Critical Discourse Analysis, An overview

The aim of this paper is to see what Critical Discourse Analysis is. This implies scrutinising its origins, what it has meant to the academic world as a whole, how it encapsulates various trends with different theoretical backgrounds and methodological approaches, what are its limitations and its new developments. A simple practical example will show its potential.


Introduction
In this paper, I describe the heterogeneity of Critical Discourse Analysis (CDA), its power to attract and annoy, and its most exciting traits and weaknesses, which have caused debate and disagreement. More than two decades have passed from the analysis of excerpts to the study of large corpora, from allegedly interested selection to random collection of data. Its social implications encouraged its development. Leaving mere intuition aside and exploring the trace of ideology in texts other than literary ones contributed to its scientificity and helped broaden its scope.
I will pay attention to CDA as problem-oriented social research, founded in social history, semiotics and linguistics; to scholarly approaches that are also considered critical; to the objections raised against CDA; and to new trends trying to tackle its limitations. The question of what should be understood by critical is also addressed, with the aim of resolving misconceptions associated with this label. It is equally important to clarify commonly used terms, including text, discourse and context as well as others that have a central role in CDA itself, in particular, ideology, power, dominance, prejudice and representation. Further, because CDA has its origins in textual and linguistic analysis, I will address the question of why one particular theory of language, Halliday"s (1985) Systemic Functional Grammar (SFG), has been widely adopted by CDA researchers. SFG is not the only linguistic theory used by CDA practitioners and I shall comment on the other methods of linguistic analysis that have been applied. However, linguistics is not the only, or even the most important influence on the development of CDA, so I shall also give attention to others from sociology, social theory and philosophy.
Among the key authors in the discipline, I will highlight Wodak, Fairclough, Kress, van Leeuwen, van Dijk and Chilton, who represent the major approaches I refer to here. These scholars share interest areas such as inequality, control, literacy and advertising. While I cannot analyse one specific text from each perspective, I provide clues as to how it would be implemented and point to tools CDA has put forward to attack its allegedly fatal malady, overinterpretation.
This area of applied linguistics, which has variously been taken to be a paradigm, a method and an analytical technique, was originally known as Critical Language Studies (Billig 2003). It goes by various similar names. For instance, van Dijk (2009) prefers the term Critical Discourse Studies, suggesting that this may help see it as a combination of theory, application and analysis. The interest of this cross-discipline (van Dijk 1997) lies in attending to all types of semiotic artefacts, linguistic and non-linguistic. A central aim in all the various approaches is that critical analysis raises awareness concerning the strategies used in establishing, maintaining and reproducing (a)symmetrical relations of power as enacted by means of discourse. CD analysts focus on those features contributing to the fabric of discourse in which dominant ideologies are adopted or challenged, and in which competing and contradictory ideologies coexist.

What is discourse?
The first obstacle faced by newcomers to the field is the various definitions of the concept of discourse. In a modified version of a taxonomy by Bloor and Bloor (2007: 6-7), it is possible to make the following kinds of distinction: discourse-1 is the highest unit of linguistic description; phonemes, morphemes, words, phrases, clauses, sentences and texts are below; discourse-2 is a sample of language usage, generally written to be spoken, that is, a speech; discourse-3 refers to the communication expected in one situation context, alongside one field and register, such as the discourse of law or medicine; discourse-4 is human interaction through any means, verbal and non-verbal; discourse-5 is spoken interaction only; discourse-6 stands for the whole communicative event. Wodak and Meyer (2009) associate this diversity with three different trends: The German and Central European tradition, in which the term discourse draws on text linguistics; the Anglo-American tradition, in which discourse refers to written and oral texts; and the Foucauldian tradition, in which discourse is an abstract form of knowledge, understood as cognition and emotions (Jäger and Maier 2009). Gee"s (1999) pair small-d-discourse and big-D-discourse encapsulates these senses above cogently: The former refers to actual language, that is, talk and text. The latter, to the knowledge being produced and circulating in talk; to the general ways of viewing, and behaving in, the world; to the systems of thoughts, assumptions and talk patterns that dominate a particular area; and to the beliefs and actions that make up social practices. Chilton"s (2004) language L , language l , discourse D and discourse d are very much in the same line. Cameron (2001) does not use these labels but her meaning is comparable when she distinguishes between the linguists' discourse (i.e. language above the sentence and language in use) and the social theorists' discourses(s) (i.e. practice(s) constituting objects). In a similar vein, van Dijk (1997) proposes linguistic, cognitive and socio-cultural definitions. He first argues that discourse is described at the syntactic, semantic, stylistic and rhetorical levels. Secondly, he adds that it needs to be understood in terms of the interlocutors" processes of production, reception and understanding. And, thirdly, he points to the social dimension of discourse, which he understands as a sequence of contextualised, controlled and purposeful acts accomplished in society, namely, a form of social action taking place in a context (i.e. physical setting, temporal space plus participants). Since context is mostly cognition, that is, it has to do with our knowledge of social situations and institutions, and of how to use language in them, van Dijk claims that each context controls a specific type of discourse and each discourse depends on a specific type of context.
