Paul Baker and Tony McEnery's Corpora and Discourse Studies anthologises thirteen original research articles, each exploring ways in which traditional discourse analysis can benefit from the techniques of corpus linguistics.
The editors begin by presenting some of the key concepts and tools of corpus linguistics, opposing quantitative corpus analyses, using keywords and collocations, for instance, and more qualitative methods, with concordance tables and contextual analyses. The contributions are situated relative to how the ambiguous term of ‘discourse’ is conceptualised, from the closely textual, to the more broadly ideological. A historical and institutional perspective on the ‘development of a synergy’ between corpus linguistics and discourse analysis is followed by the presentation of three debates in the field. The first relates to criticism of ‘bias’ in corpus analysis, in the interpretation of concordance lines or the prioritising of certain keywords or collocations. The editors consider that if there is bias, it ought to be associated with transparent methodology and appropriately modest claims about the advantages of corpus linguistics. A second criticism would consider some findings of corpus linguistics to be so unremarkable as to be trivial. The authors reply that this is not always so, and that in any case, corpus research provides more objective grounds for otherwise unconfirmed intuitions. The third debate concerns the ethics of corpus linguistics and text mining, notably in cases where it is difficult, or methodologically undesirable to preserve speaker anonymity, with the authors advocating a case-by-case approach to the question. The introduction concludes with a brief overview of the thirteen chapters that make up the volume.
The first four chapters (chapters 2–5) bear on discourse related to ‘modes’ or text types. The series opens with Dawn Knight's ‘e-language: Communication in the digital age’. Here Knight compares features of the e-Language of the CANELC corpus with the spoken or written English of the BNC, focusing specifically on the use of core modals as markers of speaker attitude and interpersonal functions. The CANELC corpus includes a variety of types of e-Language, including tweets, blogs, emails, etc. Interestingly, in terms of relative frequency, core modals are more frequent in the CANELC corpus (1.53%) than in either the spoken (1.44%) or written (1.26%) parts of the BNC. The frequency of modals varies considerably, however, within the individual CANELC text types, peaking at 2.73% in emails, and falling to 1% in tweets. Knight suggests that the specific interpersonal qualities of the text types involved might help us understand these variations. While one might take issue with a rather unproblematical approach to modals and modality or with the relevance of the use of the BNC as a reference corpus, Knight's chapter nonetheless poses many interesting questions, opening new perspectives for generic studies of e-Language.
Chapter 3, ‘Beyond modal spoken corpora: A dynamic approach to tracking language in context’, by Svenj Adolphs, Dawn Knight and Ronald Carter, proposes a case study that tracks the use of language through time and space, as pairs of visitors travel between sites at the British Art Show. The aim is to explore ways of accounting for the changing contexts of language. After a slightly surprising introduction on ‘social’ and ‘ubiquitous’ computing, the authors give a full account of the methodology used. Special applications are deployed to record the speech of participants at the same time as their movement through time and space, either at the British Art venues or travelling between them. This is then transcribed following certain conventions and submitted to analysis with the Digital Replay System (DRS). If the ambitions and experimental protocol of the study are impressive, the results are somewhat disappointing. The data are taken from just three pairs of speakers and the sophisticated spatio-temporal tracking leads only to a blunt ‘inside/outside the museums’ distinction. The fact that the deictics this and that are used predominantly inside and I and we outside appears unremarkable in itself. In fairness, the authors recognise from the outset the limitations of their study. They make no claims for their results, but hope rather that it may constitute a starting point for further analysis.
