Additional discussion and perspectives

1. Evaluation and re-evaluation

In this additional section we aim to take stock of the issues, topics and materials developed in the book, and to raise a number of issues attendant on the analysis of language and power. Throughout the book, we have foregrounded the value and importance – indeed the ‘emancipatory’ significance – of analysing texts with a view to exploring how relations of power are mediated through them. Across the book we have attempted to locate instances of power in (and as) language in discourses as diverse as advertising and verbal humour, print media and social media. And we hope both that the models of analysis developed have been serviceable and that the insights they have revealed through textual analysis have been productive and stimulating.

However, we are anxious not to paint too cosy or simplistic a picture of the methods of analysis of CL and CDA, where the analyst subjects a text to analysis in such a way that the text’s complicit and hidden ideological agenda unravels before our very eyes. It would be wrong to give this impression, and so in this additional discussion we want to problematize a number of aspects of the approach adopted in this book and to highlight the concerns that some commentators have expressed about the methods of CL and CDA. We will suggest a number of additional methods that can help tighten and consolidate our theory and analysis. This section will also highlight significant developments in CDA that explicitly (or implicitly) address the concerns we raise here. Following the approach of the book, we will also present a series of exercises that call upon many interrelated models.

Are there problems with CDA?

The short answer to this question would have to be an unequivocal ‘yes’. While the investigation of language, ideology and power is an undeniably interesting topic of research, this is not to say that it is free from problems in its design, application and general theoretical foundation. CDA has come under criticism of various sorts, some of it to do with how the broad term ‘ideology’ is to be understood, some of it with how the epistemological status of the term ‘critical’ is to be defined. However, the bulk of the criticism has been to do with perceived problems in the methodology of CL and CDA. This has led over the years to a number of stinging and far-reaching attacks by respected authorities who have called into question the validity of the whole endeavour. These criticisms have tended to cluster around certain concerns which, in keeping with our approach in this discussion, we have crystallized from the literature into the following interrelated issues:

CDA is an exercise in interpretation, not analysis
CDA for the most part ignores real readers and listeners
CDA is not cognitive enough
CDA is too selective, partial and qualitative
CDA is too ambitious in its quest for social change

One of CDA’s most strident critics is Widdowson who, across different publications, takes issue with the central tenets of the discipline (Widdowson 1995, 1996, 1998). Widdowson’s broad position is that CDA is an exercise in interpretation, and not a valid method of analysis. He contends that ‘interpretation in support of belief takes precedence over analysis in support of theory’ (1995: 159). This criticism, echoed in critiques of CDA by others, is that the outcome of the analysis seems to be shaped by the political pre-disposition of the analysts themselves and not by any intrinsic feature of a text’s make-up. It is the CDA specialist alone, who through ‘expert exegesis’, unlocks the door to the text’s ideological meaning. Widdowson is especially trenchant in his criticism here, arguing that what is actually revealed by CDA is the discourse perspective of the interpreter and this has no more authority than other interpretation (1995: 169).

This leads us directly to the second major area of concern, bulleted above, that CDA largely ignores real audiences (and see our comments at the end of unit D2). If the main purpose of the analysis is to uncover and challenge the repressive discourse practices of powerful, interested groups, then what needs to be considered before anything else are the effects of these practices on ordinary (non-academic) people. Reactions of real communities to what the analysts deem ideologically insidious discourse are rarely considered; instead, the academic analyst comfortably assumes the perspective of those for whom the text was intended, moving seamlessly in and out of the multiple interpretative positions of specialist and non-specialist alike. We will address this important issue further below.

A criticism linked to the previous one relates to the role of cognition in CL and CDA. Cognition describes the mental processing that is involved in both the reading and understanding of texts and discourse. Adopting a specifically cognitive approach, O’Halloran (2003) addresses two of the key stages in CDA investigations, interpretation and explanation, as outlined in Fairclough’s framework (see unit B1). O’Halloran argues that CDA has focused mainly on ‘explanation-stage analysis’, in which it seeks to account for the connections between texts and wider socio-cultural practices at the expense of interpretation. As CDA, however, claims to interpret texts on behalf of readers who might be manipulated unwittingly, there needs to be an analysis of the relationship between readers and the texts being read and this necessarily involves more focus on cognition. According to O’Halloran, there has been ‘relatively little cognitive focus on how text can mystify for readers the events being described’. Similarly, Chilton points out that CDA, by and large, has not paid enough attention to the question of ‘how the human mind works when engaged in social and political action, which is largely, for humans, verbal action’ (2005: 30).

