began to form fairly early, reached a peak in the inter-war period, and then lived a very submerged existence in European discourse. In the Baltic states, however, it was very strong indeed throughout the Soviet period and into the 1990s. The more such things may be specified empirically, the better the analysis. The ideal is to include as many representations and their variations as possible, and to specify where they are to be found in as high a degree as possible.. The first research task is to show the affinities and differences between representations in order to demonstrate^^ether theV'BeloTig^to^he same discourse. But repetition does not preclude variation or gradual re-presentation, so discourse analysis also seeks to capture the inevit-able cultural changes in representations of reality. For example, in the late 1980s, Russia was obviously heading for challenging times, and I reckoned that this would entail wide-ranging changes in relations with Europe. My basic idea was that, regardless of period, Russia's relationship with Europe had not been straightforward, yet it seemed set to remain central to Russian foreign policy as well as to Russian self-understanding. I wanted to be able to say something general about prerequisites for Soviet/Russian foreign policy in a situation where so many things seemed to be in flux. Discourse analysis is eminently useful for such analysis, because it says something about why state Y was considered an enemy in state X. how Wi£IRSgMAL3. PPjittcal opU aside. Because a discourse maintains a degree of regularity in social relations, it produces preconditions for action. It constrains how the stuff that the world consists of is ordered, and so how people categorize and think about the world. It constrains what is thought of at all, what is thought of as possible, and what is thought of as the 'natural tninf'tn dn in a given situation. But discourse cannot determine action completely. There will always be more than one possible outcome. Discourse analysis aims at specifying the bandwidth of possible outcomes. This works the other way, too; discourse analysis may also start with a specific outcome and demonstrate the preconditions for it happening, demonstrating concurrently that the outcome might have been different. *7 To map these patterns in representations, discourse analysts examine utterances. They may be texts (written statements that do some kind of work in a context). However, any sign - a semaphore, a painting or ajrrimace - may be analyzed as text, because it conveys meaning in a parHcularj:ont^xi^ince we 'read' societal processes aTlhe fmictional equivalent of texts, one may, for example, cull data from ethnography (see Gusterson in this book). Due to limits on length, I will focus in this chapter on written sources. (For an example of discourse analysis of ethnographic data, see Neumann 2007.) Acquiring a certain cultural competence is a prerequisite for discourse analysis, as for most qualitative methods. After discussing the need for basic language skills and historical knowledge, I divide my lessons for method into three concrete steps. First, one needs to delimit the discourse to a wide but manageable range of sources and timeframes. From these texts, the analyst then identifiesthe representations that Comprise the discourse, taking into account censorship and other practices that shape the availability of text. Finally, to explore change, one uncovers layering within the discourse. The more actions that the analysis may account for by demonstrating its preconditions, and the more specifi£ally_yiis_may be dpne_Jthe better the.,discourS-eni analysis. Prerequisite: cultural competence I always encourage students to draw on extant knowledge when they choose their topics; it saves time, and they start out with a competitive advantage. It also provide^a degree of 'cultural competence.' For example, I had done my conscription at the Norwegian Army Language School, where I studied Russian. Then I lived in Russia for half a year and did university courses in its history and foreign policy. All this gave me a certain cultural competence when I set out to research Soviet discourse as a doctoral student at Oxford (later published as Neumann 1996). I knew the Russian language, genres of relevant texts, and something about the general social and political setting (such as when Russia was at war with other states that it considered to be European and the extent to which European history and language were taught in schools). Thi^cujtural qomjjeterice enabled me to use tools of discourse analysis to demonstrate variations in meanings and representations. The more in-depTn^nT^enerai knowledge, the easier the specific research. For example, I knew that many Russian newspaper articles were divided into 64 Discourse Analysis Iver B. Neumann 65 two parts: a first part that repeated the so-called main line, then a part that dealt with new material that still had not been sorted in relation ' to and assimilated by that dominant representation. What is crucial < for the discourse analyst