Introduction

Mega sports events are a unique type of planned events in that they attract large audiences across the globe to experience ‘a joint “reality”’ (Billings & Hundley, 2010, p. 2). However, for most people, this experience is only possible through media. Through the choice of what and how to report, what is included and excluded, media largely control the dissemination and representations of such events. Discourse, understood here predominantly as language use, plays a crucial role in this process in that it can effectively shape public perceptions of (sport) events and the participating actors. For example, a heavy use of metaphors from the domain of war in reporting of international football matches constructs the events as armed conflicts between two nations fuelling nationalistic ambitions (Bishop & Jaworski, 2003). Celebrating a sportsman as a British champion when he wins, but as a Jamaican-born runner when the win is contested is another illustrative example demonstrating how media can frame and shift perspectives through specific language choices. Such choices are never neutral. Commentators and journalists have a large pool of language items at their disposal, but when writing about a specific event, they may prefer some words or phrases over others, thus reflecting certain attitudes and ideological stances and propagating a particular version of events (Baker, Gabrielatos, & McEnery, 2013).

It needs to be kept in mind that the media versions of events are not solely a reflection of what media producers see as interesting, relevant, and suitable. Media are driven by profit orientation and hence, what is produced also needs to appeal to the target audience. The audience is not just the immediate reader or listener, but often an imagined and more general one (Cotter, 2010). Thus, media will often draw on popular discourses assumed to be widely shared in order to respond to and attract more listeners and readers. Studying media discourses surrounding events can, therefore, offer considerable insights into how events are medially constructed and shed light on stances and ideologies that underpin such constructions and are widely circulated in society.

Commonly media discourses are examined using approaches such as discourse analysis (DA) and especially its recent faction of critical discourse analysis (CDA) (van Dijk, 1993; Fairclough, 1989; Wodak, 2001). Both approaches are qualitative in nature and are based on close readings of texts. By examining the interplay between context, text, and language, DA and CDA have been invaluable in providing insights into how media symbolically produce and reproduce norms and ideologies, especially of the powerful and dominant groups. However, both approaches operate normally with smaller sets of media data. Whilst this can produce rich accounts of a particular issue shedding light on nuanced or subtle discursive strategies and representations, findings obtained from an analysis of a handful of articles are difficult to generalise beyond the studied cases. They may also tell us little about the more pervasive and widespread ways of how media construct events. As Fairclough (1989, p. 54) observes, ‘the effects of media power are cumulative, working through the repetition of handling causality, particular ways of positioning the reader and so forth’. It is for this reason that analysts interested in media discourses are increasingly adopting the tools and methods of corpus linguistics (CL), which investigates linguistic patterns in large amounts of language data known as corpora. Corpora are normally interrogated using corpus-linguistic specialist software and quantitative, statistical techniques that allow for retrieval of frequencies, keywords, and collocations. In this way, analysts can scan large amounts of data and reveal frequent and repeated linguistic patterns that are not immediately visible to the naked eye. Quantitative results produced by corpus tools are studied subsequently in more depth using qualitative discourse-analytic techniques. The combination of quantitative corpus techniques with more qualitative methods have led to fruitful methodological synergies (Baker & Levon, 2015; Baker et al., 2008) and approaches such as the corpus-assisted discourse studies (CADS) (Partington, Duguid, & Taylor, 2013) that can reveal much more nuanced patterns of language use and representations in the media when compared with a quantitative or qualitative analysis alone. The main aim of this chapter is to demonstrate how the CADS approach can be effectively used to investigate media construction of events, in particular mega sports events. It will do so by taking as an example the media reporting surrounding the London Olympics 2012.

Sport, especially mega sports, events are an interesting site for social scientists and (critical) discourse analysts for many reasons. Sport is a major socialising force and a common cultural and symbolic resource deployed to aid solidarity, understanding, and identity formation (Meân & Halone, 2010). Sport is also a site of tensions mostly due to ideological contradictions between its underlying values and the actual practice. Although sport ethos is based on neutrality, inclusivity, and fairness with the aim to promote international harmony, more often than not it becomes a site of aggressive national rivalry and a strong reaffirmation of national pride. Despite some progress and changing social attitudes, it is still far away from being inclusive. Sport is recognised as one of the major masculine domains perpetuating the heterosexual norm (Daddario, 1994). Undoubtedly sport, due its global impact, has the potential to be a vehicle of social transformations; there are numerous examples demonstrating the impact of sport on social and ethnic barriers (for example Carrington, 1998; Woodward, 2004). At the same time, sport still continues to be an expression of privilege, domination and subordination maintained through the complex mechanisms of inclusion and exclusion (Jarvie, 1991; Long & Spracklen, 2011). It is this contradiction that makes sport ‘a disputed ideological terrain’ and a powerful site in the re/production of traditional hegemonic orders (Meân, 2010, p. 67). Thus, sport and especially its media representations are an important lens through which to examine social relations, and constructions of identities. Against this background, the paper focuses specifically on the investigation of discourses around identity that are constructed and supported by media coverage during major global sports events. The specific research questions that this study investigates are the following:

  • Which identities are foregrounded and positively valued and which are backgrounded, negatively valued, or absent?

  • Do mega sports events have an impact on the discursive ways in which national, racial or gendered identities are constructed, and, if so, what is the nature of this effect?

The remainder of this chapter is structured as follows. Section “Media Representations of Sport” offers a brief overview of research concerned with media discourses surrounding sports events. Section “The Methodology of Corpus-Assisted Discourse Studies (CADS): Principles and Tools” discusses in depth the CADS methodology, its principles and analytical tools. In Section “Discursive Constructions of Identities in Media Reporting During Global Sports Events: The Case of the London Olympics 2012”, the main results are presented and discussed in light of the above research questions. The paper concludes with critical observations regarding the benefits and limitations of the adopted CADS approach to study media constructions of mega sports events.

Media Representations of Sport

Sport as a cultural and symbolic resource has been of interest to historians and sociologists for some time. Three major themes have been addressed in particular: (1) sport’s role in promoting nationalism (Hogan, 2003; Kinkema & Harris, 1998; Smith & Porter, 2004); (2) sport, ethnicity, and racism (Jarvie, 1991; Long, 2000; Long & Spracklen, 2011); and (3) sport as a signifier of hyper-masculinity and a site of gendered ideologies (Bruce, Hovden, & Markula, 2010; Duncan, Messner, & Willms, 2005; Jones, Murrell, & Jackson, 1999; Kane & Lenskyj, 1998; Pirinen, 1997; Wensig & Bruce, 2003). Whilst it is beyond the scope of this chapter to discuss in detail this extensive body of research, some of the major conclusions need to be noted. Firstly, sport is seen as one of the major social forces that maintain a sense of national belonging. This role has become even more significant in the post-war world characterised by globalisation, mass migration, and weaker nation states. As Smith and Porter (2004, p. 2) observe, increasingly, national identities are defined through and are ‘inextricably linked to what happens on the field of play’. Secondly, sport continues to be a site of white privilege and domination (Spracklen, 2013). Although the participation of various ethnic groups in sport has considerably increased in the last decades leading some to celebrate sport as a tool of equal opportunity, research shows that racism is still experienced at all levels of sport (Long & Spracklen, 2011). Arguably, overt expressions of racism seem to be less common, not least because of the many anti-racism campaigns introduced by various sporting bodies. However, racism still prevails, though in more subtle forms, for example, as institutionalised racism that restricts access of certain groups to positions of power. Although teams in some sports appear to be more diverse, we rarely find sportspeople from ethnic groups in the role of coaches, managers, or members of sporting board rooms. Related to this is the issue of gender representation. Women’s participation in national and international tournaments has too increased significantly since 1950s, almost reaching parity with men in some domains, for example, in Olympic sports. Nevertheless, research shows that sport continues to be a male domain that maintains traditional and heterosexual gendered hierarchies (Bruce et al., 2010; Duncan et al., 2005; Eastman & Billings, 2000; Meân, 2010).