‘Corpus-assisted multimodal discourse analysis of television and film narratives’, by Monika Bednarek, opens with a discussion – and defence – of the issues involved in corpus-based studies of film and television language. Bednarek considers that the narratives of film and of television display differences but also similarities, including their essentially multimodal nature. The integration of this multimodality into the analysis raises several issues, which she considers, before listing a number of tools and resources. There is in particular a useful section on various types of existing corpora in the field, including tagged corpora, subtitles, scripts, fan-produced transcripts, etc. Bednarek's case study considers part of the Nurse Jackie television series, using fan transcripts, generating keywords from the dialogue-only transcript, compared to the transcript of dialogue and non-dialogue features. Unsurprisingly, the personal pronouns you and I are salient in the dialogue, while indications of place (prepositions, specific lexical items) are salient in non-dialogue. The analysis also throws up some unusual data resulting from the fact that the transcriber is a non-native. The chapter concludes with a short analysis of a scene from the series, illustrated with stills. The initial introduction to the field and presentation of the issues is welcome, but the final analysis falls short of a fully multimodal approach, and the qualitative focus on one short scene does not justify a corpus methodology.
Chapter 5, by Karin Aijmer, is entitled ‘Analysis discourse markers in spoken corpora: Actually as a case study’. Aijmer begins by presenting discourse markers and some of the issues surrounding them, considering how best to define discourse markers, and describing their inherently deictic and multifunctional properties. She goes on to propose a case study of actually, with the aim of comparing ‘formal and functional properties’ in four different varieties of English, sampled from the ICE corpora. Interestingly, actually appears to be far more frequently employed in Hong Kong English than in British, New Zealand or Singapore English. Hong Kong and Singapore English display much higher frequencies for initial actually, however, compared to British or New Zealand English, where medial actually is most frequent. Aijmer distinguishes five functions of actually, which she illustrates from the corpus and compares across varieties, concluding that actually appears to have developed more ‘subjective’ meanings (pace Traugott), in Hong Kong and Singapore English, and more ‘intersubjective’ meanings, in British and New Zealand English. The role of collocations in determining functions of actually might have been highlighted, as might the possible role of linguistic substrates in the development of Hong Kong and Singapore usage, but the study provides a nonetheless interesting approach to cross-varietal discourse marker usage.
In chapters 6–10, the focus is on discourse related to social practices, beginning with Cinzia Bevitori's ‘Discursive constructions of the environment in American presidential speeches 1960–2013: A diachronic corpus-assisted study’. Bevitori starts by noting the importance of environmental issues in presidential discourse. Her corpus is limited to three forms of speech: the inaugural address, the State of the Union address and the acceptance speech, from Kennedy through to Obama. The study explores a number of environmentally loaded lemmata, environment, protect, preserve, conserve, energy and clean, which Bevitori charts in terms of their use by different presidents over time, considering their collocational affinities, the semantic evolution of some of these items over a relatively short period, and how this relates to extralinguistic events. She makes use in particular of statistical indications of relative frequency and keyness in her analysis, commenting cogently on a number of contextualised examples from the speeches. Her contribution shows how longitudinal corpus studies of this type can help chart societal trends. She also recognises some interesting limitations in the approach, in that a given phenomenon may receive different names, and that meanings may drift over time.
Chapter 7, ‘Health communication and corpus linguistics: Using corpus tools to analyse eating disorder discourse online’ by Daniel Hunt and Kevin Harvey, aims firstly to apply the corpus linguistic methods of keywords, collocations and concordance analyses to online health communication texts, in order, secondly, to home in on specific features of such discourse in the tradition of Critical Discourse Analysis. The data involved are drawn from two websites on which young people post their health queries. The authors are interested in how sufferers of anorexia represent this disorder linguistically and to this end they conduct a keyword analysis, using the spoken part of the BNC as a reference corpus. This allows them next to focus on the use of the terms anorexic and ED (‘eating disorder’) in each corpus. Their conclusions highlight the frequent ambivalence of anorexia sufferers’ discourse, in an approach which, the authors recognise, is different from the more typical CDA perspective of exposing and denouncing hegemonic discourses of power. Corpus techniques in this study are used to circumscribe the data, target keywords and target concordance lines for subsequent close analyses, which might be of interest to linguists and health professionals alike.