From Widdowson"s perspective (2004), texts can be written or spoken, and must be described in linguistic terms and in terms of their intended meaning. Discourse, on the other hand, as text in context, is defined by its effect. In his words, discourse "is the pragmatic process of meaning negotiation", and text, its product (2004: 8). Co-textual relations are concerned with text and contextual relations with discourse; that is, text cohesion depends on discourse coherence. CD analysts Fairclough and Wodak (1997: 276) refer to the following senses: Language use in speech and writing, meaning-making in the social process, and a form of social action that is "socially constitutive" and "socially shaped". The concept Fairclough finally opts for is semiosis, in order to include not only linguistic communication but also, for example, visual communication, as well as to generalise across the different meanings of the term discourse. Semiosis plays a part in representing the world, acting, interacting and constructing identity, and can be identified with different "perspectives of different groups of social actors" (Fairclough 2009: 164). Discourses can be appropriated or colonised, and put into practice by enacting, inculcating or materialising them. In contrast, texts are "the semiotic dimension of events" (ibid.), where we can find the traces of differing discourses and ideologies (Weiss and Wodak 2003).
The origin of the latter ideas can be tracked back to philosopher Michel Foucault (2002: 54), for whom discourses are "practices that systematically form the objects of which they speak". In their interpretation, Fairclough and Wodak (1997: 261) add that discourses "are partly realized in ways of using language, but partly in other ways", for example visual semiosis. Texts are the only evidence for the existence of discourses, one kind of concrete realisation of abstract forms of knowledge; at the same time, they are interactive and influenced by sociolinguistic factors. In the process of constructing themselves in society, individuals internalise discourses that comprise the core of a community of practice, in the sense that such discourses control and organise what can be talked about, how it can be talked about and by whom. Social practices are meaningful and coherent in that they conform to discourse principles. As manifestations of ideologies, discourses form individual and collective consciousness, and consciousness influences people"s actions; that is, through the repetition of ideas and statements, discourse solidifies knowledge (Jäger and Maier 2009), and reflects, shapes and enables social reality. Furthermore, it can be defined by the activities participants engage in, and the power enacted and reproduced through them; thus, we can speak about feminist or nationalist discourse, doctor-patient or classroom discourse, the discourse of pity, whiteness or science, or hegemonic and resistant discourses. To Foucault"s definition, van Leeuwen (2009: 144) adds that discourse involves social cognitions "that serve the interests of particular historical and/or social contexts", represent social practices in text, and transform or recontextualise them. As will be seen later, van Dijk places considerable emphasis on this notion.

What makes DA critical?
CDA is naturally embedded within Critical Theory, a paradigm developed in the last three decades whose critical impetus originates in the Frankfurt School, especially Habermas. As Wodak and Meyer (2009: 6) recall, in 1937 Horkheimer urged social theory to critique and change society, which meant to improve its understanding by integrating social sciences, to show how social phenomena are interconnected, to produce knowledge that helps social actors emancipate themselves from domination through self-reflection, and to describe, explain and eradicate delusion, by revealing structures of power and ideologies behind discourse, that is, by making visible causes that are hidden. The scope of CDA is not only language-based. Its critical perspective attracts scholars from various disciplines, as well as activists. Their concern lies with unveiling patterned mechanisms of the reproduction of power asymmetries. Anthropology, linguistics, philosophy and communication studies, among others, may share this inclination.
From its inception, CDA was a discipline designed to question the status quo, by detecting, analysing, and also resisting and counteracting enactments of power abuse as transmitted in private and public discourses. For some, to be critical might imply to be judgemental. However, this is not the case here, because, as Jäger and Maier (2009: 36) state, this kind of critique "does not make claims to absolute truth". CDA is understood to be critical in a number of different ways: Its explicit and unapologetic attitude as far as values and criteria are concerned (van Leeuwen 2006); its commitment to the analysis of social wrongs such as prejudice, or unequal access to power, privileges, and material and symbolic resources (Fairclough 2009); its interest in discerning which prevailing hegemonic social practices have caused such social wrongs, and in developing methods that can be applied to their study (Bloor and Bloor 2007). All this makes CDA an example of research aiming for social intervention. Fairclough and Wodak (1997) add that a critical reading goes beyond hermeneutics. In their view, CDA aims at demystifying texts shaped ideologically by relations of power; it focuses on the opaque relationship between discourse and societal structure; and it does so through open interpretation and explanation, by relying on systematic scientific procedures, that is, by achieving distance from the data and setting them in context. Self-reflection concerning the research process is a must. In sum, CDA seeks to expose the manipulative nature of discursive practices, and improve communication and well-being by removing the barriers of assumed beliefs legitimised through discourse.