Our fourth criticism, that CDA is too selective, partial and qualitative, overlaps with some of the discussion above. The view here is that the analyst selects a text or type of discourse known in advance to be contentious, the confirmation for which is presented through an analysis that in essence only partially addresses certain pattern of language in the text. The linguistic analysis becomes no more than a supplement to what the analysis has decided a priori about the text. This criticism is by no means new: not long after the appearance of the first work in CL (Fowler et al. 1979), Sharrock and Anderson noted acerbically that the procedure for Fowler and his co-researchers seemed to be to look in the wrong place for something, then complain that they can’t find it, and suggest that it is being concealed from them (Sharrock and Anderson, 1981: 289; see also Simpson, 1993: 114–15). In a similar vein, Widdowson (1998) contends that CDA analysts tend to focus on certain aspects of texts while ignoring others – other aspects that might allow for alternative interpretations that might not fit their (leftist) worldview.

With respect to the distinction between qualitative and quantitative approaches, Stubbs (1997), who admits to being ‘basically sympathetic’ to CDA, challenges CDA’s methods arguing that whereas CDA might present valid arguments about text organization, the basis of its linguistic evidence is inadequate. Stubbs questions whether CDA actually adheres to its avowed aims, to what Fairclough and Wodak describe as ‘standards of careful, rigorous and systematic analysis’ (1997: 259). In other words, the analysts are seen to make generalizations about social representation and social change without the linguistic evidence to support it, and without providing systematic comparison between texts. As a corpus linguist, Stubbs believes CDA would therefore benefit from using quantitative and comparative methods (and see our observations in units D7 and D11).

While Stubbs suggests a quantitative analysis of texts to strengthen CDA’s methodology, Widdowson suggests a more ethnographic approach that would look at how the intended discourse recipients – and not only the discourse analysts – understand and interpret texts. It is precisely because CDA raises important social issues – that it has an agenda of ‘potentially very considerable social significance’ (Stubbs 1997: 114) and because it ‘seeks to reveal how language is used and abused in the exercise of social power and the suppression of human rights’ (Widdowson 1998: 136) – that enhancing its methodology should be a priority.

The final issue, no less significant than the other four, is just how effective is CDA? What has it accomplished in terms of social change and equality? Hammersley (1997), for example, argues that CDA appears too ambitious in aiming for ‘social change’, which results in a tendency to adopt a macro-social theory which reduces relations between people to a form of domination between oppressors and oppressed. According to Hammersley, this desire for social change can lead to ‘a rather naïve view of events and actions … without specifying the nature of the desired change’ (1997: 344). Importantly, researchers in CDA have tackled the issues we raise here, in recent publications progressively and increasingly so. For instance, van Dijk (2005) has taken an important step towards the integration of cognition in critical discourse analysis. Similarly, Chouliaraki and Fairclough (1999: 61–2), perhaps in response to Widdowson’s critique, acknowledge that CDA should be combined with ethnography which would require ‘the systematic presence of the researcher in the context of the practice under study’. Furthermore, ethnography, they continue, can yield ‘precisely the sort of knowledge that CDA often extrapolates from text’, such as ‘the beliefs, values and desires of its participants’. To this end, Wodak’s participant-observation study of an out-patient clinic in Vienna is an important CDA analysis that uses an ethnographic approach (Wodak 1996; and see further below). In turn, ethnography can benefit from CDA’s premise that ‘data descriptions should not be regarded as faithful descriptions of the external world’ but as particular perspectives of the social world that need to be analysed critically (Chouliaraki and Fairclough 1999: 62). This theme is echoed in the following comments by Wodak:

A fully ‘critical’ account of discourse would … require theorization and description of both the social processes and structures which give rise to the production of the text, and of the social structures and processes within which individuals or groups as social historical subjects, create meanings in their interactions with texts.

(Wodak and Meyer 2001: 2 –3)

Clearly, in the light of Wodak’s remarks, there seems to be an increasing awareness by CDA practitioners of the ‘discursive scope’ that a text can have.