Methodologically, these issues have been mostly addressed by analysing historical records or using ethnographic and qualitative methods such as participant observations or interviews. Given the importance of media on experiencing sports events, some researchers have begun to look more closely at how media represent sport and sportspeople. There is now a large body of research documenting the media coverage of various sports events pointing to patterns of systematic mis- and underrepresentation. For example, Bruce et al. (2010) reveal that on average sportswomen receive less than 10 % of newspaper or TV coverage leading the authors to conclude that ‘women do not matter in our culture’ (Bruce et al., 2010, p. 5). Although the female converge in certain domains seems to increase, this does not necessarily lead to gender parity. As demonstrated by Duncan et al. (2005), despite the increased coverage of women during Olympic Games, the reporting tends to depict female athletes in disciplines traditionally seen as feminine such as swimming or gymnastics. Such disciplines focus heavily on aesthetics as opposed to strength and contact—attributes associated with men’s sports. Thus, some researchers are sceptical about the increased female coverage in media arguing that instead of achieving gender equality, it further perpetuates the ideology of biological difference which views women as suitable for certain sports only and denies them the possibility of demonstrating their prowess in others (Bruce et al., 2010; Duncan et al., 2005; Jones et al., 1999; Pirinen, 1997). This comes particularly into view in the ways in which female and male athletes are depicted. As Eastman and Billings (2000) show, sportswomen are more frequently described by references to their age, emotions, dating habits, or family, attributes that are rarely used to refer to sportsmen. Successful female athletes are also often compared to famous sportsmen. In contrast, successful male athletes are never compared to famous sportswomen, but to religious, mythical, or fictional figures of power such as Jesus or Superman. Eastman and Billings (2000), p. 208) conclude that such comparisons place sportsmen several levels above sportswomen and install the male athlete as the prototypical athlete. Early research by Daddario (1994) also shows that sportswomen are likely to be represented as sex objects or in heterosexual roles pointing to the impact of sports media in perpetuating the norm of compulsory heterosexuality (Kane & Lenskyj, 1998).

Similar mis- and underrepresentation has been shown in research examining the construction of race in sports media. Billings and Eastman’s (2002) study of American telecasts of the 2000 Sydney Olympics demonstrate that white American men athletes were the most mentioned and most positively portrayed. Their success was associated with commitment, whereas black athletes were portrayed as having innate sporting abilities. Similarly, in a longitudinal content-analytic study on the representations of black athletes, Goss, Tyler, and Billings (2010) demonstrate that certain features such as strength and hyper-aggressivity continue to be strongly associated with black players. Intelligence and leadership are, in contrast, more often attributed with white sportsmen (McDonald, 2010). Overall, sports media tend to over-signify blackness, whereas whiteness remains the unquestioned and ‘unmarked’ norm (Spracklen, 2013).

All in all, sports media have been shown to construct sport as a domain of hyper-masculine, white, and heterosexual identities marked by aggressive and competitive attitudes (Meân, 2010). Furthermore, hegemonic gender and racial representations seem to be strongly tied with national identities. Studies that examined the intersections of race, gender, and nationalism in the coverage of international sports events demonstrate the prevalence of nationalistic biases (Billings & Angelini, 2007) and the homogenising ideologies of nation states (Hogan, 2003). In fact, mega sports events are often directly utilised by governments to project the idea of a unified nation, whilst masking social inequalities and divides (Hogan, 2003). On the one hand, an increased coverage of women or members of ethnic groups serves this nationalising project in that it helps propagate the idea of a whole and ‘happy’ nation. On the other hand, media representations of sports events draw heavily on national discourses or in Hall’s (1992) sense, narratives of a nation, that is, images, stories, symbols, myths, metaphors, and stereotypic ‘national’ characteristics that all together construct an experience of shared history and strengthen the sense of national unity. These are also adopted to create divides between nations, which further fuels nationalistic rivalry, as Bishop and Jaworski (2003) show in an analysis of media reporting surrounding the football match between Germany and England during Euro 2000. In this way, sports media perpetuate the homogenising and nationalistic ideologies and are significant contributors to the process of ‘nation imagining’ (Anderson, 1983).

Research studies concerned with media representations of sport discussed above highlight the role of sports media in constructing and perpetuating traditional, hegemonic hierarchies and ideologies. Due to various awareness-raising campaigns, overt expressions of racism or sexism are rare. Nevertheless, biases still prevail albeit they tend to be expressed in more subtle or ambivalent ways, for example, through the amount of attention given to and attributions associated with certain groups. Whilst the conclusions from this research are compelling and consistent, there are some methodological issues. The dominant approach used is that of content analysis (CA). Whilst CA has a number of strengths, it is based on coding schemes and categories that are normally set ex ante. The data is subsequently scanned for the absence or presence of the set categories or themes. Whilst this may capture a number of relevant topics, some topics or categories may be omitted. Also, most of the studies are based on the analysis of a small amount of media data, making generalisations difficult. Finally, many of the studies highlight the importance of discourse and language but, with a few exceptions (Bishop & Jaworski, 2003; Meân, 2010), most of them are not systematic analyses of discourse or linguistic choices. The CADS approach, which is a linguistic data-driven methodology, can help address some of these weaknesses and contribute to a more systematic understanding of the role of discourse in constructing events in the media. The next section moves on to outline the principles and key analytical tools of the CADS approach.