The origins of CDA
The philosophical and linguistic bases on which CDA is grounded are certain branches of social theory and earlier discourse analysis, text linguistics and interactional sociolinguistics. Certain proponents of CDA are influenced by Marx"s critique of the capitalist exploitation of the working class, his historical dialectical method, his definition of ideology as the superstructure of civilisation (Marx andEngels 1845/2001), and his notion of language as "product, producer, and reproducer of social consciousness" (Fairclough and Graham 2002: 201). Some also draw on Althusser"s (1969Althusser"s ( /1971 conception of interpellation, which describes the way an individual can be aware of themselves as a constructed subject within discourse on their becoming part of someone"s utterances. Likewise, Gramscian hegemony (1971) influences a number of CDA scholars. It formulates the idea that power can be exercised and domination achieved not only through repressive coercion, oppression and exploitation, but also through the persuasive potential of discourse, which leads to consensus and complicity.
Habermas (1981) is frequently cited by CDA writers. His key contribution in the theory of communicative action is the notion of validity claims, which, according to him, are universally presupposed in all discourse. He further maintains that language can be used either strategically or in a manner oriented to understanding. In the latter, validity claims can be challenged and defended in a communication situation that is free from coercion, is only based on rational argument, and permits access to all who are affected by the discourse. These characteristics are absent from the strategic use of language; it is to challenging the strategic use of language that CDA pays attention. Foucault (1972), in contrast to Marx and Habermas, thinks that consciousness determines the social production process. Despite contesting the existence of an autonomous subject, he believes in the individual"s involvement in the practical realisation of power relations. Discourses are produced by all individuals, then, especially those who have the right to use all resources (Jäger and Maier 2009).
In the late 1970"s, the University of East Anglia nursed a new trend of analysis, as linguists and literary theorists were interested in linguistic choice in literature (see Fowler 1986). Later on, they would focus on other texts of relevance in the public sphere, especially the mass media. This did not mean only a terminological change (i.e. from linguistic criticism to critical linguistics). The new label, which is sometimes taken as synonymous with CDA, implied a new attitude in academe: The scholar"s commitment against social injustice. The East Anglia School proposed Hallidayian linguistics for the analysis of news texts (Hodge and Kress 1993). Language as social semiotic, the three metafunctions, and transitivity and modality became staples in this new discipline. Chomsky"s grammar (1957) was also appropriated since one of its main concerns is describing the implications of syntactic transformations: Passivisation and nominalisation have been the focal point of many a CDA work.

Examination of approaches to CDA
Notwithstanding obvious similarities, especially as regards agenda and scope, proponents of schools of CDA differ according to theoretical foundations or methodology. Some tend towards deduction and others proceed inductively. The former base their explanations on a few examples; the latter scrutinise a larger collection of data; without doubt, this can be more time-consuming but absolutely reliable and unbiased.
All in all, they all generally attend to categories such as tense, deixis, metaphor, attributes or argumentative topoi.
Fairclough"s Dialectical-Relational Approach to CDA is an essentially Marxist framework, anchored in his (1989,1995) research on language, ideology and power, where we find a very influential terminology, including dominance, resistance, hybridisation of discursive practices, technologisation of discourse and conversationalisation of public discourse. As Wodak and Meyer (2009) explain, Fairclough highlights the semiotic reflection of social conflict in discourses, which translates into his interest in social processes (i.e. social structures, practices and events). A pragmatic side of this approach is his support for critical language awareness, which to him is essential in language education (Fairclough 2007). Fairclough describes the following procedure: The scholar looks at one social problem with a potential semiotic dimension. This dimension is analysed by identifying its styles (or semiotic ways of being), genres (or semiotic ways of acting and interacting) and discourses (or semiotic ways of construing the world). Later, the differences between styles, genres and discourses are identified. Next, the researcher studies the processes by means of which the colonisation of dominant styles, genres and discourses is resisted. The focus then shifts to the structural analysis of the context, and the analysis of agents, tense, transitivity, modality, visual images or body language. Eventually, interdiscursivity is dealt with. Regardless of the apparent neatness of this methodology, Fairclough (2009) denies there is one single way of analysing any problem. Interestingly, he believes that, after selecting one research topic, the scholar constructs their object of research by theorising it. Its transdisciplinarity is one of the outstanding strengths of one approach where researchers may prefer (detailed but not always too rigorous) analysis of few data, selected, sometimes, by using somewhat unclear methodology and, to some extent, opaque style encouraging less critical thinking than one might expect.
Van Dijk"s Socio-Cognitive Discourse Analysis is an approach characterised by the interaction between cognition, discourse and society. It began in formal text linguistics and subsequently incorporated elements of the standard psychological model of memory, together with the idea of frame taken from cognitive science. A large part of van Dijk"s practical investigation deals with stereotypes, the reproduction of ethnic prejudice, and power abuse by elites and resistance by dominated groups.