is the separation of these two parts by one codeword, odnako, which is best translated as 'however.' If one knows such conventions, the reading of texts becomes easier: I could rush ' through part one, which is a simple re-presentation of an already known reality, and concentrate on part two. Similarly, the expression en principe in French signals that one is putting forth a representation which one generally shares, but from which one nonetheless is going to deviate. , Of course, some things may be learnt on the job. As a British-trained j Norwegian Russian specialist, I needed to work at mastering phrases like j 'to go' and 'drag it through the garden' to buy a hamburger in the United * States. But there are other things that you have to know before you can | start. When I turned to the analysis of discourse in the United States, ; it was inconceivable for me not to know references such as 'I have a , dream' (a speech by Martin Luther King, Jr), 'beam me up, Scotty' (a line from the television show Star Trek) or 'I pledge allegiance' (to the flag). Thejx>int is that a researcher needs a basic level c£cultural competence : to recognize the shared undersfahcTings that create a commoirframe^of ^ reference, jwhich.makes it possible for people to acUn.relatipn to one t -another. Let us not forget that the analyses we write up are written for somebody. What is adequate cultural competence for a specific discourse ] analysis hangs, among other things, on whether the resulting analysis i may tell the intended readers something new. Ideally, a sdejjtificjejct should tell every conceivable reader something new.'TTTat is a situation that is very rarely reached, however. The world is full of researchers who * produce texts that do adequate jobs in adequate settings because they j are new in those settings, and not necessarily anywhere else. There is a trade-off with cultural competence. Culture appears to be ' shared. Close up,_it. turns, out not to be.. Phrases may mean a number » of different thing&jgr,they may be used without the user knowing all their cultural references^ or implications. The challenge is not to get natuxaJized..^,nQt,jtp 'become' pajt of the universe studied - but to denaturalize^ If you are a native speaker and know a culture as only a,jaatiKe.canr then you..dajiQlJia3aeJ±Lat marginal, gaze wherg_things look strange enough .to present themselves as puzzles. "You will also lose touch with your own biases. you\3emmewh^LM]^l9J9^S&ls^ call 'hame-iilind.' For example, I once submitted an analysis of US foreign policy discourse which used a quote from the then chairman of the ' Senate's Foreign Relations Committee, a key Republican senator of long standing, as representative of American discourse. The two reviewers, who wrote flawless American English, objected to my treatment of him as an elder on the grounds that they considered him out of touch and a crackgot, respectively! These readers were definitely 'home blind.' It is fully possible to do discourse analysis in the culture you know best, but you still need some kind of distance. You cannot be too much at home. An historian or anthropologist would at this point ask, whose representations, whose culture? (See also Leander and Ackerly in this book.) We are talking about cultural competence regarding the culture that spawns the representations to be analyzed, not necessarily for other related cultures. When I was done with my discourse analysis of Russian representations of Europe, I noted that I had documented what I held to be so much arcane and downright silly Russian representations of Europe that I felt I owed it to the Russians to analyze European representations of Russia as well, presuming that just as much arcane and silly stuff would crop up. (It did; see Neumann 1999.) For that analysis, I needed neither Russian nor much knowledge of Russia. Instead, it was important to know German, French, and English. It was a problem that I could only cover Spanish and Portuguese representations in translations. But I still felt warranted in talking about European representations of Russia, for there were strong regularities between German, British, French, and Scandinavian representations of Russia at any one given time during the last 500 years that presumably could be generalized to 'Iberian representations.' As in any other research, this lacuna has to be stated, and it will serve as a challenge to new researchers. (I have tried, so far unsuccessfully, to get a doctoral student to write about Iberian representations of Russia.) Methodologically, this points to the importance of being explicit about your sweep: the bjmderjt is, the more general knowledge you need, and theje£sjis^ No good Russianist would assume cultural competence about Serbia, and old cultural competence from the Soviet era may not necessarily be applied to Ukraine after its formal political separation from Russia. Knosyjn&thf .fiypr-rhan#^ may be as important as knowing its contents. Step one: delimiting texts Discourse analysts read texts. But what texts? In certain cases this is a