The Methodology of Corpus-Assisted Discourse Studies (CADS): Principles and Tools

The term CADS was first introduced by Partington (2004) in order to account for the growing body of research interested in discourse and adopting the techniques of corpus linguistics. Generally speaking, CADS is an offspring of the parent methodology of corpus linguistics (CL) and was much inspired by the early work by Stubbs (1995); Hardt-Mautner (1995), and Krishnamurthy (1996). This section begins by describing the main concepts and tools used in corpus research. Subsequently, the specificity and tenets of CADS are discussed.

CL is primarily concerned with studying language on the basis of large electronic collections of real-life linguistic data normally known as corpora (McEnery & Hardie, 2012). Corpora can be quickly and reliably searched using specialist corpus-linguistic software programmes known as concordancers such as AntConc (Anthony, 2011), WordSmith Tools (Scott, 2008), and Sketch Engine (Kilgarriff, Rychlý, Smrz, & Tugwell, 2004). Although the starting point for analysis will be a statistical output of some sort, for example a frequency list, corpora can also be interrogated qualitatively by studying selected outputs, for example, concordance lines. Generally speaking, CL is interested in studying quantitatively and qualitatively lexical and grammatical patterns across different varieties (e.g, national varieties such as British English), domains (e.g. business communication), or specific genre (business emails). Insights derived from corpus research have increased our understanding of language use by providing empirical evidence for the existence of regularities and patterns that are not immediately visible to the naked eye. As the father of CL John Sinclair once pointedly remarked: ‘The language looks rather different when you look at a lot of it at once’ (Sinclair, 1991, p. 100).

The key analytical tools underpinning most of corpus work are frequency, keywords, collocation, and concordance. These will now be discussed in more detail. In corpus-linguistic terms, frequency refers to the count of items in a corpus, whereby item can be a word, a part of speech, or a keyword. Frequency lists are a useful entry point to a corpus in that they can reveal items that occur often (on top of the list) or rarely (towards the bottom of the list). Frequent words can point to certain regularities, which, in turn, can be a sign of importance and salience. Another good way of interrogating a whole corpus is via keywords. In corpus linguistics, a keyword is a word which occurs unusually frequent in a given corpus, as compared to another mostly larger reference corpus (Scott, 2010). The keyness of an item is usually established using a test of statistical significance, mostly log-likelihood. Whereas a frequency list indicates absolute or raw frequencies, a keyword list shows relative frequencies and it can be a useful tool in revealing the main themes of the studied corpus. Researchers often categorise retrieved keywords manually into semantic groups in order to see the wider range of topics covered in a given corpus.

Whilst frequency and keyword lists are useful in revealing frequent or unusually frequent items and can signpost major themes, they only show single lexical items and in isolation. We know from studies in phraseology that a single word does not necessarily encompass all the meanings that the word in question has. These often arise from the typical combinations of the word with other lexical items, that is, from collocations and from their use in context. Stubbs (2001, p. 105) gives an example of ‘cosy’, whose meaning generally tends to be positive, but in the collocation with ‘little relationship’, it expresses the negative meaning of ‘cliquey’. In corpus linguistics, collocation is understood as the co-occurrence of two or more words within a certain span, for example four items to the left and four to the right and a certain cut-off point (e.g. occurring five times or more). A distinction is normally made between co-occurrences that are determined on the basis of raw frequency or significance testing (Barnbrook, Mason, & Krishnamurthy, 2013; McEnery & Hardie, 2012). The latter helps establish the strength of associations, and commonly t-score, mutual information, or log-likelihood is used for that purpose. It needs to be borne in mind that each test favours different types of words and hence yields different results (Baker, 2006, pp. 100–104).

Collocations are very useful pointers of recurrent, typical lexical choices that are frequent and hence preferred or salient in a given data set. Such recurrent preferences are not just a matter of individual choices, but largely reflect established practices and evaluations and are often a means by which communities express, interpret, and evaluate people and actions. For example, Stubbs (2002) shows how the word ‘gossip’ is mostly used with terms denoting or implying a female speaker highlighting the stereotype that only women gossip. Frequent patterns of co-occurring choices can, therefore, indicate how a phenomenon, group of people, or events are persistently framed in discourse. As Stubbs (2001, p. 35) states, collocations are not simply lexical items, they ‘are also widely shared within a speech community’ and can act as ‘nodes around which ideological battles are fought’ (ibid., p. 188).