Van Dijk also emphasises the control of discourse dimensions as a means to gain access to power. A further element in his account of discourse production and comprehension is the K-device, which is shorthand for personal, interpersonal, group, institutional, national and cultural knowledge (van Dijk 2005). Cognition, realised in collective mental models as a result of consensus, is the interface between societal and discourse structures (van Dijk 2009). While societal structures influence discursive interaction, in the latter the former are said to be "enacted, instituted, legitimated, confirmed or challenged by text and talk" (Fairclough and Wodak 1997: 266). Van Dijk (2009) believes CDA needs a model of context such as Moscovici"s (2000) social representation theory: One individual"s cognition is informed by dynamic constructs known as social representations, that is, the concepts, values, norms and images shared in a social group, and activated and maintained in discourse. He advocates the analysis of semantic macrostructures, local meanings, formal structures, global and local discourse forms, specific linguistic realisations and context. The aspects he focuses on are coherence, lexical and topic selection, rhetorical figures, speech acts, propositional structures, implications, hesitation and turn-taking control. Despite its power, in this approach, it is my belief that intersubjective agreement between scholars is not fully guaranteed by a slightly deficient explanation of how to apply some of the rules identified by van Dijk in discourse practice; thus, method and conclusions are open to multiple interpretation.
The Discourse-Historical Approach (DHA) (Wodak and colleagues) attempts, inter alia, to describe those cases where language and other semiotic practices are used by those in power to maintain domination (Reisigl and Wodak 2009). Initially, DHA was concerned with prejudiced utterances in anti-Semitic discourse. Recent developments include the discursive construction of national sameness and the social exclusion of out-groups through the discourses of difference, and the reconstruction of the past through sanitised narratives. The general approach reflects sociolinguistics and ethnography; it also gives an important place to Habermas"s notion of the public sphere and to strategic communicative action as opposed to ideal communication oriented to understanding. Its central tenet is the importance of bringing together the textual and contextual levels of analysis. The model of context used in DHA invokes historical knowledge understood in terms of four layers: (a) the linguistic co-text, (b) the intertextual and interdiscursive level, (c) the extralinguistic level, and (d) the sociopolitical and historical level (Wodak and Meyer 2009). The interconnection between various texts and discourses leads directly to the notions of de-contextualisation and recontextualisation, processes in which elements typical of a particular context can be taken out of it and inserted into a new context with which it has not been conventionally associated. DHA has further produced a series of analytical and descriptive tools, drawing on linguistic models and argumentation theory. In particular, DHA lists six strategies for identifying ideological positioning (i.e. nomination, predication, argumentation, perspectivisation, intensification and mitigation) which are analysed as part of a larger process that includes also the characterisation of the contents of a discourse, linguistic means of expression and contextdependent linguistic realisations of stereotypes. One of the strengths of DHA is the emphasis on the combination of observation, theory and method, and the continuum between application and theoretical models. Its historical, political and sociological analyses are also, in my view, an important part of its methodology, especially in relation to systems of genres, although the lack of a fully systematic procedure in this regard is one of its weaknesses.
The Duisburg School is heavily influenced by Foucault"s work. A particularly strong underlying conviction is that it is discourse that makes subjects (Jäger and Meier 2009). In other words, an individual"s sense of who they are arises from their imbrication in systems of historically contingent meanings communicated by institutionalised patterns of behaving, thinking and speaking. This kind of framework, sometimes referred to as Dispositive Analysis, also draws on social constructivism (Laclau 1980) and activity theory (Leont"ev 1978), and claims that social selves are constituted in a semiotic network that includes not only linguistic mediation of various kinds but also architectural arrangements, legal practices, customs, rituals, modes of moral thought, social institutions and so forth. Their notion of discourse is built upon "an institutionalized way of talking that regulates and reinforces action and thereby exerts power" (Link 1983: 60). While Foucault"s approach is relatively vague as regards discourse in its linguistic manifestation, the Duisburg approach pays attention to metaphors, references, style, implied meaning, argumentation strategies, the sources of knowledge, and agentive structures and symbols. Like DHA, the Duisburg approach advocates an analytic procedure. After selection of a particular subject matter, analysis is focussed on one domain, such as the media. This is followed by the structural analysis of one discourse strand ("what is said and sayable at a particular point in time", Jäger and Maier 2009: 46) and of typical discourse fragments, that is, the different topics each text deals with. Linguistic-discursive practices are explored through the analysis of texts; non-linguistic discursive practices, through ethnographical methods; and materialisations, through multimodal analysis and artefact analysis. The interest of an approach like this, concerned, inter alia, with everyday racism, patriarchy in immigration discourse or the discourse of the right, may be diluted, however, behind the imbalance between its complex theoretical apparatus and what may seem to look as only content-based analysis.
There is a prominent strand of CDA that advocates the use of Halliday"s SFG. This is the framework of linguistic description used by Fairclough, as it was also by Fowler et al. (1979) and Hodge and Kress (1988). In those studies that make use of SFG, different linguistic descriptions of the same piece of reality are claimed to stand for different constructions of that reality. Thus one and the same historically occurring event can be described as a riot, a demonstration or a protest; and social actors can be presented as agents or victims by selection of grammatical coding. More generally in this approach, text types represent social practices (i.e. regulated ways of doing things), which involve participants, actions, performance modes, presentation styles, times and locations, resources and eligibility conditions. Theo van Leeuwen (2009) has developed SFG"s formal framework for the classification of the semiotic system of social actor types and for the classification of the different ways in which social actors can be linguistically represented. According to this author, deletion, substitution, rearrangement and addition are the transformations that elements of a social practice undergo through discourse.