simple question to answer. If one is to study party systems, then party 66 Discourse Analysis Iver B. Neumann 67 programs, election laws, and articles as well as speeches by party leaders are typical primary materials. Still, the quantity of material is usually enormous, especially if one includes the secondary literature. It is crucial to draw some lines, but problems of delimitation are inevitable.JThe_ choices applied to each individual discourse analysis always have to be defended. For example, if one repeatedly finds statements such as 'scrape a Russian and the Tatar will appear,' it would be mistaken to omit representations of Tatars from an analysis of Russian identity. A given discourse cannot be entirely detached from all other discourses. They are ordered and scaled in relation to one another. Russian identity, therefore, must be studied as something Russian and something non-Russian. However, which relation or relations to study -between Russia and Asia, Russia and Europe, Russia and Germany, Russia and Tatarstan, Russia and the Jews, Russia and the feminine - is not given. Ideally, all should be covered. In practice, that is rarely possible. The choice of which relation(s) to single out may be theory driven (let us selfwhat happens if we bring a feminist standpoint perspective to the study of Russian identity and look at the constitutive role of gender), utilitarian (I need to illuminate the identity aspects involved before I can get a handle on Russian-German energy relations; how do Russians think of Germany in general?) or ludic (my own favorite: why is it that Russians treat me the way they do? This must have something to do with general Russian ideas about Europeans.). Insofar as politics is a struggle between named groups and people, politics is conflict. Conflict should therefore attract the analyst of political discourse. One will often find direct references to texts that are being attacked. It is usually apparent who is attacking whom. When there is such a racket, it is because something new is happening, something that is meeting various attempts at limitation from those who dominate the discourse (see Lukes 1974). However, the puisuit.of commotion can be a methodological problem, since realities are maintained by the frequent repetition and confirm-ationjit represe^ta^^^^ does not mean that the discourse in question is non-conflicfual. OjaLhisJto use more time~and mentaT energy to'work out how and why things remain unaltered. ConcentratiBg gn the texts that produce thej^atestjacj^et mi^hj_inean that one. a^omatically j^vlleges__ the dominant, tepres-entation, which usually,vvill be the.loudest (Waever 1999). Some texts remain" unpublished when censorship is successful. Challengers may remain undetected for other reasons, including socially distributed lack of writing skills. One may also turn this around: publications that only repeat or incrementally expand the main representation tend to pass I relatively quietly. If one fails to detect these processes of power, then the J analysis easily becomes a shallow one of the boundaries of the discourse ! and its domination. ' Also, social and political life is full of cases where somebody writes something new and intriguing, with no immediate reception whatsoever. It may simply be that the textjs so new or differentJn_xejation ». tcnj^t4]j3ea4yr,(^tej|iant goes unnoticed for this very reason. There are existing texts as well as future texts^"atl^rsu!Kf ITBsTateTB a * text from a relatively obscure source becomes central - as did Francis * Fukuyama's 'The End of History?' in The National Interest - then it is a | research task to demonstrate how the text overcame the odds. Some Jtexts vMjhow MR. a§_crossroads or anchor points, such as short government treatises outlining policy (called white papers in most { English-speaking countries). These are called canonical texts or monu- mejaiS_(compare Laclau and Mouffe 1985 on nodal points). In my dissertation research, I was actually able to identify the textual canon by starting with the secondary literature, because it proved to be well » informed. I took the 'monuments' to be the works that were generally cited in the secondary literature. I read these works, and indeed I found that they tended to refer to one another. This, as well as the negative finding that there were few additional central texts, confirmed them as monuments. It is useml to_select texts around these monuments, since monuments also contain references to other textsTwhich"again pointed me to others that_were related. One discovers that some texts are 'canonical/ in the sense that they have a broad jecegtion and are often cited. If one identifies these texts, reads them, and then reads the central texts that these texts in turn refer to, soon one is able to Jdentify the main positions and versions. In most contemporary Western nationality discourses, for example, the representation of history for political purposes is widespread. However, it is not always possible to goback to antediluvian events, so one^mjgtdeiimjtthe timeframeTFor