Finally, corpus researchers can study selected words or phrases in context by examining concordance lines. Concordances are lists with lines that display all occurrences of a search term. The word or phrase is normally positioned in the middle with a few words to the left and to the right. Such a display is known as a KWIC, short for key word in context. Figure 8.1 provides an example of a random concordance for the plural noun ‘women’ from the COR-OLYMP corpus, which includes press articles published in major national British national newspapers during the London Olympics 2012 (see below). A close examination of concordance lines is a qualitative technique that enables the researcher to discover lexico-grammatical patterns offering clues to the uses and meanings of the search term in context. To illustrate an example, we will have a look at the concordance lines in Fig. 8.1. As can be seen, the term ‘women’ often collocates with the modifier ‘British’, ‘Britain’s’, or ‘Great Britain’s’ highlighting potentially a national bias when referring to female athletes in the context of the Olympic Games. Another interesting pattern is the use of ‘women’ in the second position after ‘men’ providing some evidence for the male-firstness observed in general English usage (Baker, 2014). In this way, a concordance analysis can reveal associations that would otherwise be difficult to detect and could provide evidence for the existence of certain discourses and biases.

Fig. 8.1
figure 1

Concordance lines of ‘women’ in COR-OLYMP

Full size image

The CADS approach utilises the quantitative tools offered by CL, but it extends the methodological paradigm by integrating techniques commonly associated with qualitative discourse analysis in order to understand the discourse in question and its context as much as possible. Alongside frequency or keyword lists, CADS researchers interrogate the data in a variety of ways, for example, by close reading, watching, or listening to its subsets (Partington et al., 2013, p. 12). In most cases, the studied phenomenon is further contextualised by considering and examining its social, political, or historical context (Partington, 2014; Taylor, 2014). In contrast to general corpus research, CADS researchers prioritise a comparative approach. As Partington et al. (2013, p. 12) highlight: ‘we are not deontologically justified in making statements about the relevance of a phenomenon observed to occur in one discourse type unless, where it is possible, we compare how the phenomenon behaves elsewhere’. Comparisons can take different forms. We can compare and contrast one discourse type in different sources, for example, national print newspapers versus political speeches. We can also examine the same discourse produced under different circumstances, for example, during different events. Another type of comparison is a diachronic one, which involves studying a particular type of discourse at different points of time (Marchi, 2010; Taylor, 2010). This enables the researcher to detect changes and shifts over time. Finally, we can combine all the different types of comparison depending on the research questions.

The CADS approach has been successfully used to study a variety of discourses including discourse of science in the British press (Taylor, 2010), the discourse of morality also in the British press (Marchi, 2010), nationalism and language ideologies in the Canadian context (Vessey, 2013), the representations of Islam in the British media (Baker et al., 2013), and the portrait of immigrants in the British and Italian press (Taylor, 2014). Representations of specific events have not thus far been examined by using this methodology. Therefore, the next section presents a case study demonstrating how the CADS approach can be employed to investigate media constructions of events, especially the discourses of identity that underpin such constructions. The focus is on the London Olympics 2012.

Discursive Constructions of Identities in Media Reporting During Global Sports Events: The Case of the London Olympics 2012

This case study is based on the analysis of a large corpus of newspaper articles (COR-OLYMP) that were published in four major national British newspapers with their Sunday publications from 27 July to 12 August 2012 (see Table 8.1). The data was collected using the database Nexis. Instead of focusing solely on texts that were specifically about London Olympics, the decision was made to retrieve all articles published during this period. This allowed us to assess better the extent to which media construct sports events alongside other newsworthy events and which discourses are foregrounded and backgrounded in this context. Following the comparative CADS credo, we also decided to compare the media reporting diachronically and created two control corpora encompassing articles from the same publications with the same time span, but one year BEFORE and one year AFTER the event. In doing so, we were able to evaluate the impact of mega sports events on the discursive ways in which such events and specifically identities during such events are medially constructed. Table 8.1 shows the data in all three corpora. Because of the unequal sizes of the data sets, results were normalised whenever necessary. The corpora were interrogated using the software programme WordSmith Tools Version 5 (Scott, 2008).

Table 8.1 Corpus data
Full size table

As discussed in Section “The Methodology of Corpus-Assisted Discourse Studies (CADS): Principles and Tools”, frequency lists are a good point of entry into a corpus and thus, the analysis was started by examining and comparing the frequency lists retrieved from the three corpora. We included content words and personal pronouns only, as they are more likely to reveal aspects of identity constructions. Table 8.2 shows the most frequent content items in the three corpora.