In spite of the interconnected criticisms raised here, there is no doubt that in the broader context of language study, the practice of CL and CDA has brought to the fore important issues to do with ideology and power. Notably, it has helped to contribute to our understanding, among other things, of the discourses of racism and sexism. Another academic reflex of the CL/CDA method is that it has encouraged people, especially the young, to interrogate the discourses that surround them in their everyday lives. Although not setting an agenda in CDA per se, Carter (1990: 109) encourages the development in schools of tools for talking and writing discriminatingly about language, tools which encourages a critical awareness of the way language works in a context of culture. Rejecting traditional grammar drills and rote learning as arid and reactionary, Carter sets his store by materials that comprise real discourse, in real contexts of use. And while this may seem a common-sense position to adopt, it is one which still meets with resistance in political circles, especially from the political right, whose agenda is actively to discourage pupils and students from, to use Carter’s phrase, ‘seeing through’ language and discourse.

2. Language and power: extending the analysis

We will now develop this critique of CDA by suggesting some useful ways in which a CDA-style analysis can be extended, enriched or otherwise made more rigorous. We do this in a positive way, addressing head-on a number of the reservations raised about CDA in this discussion, but in such a way as to offer a programme of analysis onto which the core procedures of CDA can be mapped. Of necessity we restrict ourselves to three main areas: the corpus approach, the ethnographic approach and what we call the ‘mediated reaction’ approach. Each of the areas will be outlined in turn, although they comprise just three of a number of ways of enlarging our analytical apparatus to make our analyses stronger and more convincing.

Corpus studies in CDA

We have reported above on how a number of the criticisms levelled at CDA expressed concern over the lack of quantitative (and comparative) methods in what is a largely qualitative method of inquiry. As we pointed out, CDA researchers have traditionally been subjective and to some degree partial in the way they approach the texts (spoken, written, multi-modal) they wish to analyse critically. The benefit a quantitative analysis offers is to bolster and firm up interpretations made on linguistic evidence. This is not to propose an exercise in soul-less ‘number crunching’ for its own sake, but rather to develop a fuller and more expansive picture of patterns of language across the whole text on which the interpretation will be based.

Of the three issues surveyed here, the recent introduction of corpus techniques has been the most marked change in CDA’s methodology, with many studies balancing qualitative and quantitative dimensions of textual analysis through a marriage of CDA with the methods of Corpus Linguistics. Garzone and Santulli claim that because CDA practitioners are especially preoccupied with sociological and political issues, they ‘tend to focus their attention on larger discursive units of text’, often at the expense of ‘linguistic analysis proper’ (2004: 352). They continue thus:

Corpus Linguistics tools in a CDA framework have an important potential in helping overcome some of the questions that have afflicted the discipline and, in particular, the problem of the representativeness of the samples of language analyzed and the need to check the hypotheses developed in qualitative analysis against empirically verifiable data, chosen on the basis of explicit and objective criteria and collected using rigorous scientific and statistical procedures.

(Garzone and Santulli 2004: 353)

This fusion of Critical Discourse Analysis and Corpus Linguistics has been described positively as a ‘useful methodological synergy’ (Baker et al. 2008), a conceptual ‘co-penetration’ (Partington 2004: 11) and an ‘aid to research’ (Garzone and Santulli 2004: 351).

However, the chief methodological difference between Critical Discourse Analysis and Corpus Linguistics still resides in their respective approaches to text selection. Corpus linguists construct corpora from a repository of hundreds or thousands of representative spoken and written text fragments, unlike critical discourse analysts who base their analyses ‘on in-depth studies of small collections of [whole] texts’ (Mahlberg 2007: 196). While Corpus Linguistics inhabits similar theoretical ground to CDA in terms of it interest in examples of ‘real life’ language use (McEnery and Wilson 1996: 1), it does tend to place more emphasis on quantitative scope. This means that it seeks to substantiate claims about the production and reception of synchronic and diachronic discourses by tapping into millions of words of running text. Whereas early developments in Corpus Linguistics were concerned with the compilation of large ‘general corpora’ such as the Brown Corpus (Brown University Standard Corpus of Present-Day American English, 1964) and the Lancaster-Oslo/Bergen (LOB) Corpus of British English in 1978 (see Kennedy 1998: 19, 23, 27– 8), the integration of corpus techniques as an enhancement to practising CDA has since led to the construction of more specialized, purpose-built or ‘monogeneric’ corpora (Partington 2004: 13).