example, once I had my dissertation topic, I read up on the secondary literature in order to identify cut-off gojnts^ An obvious one would have been the coming to power of the great Europeanizer Peter the Great in 1694. In order to trace discourse in more depth, I chose the Napoleonic wars that really brought Russia into the heart of European great power politics, and treated the period from 1694 to 1815 cursorily as a prehistory. The other cut-off point presented 68 Discourse Analysis Iver B. Neumann 69 itself during the work, as the Soviet Union split up in the autumn of 1991. In specifying the sweep of the analysis, it is also important to keep in mind your reader. I later did a discourse analysis in my native Norwegian on Norwegian representations of Europe (Neumann 2001). I tried to tackle the question of home blindness by going way back in time -who is really 'at home' in the Middle Ages? In this case, the main intended reader was an informed Norwegian. I therefore needed to be fairly detailed in drawing up representations from the last 50 years. Yet I did not present context that was already fairly well known, which would not be particularly interesting to the prospective reader. When I did a shorter version in English (Neumann 2002), the intended readers were different, so I dropped detail and filled in context. A doctoral student in Europe, who has little idea who his readers will be, will tend to write differently from an American student, who has a committee from the outset. And how do you weigh writing for your examiners against writing for a more general audience that may also be interested in the texts? There are authorial decisions to be made - different strokes for different folks; broader ones for non-specialist foreigners, dense professors and academics working in outer disciplines. Participants themselves also delimit their discourses. For example, medical diagnosis relies upon the definition of diseases and syndromes, upon which doctors draw. Analyzing the struggles over these definitions, and the process of getting them registered as such, form part of the research. If the chosen discourse is international intervention (to distinguish from medical interventions), then the struggle over the characterization of certain policies as 'humanitarian' is decisive. The main task is to dig out the production of specialized knowledge. In analyzing Norwegian human rights law, for example, there will be a number of relevant texts in legal journals and government policy papers. One can compare related professional discourses in other countries. However, the connection to general political discourse may not be explicit. Some texts can acquire importance from the medium through which they are published. For example, a private letter from the 1830s threatened the dominant Russian representation of Europe after it got a wide reception through the circulation of copies in the saloons of St Petersburg, even though the author was declared mentally ill and incarcerated. It is important to bear in mind the values which different media give texts. If one is to carry out a discourse analysis of peace onera-tinns in the 1990s, it is important to distinguish between those journals that aim at operative milieus (Foreign Affair'; or Sumval), those that are written for a more general audience (International Affairs in Europe), and those that are mainly read by academics (International Peacekeeping). But what if there is a Russian letter or unpublished manuscript from the 1930s, unseen by more than a handful of people, which projects a representation of Europe that makes my analysis incomplete? In terms of the history of ideas this would be very interesting, precisely because of its originality and its lack of reception. Its discovery would provide a more accurate definition of the borderline between possible thought and the communication of that possibility. In terms of politics in the 1930s, however, it would be a non-event, because the analysis concerns texts that are socially communicated. What if it turns out that there are a number of texts that are systematically overlooked, which jointly document that there was a main representation that previously had not been included in the analysis? In the area of women and war, one can at least imagine the possibility that a systematic reading of all available sources on the national service in Norway written by women would result in a revision of previous views of the national service institution (see Ackerly's chapter on subaltern discourses in this book). Mikhail Bakhtin's concept of genre is useful. Genre carries its own memory, in the sense that every text relies on its predecessors and carries with it their echoes. If previous analysts have for si?nieje^on.oyerlooked an entire genj^Vthen it is animportant research task to cast light upon how this has happened. This will change the way wÍĽI£2iÉI!]