Recontextualisations add the what for and the why (or why not) of a social practice. In discourse, van Leeuwen hypothesises, social actors can be included or excluded; actions can be represented dynamically or not, as if there were no human agents or the opposite; we can generalise them, or make them stand for specific references, abstractions, symbols; as for practices or their elements, these can be set in a context, or reallocated. By making explicit the ranges of ways in which texts represent social actors, their actions and purposes, van Leeuwen seeks to analyse how specific discourses legitimise some of these actors and practices and intentions rather than others. His concern with an overall comprehensive analysis of complex semiotic phenomena (the language of images included), which is not yet complete, by way of linking various disciplines may be both one of its strengths and one of its drawbacks at the same time.
Finally, I will mention Mediated Discourse Analysis (Scollon and Scollon 2005), which highlights the role of ethnography and semiotics. As with the DHA, this emphasises the diachronic dimension, and texts are viewed as situated discourse (Scollon 2003). Further, Scollon and Scollon revisit Bourdieu"s notion of habitus (a system of permanent, identical characteristics, which, by integrating past experiences, mediates perceptions, judgments and actions), and develop it by appropriating Nishida"s (1958) concepts of action-intuition and the historical body (a combination of the social and the psychological). Scollon and Scollon (2005) claim that discourses are always present at any moment. They depict the individual as an actor "embodied in the society of various social groups" (Scollon 2003: 172). Subsequently, one of their goals is to find the link between individual action and public discourse, so as to achieve an understanding of how we internalise the social world and how, through action-intuition, the historical body of a social actor is externalised. I must admit that their optimistic standpoint as to the possibility of researchers acting in the world in order to make a difference in the actual world and their concern with problems in intercultural communication is more than exciting; however, some relative lack of very detailed concrete methodological guidelines may deter scholars.

Critiques of CDA
The merits and demerits of CDA research have been the object of a certain amount of critique. The problems that have been picked up concern context, cognition, partiality and the linguistic model employed. Most critics do not call into question the existence or epistemological relevance of CDA, perhaps with the exception of Widdowson and Chilton, but are aware of its shortcomings: Its theoretical foundations are quite tangled in many cases, and the use of concepts and categories may seem to be inconsistent, which does not encourage the production of a systematic theory. Eclecticism, if lacking in justification, can be a source of contradiction.
Although Widdowson (2004) does not oppose CDA"s cause, he casts doubts on its modes of analysis. He cannot agree with the way CDA uses SFG, where meaning is understood as a condition of texts, taken from them, not put into them. He points out that there is a gap between addresser meaning and addressee interpretation of this meaning, on the grounds that the perlocutionary effect is not a feature of texts but a function of discourse, in which the addressee"s assumptions are shaped by their knowledge and beliefs. Hallidayan grammar offers interesting devices for the description of semantic meaning (or signification); however, to Widdowson"s eye, this is defective because it fixes on isolated sentences instead of utterances. He adds that, in this framework, the concept of context is as essential as it is indeterminate. If meanings are understood as properties of the interaction between words and contexts, interpretation is an imprecise process. In a nutshell, Widdowson regards some CDA approaches as examples of the functional fallacy, by which he means the idea that pragmatic meaning (or significance) may be produced directly by signification. He maintains that abstracting sentences from their contexts and choosing examples relevant to the ongoing research does not lead CD analysts to produce analysis in the strict sense of the term: Pretexts influence how to approach texts and the type of discourses derived from them. To him, CDA is a biased, unprincipled, conventional, decontextualised cherrypicking of linguistic features, closer to impressionistic commentary, which supports interpretation and yields simplistic findings. Widdowson argues that CDA is critical in the sense that it has moral appeal, sociopolitical justification and liberal ideological positioning. And he accepts that the CD analyst observes issues that are relevant in areas other than the scholarly world and addresses how control is exercised though language. However, he strongly urges that CDA should adopt a critical attitude towards its own purposes, methods and practices, be explicit in methodological procedures, which must be replicable, and apply consistent principles and systematic linguistic theory. In all, CDA should comprise systematic analysis of entire texts, co-texts and contextual relations.
Although Chilton has contributed many papers on discourse that have a social-critical intention, his (2005) paper is critical of CDA, maintaining that what CDA lacks is a cognitive theory of language that could show how discourse affects social cognition and vice versa. Cognitive frame theory, conceptual metaphor theory and blending theory can explain better than traditional CDA approaches (including SFG) why phenomena such as racism and prejudiced thought can occur. In addition, Chilton hints that the work by critical analysts is based on no particular scientific programme and may simply reflect a universal ability in nonexpert cheating-detection, going so far as to suggest that, taking into account its audience and scope, CDA may be of limited social import. Billig (2003) thinks that CDA has the crucial characteristics of a critical approach: The claim to be critical of the current social order and of the approaches which do not critique the current social order"s domination patterns. He also recognizes the importance of CDA"s claim that non-critical approaches prevail in the academic world, resulting in keeping existing power relations unchallenged; and he supports CDA"s insistence that an interdisciplinary approach is needed. Nevertheless, he recognises that, through naturalisation and institutionalisation, a critical approach may itself become a dominant discourse and, consequently, a dominant discipline, with the shortcomings of the approaches it criticises.