A growing body of research literature along the CDA/CL interface is evidence of this enhancement. Stubbs (1994), for example, examines a small corpus of school textbooks (totalling 110,000 words) and a reference corpus of one million words of written English, tracing the expression of causativity in ergative constructions, and the expression of modality in projecting ‘that’ clauses. He utilizes corpus methods as a basis for exploring the ideological stances to emerge from the school textbooks. De Beaugrande (2001) uses a combination of Corpus Linguistics and Critical Discourse Analysis to discuss the discourse of an academic paper written by H.G. Widdowson (subjecting the arch-critic to perhaps a taste of his own analytic medicine!). Teubert explores the keywords natural, human rights, work and property in a corpus of religious (Catholic) texts, and claims that Corpus Linguistics is ‘an adequate approach to observe the construction of social reality in a given discourse at a given time’ (2007: 89).

Aside from institutional and religious discourses, media discourse, particularly newspaper discourse, offers fertile ground for testing out a combined CDA/CL approach. Garzone and Santulli (2004) examine responses by the British print media to the events of 11 September 2001, drawing upon 102 articles (approx. 150,000 words) of text from the Daily Telegraph, the Guardian, the Independent and The Times. Orpin (2005) examines the ideology of the words sleaze and corruption in another corpus of 800 newspaper texts, focusing on lexical choices in the Guardian, the Independent, The Times and Today (no longer in print) derived from the Bank of English corpus between 1990 and 1996. Baker and McEnery (2005) also exploit a corpus of newspaper texts, as well as the United Nations High Commissioner for Refugees (UNHCR) website, to examine concordances and collocations of the terms refugee(s) and asylum seeker(s), finding discourses of such people being framed ‘as packages, invaders, pests or water’. This project is broadened in subsequent research by Baker et al. (2008) and Gabrielatos and Baker (2008), who analyse the discursive construction of refugees and asylum seekers in a corpus of 140 million words of UK press articles between 1996 and 2005 (the RASIM corpus of refugees, asylum seekers, immigrants and migrants). Mahlberg (2007) uses a corpus of Guardian newspaper articles from 2002 to identify local textual functions of the phrase sustainable development.

Corpus techniques (such as keyword searches, semantic and part-of-speech tagging, concordance and collocational analyses) using specialist corpus software such as Wordsmith Tools and WMatrix, do add quantitative weight to a traditional CDA analysis (see further the exercises below). There are limitations, however. As Gabrielatos and Baker point out, it is still the job of the analyst ‘to make sense of the linguistic patterns thrown up via the corpus-based processes’ (2008: 33). Corpus Linguistics might go some way in alleviating the criticisms pitted against CDA in helping it become more objective and verifiable, but qualitative and subjective assessments will still be apparent in any joint CDA/CL approach.

Below we guide readers through a basic corpus analysis of a news article, where we work through, in steps, an application of the WMatrix tools referred to above.

An ethnographic dimension in CDA

We have reviewed above some of the more strident criticisms of CL/CDA, of which Widdowson’s complaints were seen as especially trenchant (Widdowson 1995, 1996, 1998). As we noted, Widdowson (1995: 169) contends that what is actually revealed by CDA is no more than the ideology of the analyst, and this has neither legitimacy nor authority. In one place, Widdowson (1996: 60–9) targets a specific analysis by Fairclough which has now come to be known, with some notoriety, as the ‘baby book analysis’ (Fairclough 1992: 169ff). Fairclough analyses two medical texts for pregnant women, with the main focus on The Baby Book and some contrastive focus on the Pregnancy Book. It should be noted at the outset that this section of Fairclough’s otherwise stimulating book is marked by numerous technical and related inaccuracies: for example, Trew’s famous study of civil unrest in pre-independence Rhodesia is relocated to South Africa (p. 181), the concepts of ‘cohesion’ and ‘modality’ are confused (pp. 173–4), nominalization is ascribed to noun phrases like ‘medical disorders’ and ‘kidney disease’ which aren’t nominalized processes at all (p. 179), while in what is a generally weak application of Halliday’s functional model (see B2), certain verbs (like ‘ask’, ‘say’ and ‘tell’) are classified as material processes when they are in fact verbal processes (p. 178). The inconsistency in analysis is compounded by the broad-based generalizations that are drawn from it about the effects these patterns of language are alleged to have on the pregnant women who read the books. This is embodied in comments about the ‘decentering of women’ in The Baby Book or of that book’s language giving the sense that pregnant women are being ‘subject to anonymous and invariant procedures’ (p. 179).