ííš^ is politically relevant to today's situation. Excavate one text on women and war, and you have an idiosyncratic voice and an indication that a group has not met the preconditions for action to make itself heard. Excavate many, and you have documentation that an entire group has been silenced. It is also possible that there are too few texts published, making it difficult to get started. One can carry out a discourse analysis of material that has not been in general circulation (for example, of classified material). If the reason for the lack of text is the novelty of the specific discourse, with, for^^mpiej^mlx^U^ include a small text-based analysis of this material in an analysis that also draws ugoajBJtjer methods, of data, collection, for example interviews, surveys or participant observation. When does one have enough material? The ideal situation is that one covers a maximum^f^eventuajitjes, by reading_as much as possible frgmas many genres as possible. Foucault insisted that one should 'read everything, study everythingJl.Thic 'S not feasible in practice, and there will therefore always be a risk that some relevant texts are not included. 70 Discourse Analysis Iver B. Neumann 71 However, almost regardless of the extent of the discourse, relatively few texts wiH constitute the main points of reference. Therefore at some point one has to beabletod^dde^^ has not read everything. Only if a text emerges that cannot be subsumed under one of the main positions must the analysis be adjusted - or perhaps even rewritten entirely (see Hansen 1997). Step two: mapping representations A discourse usually contains a dominating representation of reality and one or more alternative representations. Discourse analysis therefore is particularly well suited for studying situations where power is maintained by aid of culture and challenged only to a limited degree, that is, what Gramscians call 'hegemony.' Structuralists and post-structuralists disagree over whether one can take a small part of the discourse and read it as symptomatic of all representations. Post-structuralists find the notion of a latent structure simply too deterministic. One must think flow, not control. Theja.sk is to search out and identify these various representations and possible asymmetries betweenjhem. TJTejmaJyjit_accepts and works with the inherent conflict between represpntai-inns Monuments frequently position themselves in the discourse by referring (adversarially or sympathetically) to texts that were previously considered monuments. Reading monuments in Russian foreign policy discourse, for example, helped me identify adversarial representations (for instance, 'Europe is vital, we should learn from it' versus 'Europe is rotten, we should isolate ourselves from it'), since these texts, often written at the same time, referred directly to one another. The advantage of a marginal position emerges clearly here for setting up an inventory of representations. Researchers question how uncertain or challengeable a given representation is. The limits of discourse are inscribed with varying means and degrees of violence. If there is only one representation, the discourse is closed. This of course does not mean that it is not political, because it takes a lot of discursive work to maintain a situation where this representation cannot be challenged openly. If moves to do something new by the text-writer are not successful, it is not necessarily because the discourse is successfully policed. On the other end of the spectrum, the field can be said to be open if there are two or more representations and_ none of them are don^at_ing;_(See Leander's chapter in this book on the boundaries of Bourdieuan fields here; historically Bourdieu formulated his theory among other things as an extension and correction of Foucauldian discourse analysis.) YetJ£is_djjficiilt_Lo. imagine a discourse that is_entirely oj)en or closed pyej time. Social relations will always be inJ^nie^degreejjfjljax^ Therejs a second problem in addition to specifying^ the discourse's degtee of openness. On the one hand, the number of 2enmotati_ons of relevant signs is endless, so the range of meanings is in principle infinite. On the other hand-L politics^ involves contestation between relatively clearly defined positions, which compete to find resonance among a illS^LSjLEHli^.-!^!^""5 it: is aesirabie to identify these positions. Typic-all^on£..pjQ.ali^^ will challenge it on certain points. The dominant position will either pjesenMtself as.being the way things have 'always' been (for instance, a democrat: humans are born free) or hark back to an idealized beginning (a democrat: Athenian democracy broke out of benighted despotism). Terms mean different things in different epochs, but carriers of a position will tend to tap the advantages of having a long (and presumably dignified) history by acting as if this were not the case (Koselleck 1988). It is important that the discourse analyst start with the representations themselves - the stories of howtfnnj^have^ 'always' been like this or that. For example, Athenian democracy was hardly a democracy by the lights of the 21st century. Neither was the ante-bellum United States. Arguing that every man is born free and has rights while