Other critical voices include those of Martin and Blommaert. Martin"s approach (1992) critiques CDA because of its inability to put into practice its social-based ambitions, so that in the end it observes social phenomena we mostly dislike, producing very persuasive materials on why they are offensive, but failing to suggest practical action. Instead he proposes Positive Discourse Analysis (Martin 2000). He argues that, for the purpose of social change, scholars should analyse not only texts they find objectionable but also texts they find admirable and motivating. As for Blommaert (2005), he deals with the discursive production of inequality and the need for self-critique in CDA. Drawing on linguistic ethnography, he refers to a particular kind of parochialism. He considers that one of the shortcomings of this field is that generally most work produced to date pays attention to texts of relevance in the West since, as a rule, CDA is not applied to societies other than the First World. He further criticises CDA on the grounds that it has demarcated boundaries around itself as a field: Much discourse analysis can be critical, says Blommaert, without subscribing to the underlying assumptions and methods of CDA.

New directions in CDA
New formats and materials involved in communication have encouraged new avenues of CDA research, analysis of the multimodal properties of texts being one significant innovation. Other trends include developing the connection between CDA and cognitive linguistics, analysing gender semiosis, and bringing corpus linguistics into CDA.
In recent decades, attention has been drawn to just one communicative mode, verbal language. However, music and pictures are the basis for the meaning-making process in the audio and visual modes; the size, colour and frame of a news report are important to guide the addressee"s engagement with the text; the distribution of images and the timing of news are significant in TV and the press; body posture, gestures and the use of space help construct our text and talk (Hodge and Kress 1988;Kress and van Leeuwen 2006).
Having underlined CD analysts" tendency not to go beyond the verbal detail and the sentence level, so that it is hard to explain what happens in the mind when reception takes place, Chilton (2005) stresses the necessity to look at how we construct knowledge, that is, which procedures are necessary for individuals to share views and build mental models. This implies, for instance, his concern with the role of cognitive frames in facilitating human processing of information, and with the cognitive strategies deployed in order to infer other people"s intentions. In the same vein, O"Halloran (2003) had already concluded that there must be a link between sentential structures and mental representations, all of which seem to be controlled by discourse rules, an idea that needs to be reassessed in the light of connectionism (McClelland and Kawamoto 1986). In his model of reading for gist, O"Halloran discusses the extent to which lay readers attend to absences from a news text (especially as far as causal relations are concerned), and considers how to avoid mystification in the interpretation stage (Hart 2010).
Feminist Critical Discourse Analysis aims to analyse the relationship between gender and language, which mostly means examining enactments of power by men and women in the public domain (see papers in Lazar 2005). Sexism, victimisation, emancipation and the construction of identity are key issues in question. Feminist research has been prolific in investigating the role of gender in politics, the media, the workplace and the classroom context. The evolution from the deficit, difference and dominance approaches to a shift to discourse (see Litosseliti 2006) has been crucial. In current research, attention is not directed towards whether men and women speak differently, or whether the language of females is a deviation from the male norm (traditionally, the excuse to explain male superiority at some levels), but to comprehend gender as a dynamic construct (Cameron and Kulick 2003;Sunderland 2004).
Finally, it has been argued that the use of corpus linguistics techniques in CDA may help to avoid or reduce researcher bias. Quantitative computer-aided corpus approaches can address large data sets, and the focus of analysis can take into account collocations, keyness, semantic preference and semantic prosody. What appears to be obvious and is taken-for-granted are checked against the data at the same time that unexpected findings arise in the procedures of description and analysis (Mautner 2009). Patterns of preferred and dispreferred lexis and structures (Baker 2006) facilitate detecting of the ideologies of hegemonic discourses associated with particular texts. The limitations of corpus approaches to CDA are that they work with very little context, may hinder close reading and can help us learn only about the verbal domain. Corpus-informed CDA can give the impression that it is a mechanistic or positivist approach. O"Halloran and Coffin"s (2004) discussion of over-and under-interpretation counters this view, however.
8. An example of CDA Analysis in CDA can be top-down, where analysts begin with their understanding of the content; or bottom-up, where the starting point is the linguistic detail. In practice, however, some combination of both is in play. The analyst looks for what is encoded in sentences (i.e. signification) and its interaction with context (i.e. significance). In this respect, the analyst is merely doing what an ordinary reader would normally do, but with more conscious attention to processes of comprehension, their possible effects, and their relationship to a wider background knowledge than the ordinary reader may assume to be relevant. Depending on the approach, various linguistic devices are paid attention to with a view to understanding their contribution to some potential strategically intended meaning that may be linked with ideology. The clues found are interpreted, and some explanation is expected about them and their implications. For example, many CDA writers use Hallidayan linguistics, focusing on the ideational and interpersonal metafunctions of language, specific deployment of which may, it is claimed, reveal author-to-audience power relationship or be connected (by further interpretation) with the author"s positioning. Here I show CDA at work by studying the British Press discursive construction of the killing of Saddam Hussein"s sons.