It is this aspect of the study, the unwarranted ascription to pregnant women of certain kinds of reactions, with which Widdowson is most concerned. Quite rightly, Widdowson points out that neither the consumers nor even the producers of the texts are consulted. Instead, the CDA analyst selects the text as preconceived evidence of the problems assumed to exist in it, and the resulting, partial analysis only endorses this position. For instance, it is argued that a problem with The Baby Book is the sense it gives of the text producer writing from a position of ‘insider knowledge’ (Fairclough 1992: 173). Yet we would argue that this could be a positive feature of the book, where the insider knowledge of a medic, rather than, say, the speculative improvised commentary of a non-specialist amateur, might offer an authoritativeness that would be reassuring to pregnant women.

These serious concerns about the methods of CDA form part of an ethnographic study by McFarland, the full version of which can be located elsewhere in the web resources (see McFarland 2006). McFarland’s procedure is simple. First, she offers a more robust functional analysis of the same ‘baby’ texts which, when set alongside Fairclough’s original investigation, makes for a useful comparison of the potential form and direction an analysis of text can take. Then, heeding Widdowson’s call for direct consultation with those for whom the books were intended, McFarland designs an experimental procedure for gathering reactions to the texts from real mothers. This ethnographic stage of the experiment focuses on a ‘mums and toddlers’ group which comprised women from a range of social and occupational backgrounds. Using carefully designed questionnaires, as well as a range of alternatively worded versions of the relevant texts, McFarland is able to elicit quantitative data through the responses of the twenty-four members of the group. The conclusions drawn by McFarland’s study are important because they show how the women’s reactions to the texts were different in significant ways to the positions attributed to them by Fairclough. While the survey tended to confirm some of Fairclough’s observations about the different ‘voices’ in the texts, the general attitude of the women was much more positive, and contrary to Fairclough, they welcomed the texts’ attempts to try to ‘reassure’, ‘rationalize’ or ‘calm’ their intended readership.

In sum, adding an ethnographic dimension to our analysis can be a productive step, and as can be seen, it is often not a difficult step to take. Our intention here has not been to criticize Fairclough by picking out flaws in the narrower context of a single analysis – not even the best-known figures get it right all of the time – but rather to engage in the wider discussion the analysis prompts and then develop it to the next level. Fairclough is after all the one who has raised this agenda, and Widdowson the one who has challenged it. Our task as analysts is therefore to heed and perhaps absorb both positions before thinking of how we might extend and enrich our methods of analysis.

Analysing ‘mediated reactions’

Given that the relevant material is elaborated at length in units B12 and C12 in the main book, we need here only to touch upon the third of our extensions to the basic analytic procedures of CL and CDA. Overlapping in some respects with the ethnographic methods suggested above, this extension essentially involves going beyond the text by focusing on the reactions and responses of people, expressed through various media, to the materials under scrutiny. In other words, the researcher is able to position their own analysis and interpretation against the voices and opinions of other, non-academic people. This type of extension not only offers an additional interpretative dimension, it also avoids, pace our comments above, the pitfall of assuming too much about the thoughts of the people for whom the text was intended.

In unit B12, the framework of analysis used to explore mediated reaction was the model of ‘universal pragmatics’ proposed by the Frankfurt School philosopher Jürgen Habermas. The task of a such a model is to identify and reconstruct universal conditions of possible understanding; to account, in other words, for the general presuppositions of communicative action. Habermas points out that communication raises universal validity claims which we glossed in B12 as truth, sincerity and appropriateness.

The work by Martin Montgomery which is addressed in C12 of the book is an excellent illustration of how an analysis can be bolstered substantially by examining the reactions of people to a particular discourse event. Using the validity claim model, he looks particularly at media and vox pop reactions to speeches delivered on the death of Diana, Princess of Wales. Additionally, an earlier co-authored study (B12) applies the Habermas model to two British election broadcasts from 1992 (Edginton and Montgomery 1996: 114–20) and to the mediated responses which followed them in the popular press.