having a number of living beings around who visibly are not born free and have rights (as slaves, or women, or children) reveals that the discourse is not open to the possibility that 'man' may be someone other than an adult, white male. Within the boundaries of his own political discourse, thus, it was not a problem that George Washington remained a slave owner throughout his adult life. However, a good discourse analyst should^ajsojpe able to dejngnstrate that where the carriers of a position see continuity, there is aljno^aJvvay_s_ change. Because of the nature of politics as a structured activity between groups, a discourse is politicized, precisely through the evolution of two or a few patterns of meaning, which is the discourse analyst's task to uncovex.Jt is possible to distinguish between the basic traits of such a meaning pattern (what unites the position) and varieties of it (what differentiate it). ,In principle, the discourse will carry with it the 'memory' of its own gcniesis. Showing how'each text is made possible by the preceding texts, often it is possible to find._.a. prehistory to the main representation. It is, for example, hard to think of Stalin's funeral oratory for Lenin without having the model of the Russian Orthodox oratory in mind. 72 Discourse Analysis Iver B. Neumann 73 Methodologically, this is significant because, as. a given representation esj^sb^jtselUnj^ j texts' that foreshadow it "This allows us to make a prediction: if a new ] main representation of Europe surfaces in Russian discourse during the ' next years, more likely than not it will be churned out of material that is already present in the discourse. Tlipre are a pumb.er of formal and informal practices that determine which representations are allowed into the discourse, and that make it posslbiFfor thTanaiyst to map meaning are legal systems and censorship, whereby sanctions against violating the t boundaries of the discourse are threatened explicitly. An example: in Norwegian nationalist discourse of the 1990s just using the word 'race' activated a set of sanctions, foremost among which are laws that prohibit what Americans call 'hate speech.' The fact that there is no comparable Norwegian concept for the phenomenon, and that the American term is used regularly, are data for a discourse analysis of 'race.' (See also Klotz's discussion in this book of the concept of race in case selection.) One can also examine what kind of self-censorship different types of mass media apply and what deviation it takes to provoke more formal sanctions. Legal verdicts on the borderline between incitement to violence and freedom of speech, and the debates surrounding it, would be one of several clues. To study nationalist discourse in the Soviet Union in the 1930s, where every newspaper, radio, and television station sifted what was printed and broadcast, one must start by examining the formal censorship instructions. Thereafter one might look at what unpublished and imported texts circulated, and what incidents resulted in Gulag sentences. One should not overlook cultural artifacts with a widespread, so-called popular culture (see Dunn's chapter in this book). Discourse analysis is, for example, a useful way to examine film, understTOSTas text. Rather than looking at museums, one can look at the reality production that happens in soap operas. If one is to examine the reality of 'Germany' in British discourse, then in addition to cases such as bilateral political discourse, EU discourse, and so on, it will also be of interest to look at representations of Germany in magazines, pulp fiction, and imported B movies (where it is still not unusual to find narratives where German Nazis are the crooks). I.would argue that the discipline of International Relations is not at present.paying enqugtrheed to artifacts of popular atltairglrbut..iuch_ an^analysis must be situated, in the sense that one must be able to pjomtjjutjhejnter-relation between representations of, say, Germany in popular culture and political discourse about Germany. How does popular culture appear in and relate to political discourse? To what degree do representations from the former result in truth claims in the latter? 'Situating' (showing where something can be found, where it is in situ, 'in place') omj^specified as_pjoving jnter-textuality between expressions/ texts, and discourses (see Neumann and Nexon 2006). ^fincj[raphy^^ are similar jpjhat the^jpayj or should pay, a lot of attentionto how thejyialyjit is sjtuated injelation to the data. In the 1980s, a key development in ethnography was an intensified attention to the writing up of the ethnography, and this turn was directly inspired by discourse analysis (Clifford and Marcus 1986). Typically, however, discourse analysis would splice data collection methods such as fieldwork or memory work with the analysis of writtenJexYs.lFwouTd also typically turnjowritten texts first, and think of other data collections^ or substitutionary. Certain analysts are more formal in their mapping than others. I see heuristic value in being