Jack A. Hardy, in chapter 8, proposes a ‘Multi-dimensional analysis of academic discourse’. The study begins with a brief overview of multi-dimensional (MD) research. Given that the texts of different ‘discourse communities’ may be described in terms of certain typical groupings of variables, these groupings can be detected, and quantified, through multi-factor analysis. Hardy's study focuses on the Michigan Corpus of Upper-level Student Papers, including A-graded student papers at final-year undergraduate, and first-, second- and third-year postgraduate level in a broad range of sixteen disciplines. These are analysed in terms of four dimensions, each characterised by a set of features. Dimension 3, for example, is entitled ‘Situation-dependent, non-procedural evaluation vs procedural’. The former might involve stance adverbs, hedges, verbs of communication, and the latter, nouns or passive constructions. Findings on each dimension are then presented in tabular form, by student level and discipline. This allows the author to characterise types of discourse styles by their scores across dimensions, and to observe the progression of student productions within a given discourse community as they adopt its codes of production. Hardy's conclusion places the study in perspective, evoking certain drawbacks of the very quantitative approach MD analysis implies, and suggesting that this mode of investigation might serve as a ‘stepping-stone’ to a more qualitative inquiry.
Chapter 9, entitled ‘Thinking about the news: Thought presentation in Early Modern English news writing’, is by Brian Walker and Dan McIntyre. In this chapter, the authors look at differences in modes of discourse presentation (DP), and particularly thought presentation, in the news of Early Modern English (EModE) and its Present-day English equivalent. They begin with a typology of DP, with a range of categories, from direct speech to narrative, in speech, writing and thought. Tags reflecting these categories were used to annotate a corpus of EModE news writing. The labour-intensive annotation task was performed manually, given the difficulty in pinpointing formal criteria for DP. The results show a far greater proportion of thought presentation in EModE news writing (nearly 30% of all DP), than in the PDE corpus (10%). In EModE this is often presented as internal narration. The authors offer a number of interesting hypotheses as to how the nature of news writing in the period – often based on epistolary sources – might have favoured these features. They conclude that small, manually tagged corpora, while not large enough to support reliable statistical generalisations, can enable us to formulate hypotheses which can then be tested on larger, untagged data.
Chapter 10, by Darryl Hocking, ‘The use of corpus analysis in a multi-perspectival study of creative practice’, appears to consider discourse as social praxis or habitus. Hocking works on a variety of sources relating to creative practices, specifically: corpora of student and professional briefs, audio recordings of brief meetings, tutorials, student–tutor interviews and published texts relating to art and the history of art. This heterogeneous set of data is chosen for the multiple perspectives it brings to Hocking's approach to the discourse of creativity. He uses corpus software to generate initial keywords from sections of the corpus, then further investigates the use of the key term idea(s) in the different discourse contexts considered. Hocking's study is ambitious in its aims, but raises the question of the compatibility of a corpus-based approach with such a multifarious set of data. For example, the use of the Wellington Corpus of Written English, as a reference corpus to generate keywords for a spoken corpus appears methodologically dubious, as does the comparison of collocations of the node-word ideas between corpora of very different natures.
Chapters 11–14, finally, move on to address questions of discourse related to ideology with Alan Partington, in chapter 11, focusing on ‘Corpus-assisted comparative case studies of representations of the Arab world’. Partington's introduction presents some of the concepts behind, and arguments against, the school of Critical Discourse Analysis (CDA), before moving on to the study of media representations of the ‘Arab world’. The data used are drawn primarily from the British broadsheets The Guardian and The Telegraph (2010 and 2013), as well as from The Daily News Egypt and Gulf News (2013). In the 2010 data, the Arab world was analysed as characteristically more passive – and hence less agentive – when compared to the Western world. In the 2013 data, there are new collocations, related obviously to geopolitical developments, but the tendency towards non-agentivity appears to be maintained. Comparison with 2013 data from English-language Arab world sources shows a far greater use of the expression the Arab world than in the British press – contrary to Partington's expectations. The function of ‘backgrounding agency’ in the Arab world is current in these sources too, however. Partington makes some careful methodological remarks on the sorts of conclusions his findings might warrant, in the specific context of press discourse, and on the relevance of corpus techniques in discourse analytical approaches more generally.