In this section we have highlighted various ways in which an analysis can be extended and enriched, and in such a way as to allay the worries we have addressed about the limitations of the CDA approach. Reasons of space have restricted us to a broad survey, and we do not wish to suggest that the work reviewed here is the only work to have addressed these issues. Far from it, there have been innovations and developments in many of the areas which are reflected in different Strands of this book. For instance, in the case of studying news media discourse this can mean interviewing editors and journalists or spending some time with news agencies to observe how they work. In this respect, Machin and Mayr’s critical discourse analysis of the Leicester Mercury (2007) is included in the extra website material for Strand B, this work underscores the importance of interviewing its editor about the paper’s policies on multiculturalism. More generally, we suggest that it is important to bear in mind that the study of text production and /or consumption can be usefully enriched by an ethnographic approach which investigates the processes that lie behind the production of (newspaper) texts (see Stubbs 1997).

Similarly, other work which has put the study of the reaction to and reception of discourse to the fore is Benwell’s study of readers’ responses to men’s magazines (2005) and its possible contribution to an understanding of the discourses in and around men’s magazines. This study consists of unstructured interviews with male men’s magazine readers and is therefore an important complement to Benwell’s critical multimodal analysis of men’s magazines (2002), which is our reading in D4.

3. Practising corpus-assisted critical discourse analysis

In this section we offer some practical suggestions for critically analysing newspaper discourse using techniques from Corpus Linguistics. We also include a shorter, second exercise which has more a general application and which can be approached using a number of different models of analysis. First, though, to our corpus-assisted material.

Corpus-assisted CDA

By accessing https://www.theguardian.com/uk/2006/jun/26/northernireland.race you can read an article entitled ‘Ulster justice system “institutionally racist”’ (705 words), which was published in the Guardian Unlimited on 26 June 2006. Examine the text of the article carefully, and think about the different ways in which the article may be approached using the existing toolkit from Critical Discourse Analysis. A step-by-step corpus approach to analysing the discursive properties of the article is suggested below. You may want to keep the article open on your web browser as you read through the analysis below.

The article deals with a number of issues that have already been explored in the book. As an example of newspaper discourse, it is encoded with the ideological views of the Guardian, and the intricate web of reporters, journalists and editors who work to produce the text. The article itself concerns institutional discourse, in particular the institutionally racist discourse that is deemed by the newspaper to be endemic in the Northern Irish criminal justice system. The article could therefore also be of interest to the forensic linguist, who may decide to research the linguistic foundations of this apparent institutional racism in more detail.

When we read through the article, there are recurrent linguistic patterns that seem to ‘jump out’ at the reader. The phrase institutionally racist is carried from the title into the main body of the text, where it occurs on several occasions: the phrase institutional racism crops up in the first, second and fourth paragraphs of the article. In unit C1, we saw that one of the functions of repetition and reiteration is to intensify meaning, with repetition having the strongest cohesive force. Critical discourse analysts consider this pragmatic strategy of overlexicalization as evidence of encoding ideology in news discourse. However, what if, as critical researchers of language, we wish to be more than qualitative in our assessment of the above newspaper article? The tools of Corpus Linguistics equip linguists with the means necessary to perform quantitative analyses of texts, from newspapers to novels, and from manifestos to music lyrics. Corpus tools enable critical discourse analysts to calculate, with a relative degree of accuracy, the prominence of recurring lexical items and linguistic patterns. It then becomes the job of the critical discourse analyst to decide how such patterns might serve to produce and reproduce existing discourses, such as institutional discourse or racist discourse as in the article above.

There are a number of freely accessible and subscription-based corpus tools available to the linguist, but in this section, we use Paul Rayson’s corpus tool WMatrix. Not only does WMatrix enable the researcher to view the frequency of particular lexical items, but it also ‘tags’ each word in the text with both a part-of-speech label (such as adverb, verb, preposition) and a semantic tag (such as Disease or Danger). For instance, the word party would be tagged with the part-of-speech label singular common noun, but depending on the co-occurring lexical items in the text, would be semantically tagged as either government or entertainment. At this point, we refer the reader to Paul Rayson’s introduction to the tool, accessible at: http://ucrel.lancs.ac.uk/wmatrix/ and his detailed online tutorial for using the tool, accessible at http://ucrel.lancs.ac.uk/wmatrix/tutorial/. This section presents only a basic step-by-step guide to performing a corpus analysis of the above article on institutional racism.

Step 1: Save the text of the article as a plain text file on your computer.

Step 2: Log onto WMatrix using your secure username and password.

Step 3: Click on ‘Tag Wizard’ and upload the saved file into WMatrix. Name the file ‘institutional racism article’.

Step 4: Be patient while the Tag Wizard annotates the file.