stylized. When discourse analyses are highly formalized, however, I always ask myself whether the reason is a need to appear social science like in order to get published, or whether it is actually an urge growing out of the text itself, whether it is necessary, and whether it is a market-driven or a scholarly necessity. Step three: layering discourses Not all representations are equally lasting. They differ in historical feBJJkJILJg^ m the discourse. The third task for the discourse analyst is to demonstrate this. The production of gender is an example. There are a number of biological and social traits (diacritics) that line the boundary between the sexes, from the presence of ovaries to ways of brushing hair away from one's eyes. Few can be counted as unchangeable. However, some are more difficult to alter than others. It is easier to neutralize the gender-specific aspect of the sign 'unremunerated domestic labor' than 'childbirth.' At this stage, some discourse analysts would cry foul, because they would like to insist that everything is fluid, and that nothing should be reified in the analysis. I agree that everything is fluid in principle, PJJtjhjjjginthere is that not everything is equally fluid^£urthermore, it is impossible to analyze ^something wrthouj^reifyinR_something else. Indeed, as my initial example of the child perceiving the cup is meant to bring out, it is impossible to see and to live without reifying things. 74 Discourse Analysis Iver B. Neumann 75 We have to subsume new phenomena into already existing categories in order to get on with our lives. Arguing that everything is equally fluid makes it impossible to analyze something in its social context. It also goes against what seems to be the very physiological preconditions of our existence as Homo sapiens. CertainreDresentations in a discourse wy1^^.^^?]^?^*0..?.^?®!. th,an others. Signs that are 'good to think with' (Levi-Strauss 1963) and representations of material objects will often be among these. However, now physical reality turns up. Put in everyday speech: material objects are difficult (though not impossible) to 'explain away.' But for the study of human behavior, this is not a problem. As Laclau and Mouffe illustrate, An earthquake or the falling of a brick is an event that certainly exists, in the sense that it occurs here and now, independently of my will. But whether their specificity as objects is constructed in terms of 'natural phenomena' or 'expressions of the wrath of God' depends upon the structuring of a discursive field (Laclau and Mouffe 1985: 108). Meaning and materiality must be studied together. It is possible to take as one's starting point for a reading of a social event, such as the reasons why Sweden went to war in 1630, that there are a number of material 'facts,' including archaeological objects. Any valid representation of the social event must relate to and at the same time study the various representations of the social event without having to hunt some kind of 'truth' about it beyond accounting for these_obi£Cls (see Řingmar 1996; Neumann 1997). The question is what the scope or degree gf.sQgiaL construction is in the relationship between 'fact' and 'representation.' We should expect greater 'inertia' in the representation of material objects than that of other things, but this still does not ensure the place of the objects in the discourse. This issue also lays bare the metaphors on which the discourse approach rests. Foucault wrote about archaeology and genealogy, the basic idea being that of things emerging, with some things remaining the "am£x. 9_nfl .."tfrťT.S rh.3Pging- An archaeological site will contain certain artifacts that tell of continuity - there will be shards of pottery and traces of funeral rites - and these will vary with the period. But, in a particular site, certain things will remain stable whereas others change. The key, in archaeology as in social analysis, is to specify, what changes and what does not, and how. The same is true of genealogy, the basic meaning of which is that you start with one human and trace his or her ancestry. You will tend to find people who become less and less interrelated to one another the further back you go. At some stage, all they have in common is that they are all the ancestor of that particular human. If-some traits unify and some differentiate, it is reasonable to think tMLthg.traitg.jhat unite are more difficult to change. For example, if one chooses to study German identity, one will find endless variations on which things are thought to be German. If one looks at the question of how the state is related to the nation, the range of meaning will be lesser, perhaps only covering two possibilities: one, that the nation defines the state by being its cultural carrier, Kultumation, or second, that nation and state are both anchored in citizenship, Verfassungspatriotismus (see Waever 1999). In my doctoral thesis, I approached this question of layering by postulating explicit and implicit family resemblances across time. The element of Europe as a place to