Chapter 12, by Paul Baker and Tony McEnery, is entitled ‘Who benefits when discourse gets democratised? Analysing a twitter corpus around the British Benefits Street debate’. The authors investigate discourse surrounding recipients of state benefits, using a corpus of tweets reacting to a Channel 4 documentary series, Benefits Street, on people living in a particularly underprivileged area of Birmingham. Tweets were targeted by hashtags and content, keywords were identified using AntConc, categorised manually by topic and analysed for collocations and clusters, in order to identify regular patterns. This leads the authors to isolate three main discursive positions or ‘discourses’, in the CDA sense, described as the ‘idle poor’ discourse, the opposite, ‘poor as victims’ discourse, and the ‘rich get richer’ discourse, representing three different ‘discourse communities’. Within these communities, the authors identify ‘two loose groups of “haters”’, and another more serious groups of ‘activists’, who regularly retweet, and whose use of language tends to be less abusive. The methodological use of corpus tools and enquiries to identify types of discourse within the noise of Twitter represents an innovative development in the field. Baker and McEnery provide a thoughtful discussion of the ethics and practicalities of using material which is, in data terms, both traceable and noisy.
In chapter 13, ‘Representations of gender and agency in the Harry Potter series’, Sally Hunt, makes a collocational analysis of volumes 1, 4 and 7 of the Harry Potter series, to focus on how female and male protagonists’ body parts are used. After briefly considering previous research in the field, Hunt presents her use of corpus tools to locate semantic and discourse prosodies which will then be considered in a CDA perspective. The top ten body parts mentioned are located, then labelled manually as male or female. The author then looks into verbal collocations, categorised threefold into changes of body position, communication with the body and acting on or responding to the world. The prosodies her study uncovers enable her to conclude that, in the Harry Potter series, agency is clearly gendered, as males are shown as physically active and agentive, females as physically insufficient and passive. This is, in Hunt's view, ideologically problematic in that children's books can be seen to play a formative role in conditioning the future adult's worldview. Corpus linguistics is in this respect well placed to reveal this type of almost subliminal bias.
The final chapter, ‘Filtering the flood: Semantic tagging as a method of identifying salient discourse topics in a large corpus of hurricane Katrina reportage’, is by Amanda Potts. As in the previous three studies, Potts works from a CDA perspective, using corpus techniques to identify covert ideological stance through the identification of collocations, and semantic prosodies. Specifically, a news article corpus of over 36 million words was automatically tagged for parts of speech and by semantic category, and then analysed using Wmatrix, which allows for hybrid semantico-syntactic enquiries of this type. The adjectival collocates of people belong most frequently to the semantic categories of ‘Numbers and Measurement’, ‘Social Actions, States and Processes’, ‘Money and Commerce in Industry’. Further analysis of secondary collocates for black people and white people, for example, shows categorisation in terms of wealth: poor black and rich white, etc. Potts concludes that much of the press reporting around Katrina involves the reinforcement of existing bias in readers, through recycled stereotypes, and the stigmatisation of charity, in a strategy that makes use of discourse to alienate hurricane victims.
In summary, Corpora and Discourse Studies: Integrating Discourse and Corpora provides a useful panorama of approaches to this growing interdisciplinary area. The present reviewer might find issue with certain biases, might question the relevance of certain methodological procedures or contest the usefulness of certain claims. The editors have, however, anticipated these quibbles in their cogent introduction, and the very fact that the contributions might foster debate in this fashion bears witness to the vibrant, thought-provoking, and on occasion provocative, nature of research in the field.