The analysis of how people speak or write about crime is interesting because, by describing the verbal construction of perpetrators, we can decipher the discourse of wrongdoing, and gain an insight into the conception of the world held by the speaker or writer. Thus, the language of the journalists who presented Saddam Hussein"s sons as agents of evil may also tell us much about the journalists themselves and the newspapers they work for. To illustrate my point, I have studied all the news articles published in the UK on 15 December 2003, both on paper and on the Internet (when available), one day after Saddam Hussein was captured by US troops. The newspaper subcorpus consisted of 56 items, including all sorts of articles and other materials (see Table 1). The web newspaper subcorpus was a 17,492 word collection of 16 articles taken from the Daily Mail, 7 from The Guardian and 3 from The Independent. A full analysis might focus on metaphor, modality, transitivity and lexical selection, among other features. Given that I want to examine the construction of opinion, I look at transitivity, which in SFG is treated as part of the ideational metafunction (see Halliday 1985), and provides a powerful tool for the analysis of how meaning is embodied in the clause.
In particular, the category of transitivity deals with the linguistic means we have for expressing, through language, our experience of the world around us. Its components are participants, the processes these are involved in and the circumstances in which processes take place. There is some tendency for clause constituents to be expected to have a specific semantic role: Subjects are generally agents; direct objects, patients; indirect objects can be recipient or beneficiary; subject complements, attributes, and so on. As for the types of processes, Halliday proposes six: Material (action, event), behavioural, mental (perception, affect, cognition), verbal, relational (attribution, identification) and existential.
In spite of this looking like a precise taxonomy, it can still be difficult to identify sometimes which is which when there seems to be overlapping (see Table 2). In my corpus, the West (the so-called Coalition) is in the main involved in material (action) and verbal (saying) processes (see Table 3 below).
Since they are identified with many of the values and beliefs supported by the privileged voice behind these newspapers, they are sayers (or, simply, speakers) whose voice is heard profusely in several ways. They are also represented as agents, or doers, who carry out generally neutral or more positive actions than their "High Value Target". The others of the West, the Iraqis, are scared, dead or victimised sufferers; very few are said to be actors or heard articulating their thoughts. That is, they are those affected by the processes and as such are often encoded as grammatical objects. As for Saddam, who is described as perverse, is accordingly demonised, dehumanised and objectified-this representation of him arising in large part also as a result of the journalist"s selection of verbs expressing the processes in which he is a participant. So far, none of this is surprising. The position of the newspapers is reflected in their negative depiction of the dictator and the marginalisation of all voices other than the non-problematic West. The ideology of the media is equally transparent in the next set of examples, which concern the dictator"s sons. Nominalisation is a powerful means of abstraction that may, in certain contexts, disguise, or make less salient, agents, responsibility and circumstances. For example, the Guardian text speaks about "the death of Uday and Qusay" (the end of their lives without indicating the cause) and "the killing in July of Saddam"s two sons" (an unmentioned someone or something caused them to stop living). In its editorial, we also read: "Uday and Qusay … perish[ed] in a murderous blizzard of bullets". Here the instrumental cause of their death is made explicit (i.e. the large number of something as annoying and unpleasant as bullets), but there is no direct reference to any trace of the animate agent pulling the trigger. The Daily Mirror reports Saddam"s sons as being "killed in a battle", and adds that "Uday and Qusay … died in July guns blazing, after waging a four-hour-battle with American forces". Readers see that there has been a change of state in the patient argument, something natural and unavoidable. Although we get information about how this took place, the agent remains unknown. This is explained in terms of the circumstances of this event (expressed, for instance, in the clause "after waging …"). The human agency is thus not stated explicitly and the instrumental causation is expressed indirectly. In some way, such indirectness may appear to justify what happened: There was a long fight between opposing groups, one of which is vaguely present; it was in July; bullets were fired quickly and continuously; and both perished. In the last scenario, there is something else: Uday and Qusay were dead because they started this battle and continued it over a period of time. The American forces only fulfil the role of goal, the Hallidayan term for the participant affected by (not the agent of) some material process.
The Daily Mail tells that they were "killed after opening fire on an overwhelming US force which surrounded them in July", "in a fire-fight against US forces", and that they "died in a shoot-out with US troops". In the first case, the journalist shows the episode as if Saddam"s sons had to be affected necessarily by an agentless action, which eventually caused their death because it was them who started shooting at the Americans. In the second case, we only know that someone or something caused their death in a battle that involved the use of guns rather than bombs or any other weapon. US forces are merely a circumstance in the whole event. As for the third, although the supposed agents of their demise are mentioned indirectly as minor participants, the fact that the journalist prefers the lexical item shoot-out helps imagine a fight in which several people shot at each other. In other words, the idea of reciprocity is presented here, in contrast to the previous examples. Furthermore, the noun troops indicates that there were other people involved in this event, that the US force was not an abstraction, and that there was some human agency behind it.