learn from was in evidence at all points in time since the latter half of the 17th century, except for the High Stalinist period (two decades from the early 1930s onwards). In later work (Neumann 2004: 21), I formalized this step by drawing up a model of Russian discourse on Europe across time, using three layers: basic concepts (state, people, and so on), general policy orientation (isolation, confrontation, learning, and such), and concrete historical examples (pan-slavism, Bolshevism, early Yeltsin years, among others). At the level of the broad historical sweep, such a mapping of preconditions for action is the endpoint of discourse analysis. As should be clear by now from the discussions above, however, there remains endless work of specification on different constitutive relations, close-ups of specific time periods, tailor-making of the analysis to illuminate specific (types of) action, and so forth. Conclusion: a discourse analysis toolkit If one should fashion such a thing as a discourse analysis toolkit, it would perhaps look like this. Tool one would be a carver that would carve_texts,out of the social world Tool two would be an eq^'alizeftnat Q3afe_£*gL£henojnen£_(for example, a semaphore, an ad, a body) into material to be analyzed on a par with texts. Tool three would be something like a h^ciirff dog that would group thlse^n^onTeTO together based on them being about the same thing. Tool four would be a slicer, cutting the phenomena into different representations of the same thing. Tool five would be some kind of optic device that would 76 Discourse Analysis Iver B. Neumann 77 make visible the meaning dimension of the material phenomenon to its users. It would come with a grading spectre that could demonstrate how easy it would be to change the different layers of a given phenomenon. And finally, the only one that I would really like to see on my desk, tool six would be a self-reflecting quill that accounted for my own weighting of the phenomena of which I wrote as I wrote. The point of such a tool kit would be to help us understand how the seemingly unchanging and 'natural' stuff of which our social worlds actually emerged as a creation of human history. Discourse analysis makes the social world more transparent bv demonstrating how fts elements interact. By demonstrating that things were not always the way they appear now, discourse analysis makes us aware that they are most probably changing as we speak. In order to account for global politics, therefore, it is not enough to study what one clerk wrote to another, how statesmen pontificate about the policies they pursue, or the technological changes that make for different kinds of warfare. The study of the meaning which these different phenomena have to those concerned also has to be included, and this means that discourses should beaccessed^ at many djiferent_points. Representations are constitutive in determining what is sensed and communicated, but they do not necessarily come with 100 percent built-in guides for action. If one has, for example, mapped Russian discourse on Europe, one has demonstrated several preconditions for foreign political action, but one has not necessarily cast any light directly upon the specific processes in the determination of such action (see Neumann 1996). A representation can make roojmfor several different act ions, and its carriers can be more or less conscious in their relationship with this representation. An analysis of representations of Europe will thereby not constitute an exhaustive analysis of Russian foreign policy. To do that, one must not only systematize the analysis of those sanctions that follow deviance, as I have already mentioned, but must also look at a number of other aspects of the materiality of discourse. To the extent that a fuller understanding of where we are and how we landed here is helpful in getting us somewhere else, discourse analysis may be 'useful' for solving problems. But it is not your first choice in a tightly scripted situation, such as answering why state X went to war against state Y at point Z in time. Rational choice may be fine for that, even though the assumptions of the two approaches are very different indeed. An a_naly_st ..may usedisccnirse analysis in order to study how structures produce agents, arul J hen decide,to Trce/.c' agents at a .specific point in time, for example at the outbreak of war. The analyst may then change tack and analyze the outbreak of war drawing on social choice theory. Such splicing of methods is highly unusual, among other things because few analysts are fluent in such diverse methods, but also since the analyst's own identity may be so tied up to one particular method as to make the very thought of mixing methods appear as sleeping with the enemy. To make a self-reflective point, why this is so may be studied by drawing on discourse analysis. Acknowledgment I thank Karen Allen, Stephanie Smith, my co-contributors and especially Audie Klotz and Deepa Prakash for a range of positive and productive comments on earlier drafts. ! \ t i I ! i I !