Step 5: You are now ready to analyse the annotated text of the article! A folder will have been created, entitled ‘institutional racism article’. Click on this to explore the work area that has been created by Wmatrix.

By performing this analysis, we were able to gauge information about the most frequent lexical items in the article. The table below lists the top ten lexical items in the text. Predictably, Northern Ireland occurs most frequently (11 times), while the words racism and racist rank fourth and seventh.

Figure Web.1a: ‘Racist’

Racist typically pre-modifies the words violence, attacks and incidents, promoting a discourse of racist plus some type of negative violent incident. Racist also occurs with xenophobic. If we look at the L1 rather than the R1 (i.e. the word occurring to the first place on the left of racist rather than the word to the first place on the right), we see that this place is typically filled by a preposition such as of or in. We can see a very discernible linguistic pattern here that is made much more explicit through the use of corpus methodology. Let us turn now to the word racism.

Racism: frequency – 6; relative frequency – 0.90

Figure Web.1b: ‘Racism’

Racism is typically pre-modified by institutional (or by the prepositions into, about and of ). Racism also collocates with attacks, and like the previous example of racist, occurs alongside a coordinating conjunction and xenophobic.

The words racist and racism were tagged semantically with belonging to a group. Other words belonging to this semantic category include institutional, racially, racial, public, ethnic, association. The concordance lines for each of these words are provided below.

Institutional: frequency – 3; relative frequency – 0.45

Figure Web.1c: ‘Institutional’

In this particular article, institutional is always found as a pre-modifier to racism.

Racially: frequency – 3; relative frequency – 0.45

Figure Web.1d: ‘Racially’

Racially always collocates with motivated in these examples, and two-thirds of these are followed by some form of negative physical behaviour.

Racial: frequency – 2; relative frequency – 0.30

Figure Web.1e: ‘Racial’

Racial occurs here with incidents and harassment.

Public: frequency – 1; relative frequency – 0.15

Figure Web.1f: ‘Public’

Ethnic: frequency – 1; relative frequency – 0.15

Figure Web.1g: ‘Ethnic’

Association: frequency – 1; relative frequency – 0.15

Figure Web.1h: ‘Association’

This is a micro-study and of course a broader range of newspapers needs to be consulted to fully investigate the discourse of institutional racism. However, by using corpus tools to analyse this article we have stumbled on some interesting results. Lemmas, or word forms, of race in this article, whether it be racism, racist, racial or racially, consistently collocate with the following abstract nouns and nominalizations: attacks, assault, incidents, harassment and violence. Those who suffer from racism are represented in the article as the recipients of physical brutality, an injustice that is perpetuated even further by the rise of institutional racism. They are being attacked, therefore, on all fronts: by those who are motivated to perpetrate violence against them, and by the criminal justice system that is designed to protect them. Therefore, Corpus-assisted Critical Discourse Analysis allows us to uncover linguistic features of a text that may have otherwise been overlooked.

Suggestions for further reading in Corpus Linguistics

Important books which survey the theory and methods of CL and CDA are Pennycook (2001) and the edited collections by Wodak and Meyer (2001), Wodak and Chilton (2005) and Weiss and Wodak (2003). Other publications which raise questions about the theory and practice of CL and CDA are articles by Toolan (1997) and McKenna (2004), while O’Halloran and Coffin (2004) uses corpus techniques to address problems of interpretation in critical linguistics.

Corpus techniques also feature in Deignan (2005), which is a study of the way metaphors are used in different contexts, while the part played by cognition in the understanding of discourse is the focus of Van Dijk’s study (1990). A generally useful collection of papers is Hoey et al. (2007) while Baker (2006) is an important illustration of how discourse analysis can be done using techniques that are grounded in Corpus Linguistics (and see also Conrad 2002). Louw (1993) offers an intriguing method for exploring irony in large corpora. More generally, he develops the term semantic prosody to describe the consistent aura of meaning with which a particular lexical item is imbued; an aura which develops through the other words with which the item regularly co-occurs. Finally, a useful survey of the principal methods of both CL and CDA is Wodak (2006).

Word	Frequency	Relative frequency
Northern_Ireland	11	1.65
Criminal_justice	8	1.20
System	7	1.05
Racist	7	1.05
Today	6	0.90
People	6	0.90
Racism	6	0.90
Attacks	5	0.75
Violence	5	0.75
Report	5	0.75