As shown above, the point I have been considering has to do with avoidance of agency. To die is a material (event) process. It happens to animate beings, and it happens accidentally or not. If we use this lexeme, we focus on the affected participant, not on the cause (e.g. an illness, misadventure), the agent (e.g. living thing acting deliberately or not), the instrument (e.g. gun, knife, rope, poison) or the circumstances (e.g. in water, in action) concurring to provoke it. If we prefer the verb to kill, we refer to a material (action) process that involves an affected goal and an actor that causes the object"s death. As for to perish, it means that the patient dies as a result of an accident or very harsh conditions; these very hard conditions were the bullets shot by the US troops, against which they fought to the death, an idea that is celebrated by the newspaper if compared with Saddam Hussein"s apparent passivity.
The images resulting from the newspapers" construction of the scene are the following: We know what happened to both (they died), when (July), where and how (in a battle, in a fire-fight, guns were blazing), and why (they waged a battle with American forces, they opened fire on a US force). Sometimes the actor or cause of their deaths is referred to indirectly as another circumstantial element (waging a four-hour-battle with American forces, in a fire-fight with US forces) or as the affected object of their actions (opening fire on a US force). Thus, these journalists depict Uday and Qusay as responsible for their own deaths, since they started the attack. Linguistically speaking, the US force is a contingent participant or suffers Saddam"s sons" actions, but it never acts. This may have been an unconscious choice. Nevertheless, in one case the writer adds clarifying contextual information: The US force was overwhelming. The journalist emphasises that an amount or quantity is much greater than other amounts or quantities. The situation is perceived as one in which one side is at a disadvantage. Therefore, the affected goal changes as well as the notion of who is the patient.
This practical exercise develops some of the main tenets of CDA: Aim, type of data and analytical approach. Methodologically speaking, I have studied one aspect such as transitivity which is very relevant to understanding people"s positioning. This is one way to get to know which your identity is, and how you see the world and how you perceive others. To me, the application of this perspective is useful because it deals with ideologically loaded material systematically. It is made even more robust by the use of the complementary tool of corpus linguistics, which has allowed me to scrutinise frequency patterns and made it easier to discover the traces of what is (and is not) essential in the text, in order to comprehend better speakers" and writers" intentions. If the latter avoid some structures, this may be an indication of their fears and their taboo areas. If, on the contrary, they prefer some others, this may show they are worried about or even obsessed with those problems around which their discourse revolves. Despite the systematicity of this type of approach, I must agree, however, this is not "a mechanical procedure which automatically yields "objective" interpretation" (Fowler 1986: 68). Furthermore, Chilton (2004: 111) is right when he says that "labelling stretches of language as serving strategic functions is an interpretative act". All in all, I believe that certain devices lead any researcher to analogous conclusions, and that certain linguistic patterns have certain implications other scholars can also examine when replicating similar experiments. The microanalysis of a text helps to support this point; its macro-analysis can be used to avoid misjudgement. Everything is meaningful in language. The selection of one item implies at the same time the exclusion of some others (Fairclough 1995: 210). In Fowler"s words, "[d]ifferences in expression carry ideological distinctions (and thus differences in representation)" (1986: 4).
Categories such as ideology and power are present in the analysis of discourse practice (Fairclough 1995: 11). That is something I find revealing, because representation has to do with power. It is the powerful that have the chance to represent others (and themselves) in a light they may find more or less convenient. Metaphorically speaking, it is those who are powerful that give (or do not give) voice to those represented in their discourse practices. Their control of the media can allow them to arrange the ordering of events, and then to obscure or give more prominence to some participants instead of others.

Conclusion
CDA is an infant discipline gradually maturing. Curiously, several of its strengths can be taken simultaneously as the source of its weaknesses. Some of the exponents of the critical paradigm may themselves be lacking in a self-critical attitude since CDA has become an established discipline (Billig 2003). However, its general critical outlook has encouraged the development of new approaches, in an attempt to answer new research questions, and allay doubts about its method and theoretical grounds.
Its inter-and transdisciplinary nature still needs to be carried forward before it yields fruit. The ambition that CDA can help raise awareness about the unequal social conditions of minorities makes it a worthy enterprise. Nevertheless, both proponents and audiences are often familiar with this asymmetry and usually hold similar views: CDA is mostly consumed by CDA scholars not by the average woman or man in the street. Furthermore, despite CDA practitioners" activist orientation, their recent achievements only range from adjustments in the perception of a particular unjust state of affairs to cosmetic changes in advertising, news reports or political speeches.
Drawbacks notwithstanding, the adventure of CDA is to look into how discourses construct participants in communication as individuals with allegiances to the collective, and to embark on the analysis of the discursive means by which the world comes into existence. If this finally may bring increased understanding of social processes and structures, and ultimately perhaps, increased understanding of effects on social actors" views and actions, CDA must have a role in the social sciences.