Table 8.2 The 30 most frequent content words and pronouns in the three corpora
Full size table

Table 8.2 points to a number of lexical items that denote national and gender identities and these seemed worthy of further scrutiny. We begin by examining the item ‘British’, which belongs to the 30 most frequent content items in COR-OLYMP and occurs much lower on the frequency list in the two other corpora. This could suggest that national identity is more emphasised during global sports events such as Olympic Games. To further test this claim, we compared the use of other typical items denoting national and regional identities used in the UK to see whether ‘British’ was preferred during the London Olympics.

Figure 8.2 demonstrates a considerable increase in the use of ‘British’, ‘Britain’, and ‘Britain’s’ in COR-OLYMP. Together with ‘UK’ and ‘GB’, these were the preferred national terms. The frequent use of GB in the context of London Olympics is not surprising, as the acronym was used in the name of the British national team. Terms denoting regional identities such as ‘England’, ‘English’, ‘Scotland’, and ‘Scottish’ experienced a drop, in some cases substantial. Overall, the major national newspapers seem to promote the concept of nation-state or big-state nationalism more strongly than aspects of regional identity, as evidenced by the lesser attention given to Scotland, Wales, and Northern Ireland in all three data sets. The London Olympics appears to have reinforced this trend. Thus, the data provides some empirical evidence suggesting that media reporting during global sports events effectively strengthen the concept of national identity and nation-state.

Fig. 8.2
figure 2

Terms denoting national and regional identity in the three corpora

Full size image

Other interesting items appearing on the frequency lists are words denoting gender. For example, the two masculine pronouns ‘he’ and ‘his’ belong to the most frequent items in all three corpora. It is notable that these pronouns always appear before the feminine pronouns ‘she’ and ‘her’, suggesting that in the media men receive more attention than women. This is consistent with previous corpus research that revealed a significantly higher proportion of male than female terms in general language usage in English (Baker, 2014; Pearce, 2008; Sigley & Holmes, 2002) and in the British media (Caldas-Coulthard & Moon, 2010). However, when we compare the normalised frequencies of the female pronouns across the three subcorpora, an interesting result emerges. In the BEFORE Corpus, ‘she’ occurs 226.3 times per 100,000 words. In COR-OLYMP, the number rises to 322.2 and in the AFTER Corpus declines again to 245.3. The same pattern applies to the use of the pronoun ‘her’ as well as other terms denoting male and gender identities. Figure 8.3 shows the trends for ‘woman’, ‘women’, ‘man’, and ‘men’. Thus, although the traditional order was still maintained, the London Olympics seems to have had some impact on the gender representations in that media gave more attention to women as opposed to the time before and after the event.

Fig. 8.3
figure 3

Distribution of gendered lexical items

Full size image

This makes gender an interesting area for further investigation. For reasons of space, this cannot be perused here in greater depth. To illustrate the advantages of corpus tools and methods, this section focuses just on one example, that is, the collocations used with the term ‘women’. Table 8.3 demonstrates the strongest collocations as measured by t-score and retrieved within a −5 and +5 span. Only content words were included. As can be seen, items associated with ‘women’ vary across the corpora. In the BEFORE and AFTER corpus, women frequently co-occur with lexical words pointing to reproductive status (‘pregnant’, ‘children’), appearance (‘beautiful’, ‘hair’), and age (‘young’, ‘aged’, ‘older’). In the AFTER corpus, we also find references to crime (‘violence’, ‘rape’) and sex (‘sex’). Previous corpus research on gender representations in the media points to the same tendency: women seem to be more strongly associated with appearance, sexuality, and violence in contrast to men, who tend to be linked with careers, roles, and status (Caldas-Coulthard & Moon, 2010; Pearce, 2008).

Table 8.3 The 20 strongest collocations of ‘women’ across the three corpora
Full size table

Of interest, however, is the absence of the ‘regular’ associations with women in the media reporting during the London Olympics. The collocations that appear in COR-OLYMP are, in some sense, ‘unusually’ positive and point to different discourses about women. Firstly, we find here two modifiers ‘Britain’s’ and ‘British’ suggesting that in the context of global sports events women are strongly attributed with the nation-state. Interestingly, the association between Britain and women appears predominantly in the context of celebrating sporting success. A quick look at concordance lines in Fig. 8.4 illustrates this point: the attribution with ‘British’ and ‘Britain’ is clearly linked with winning, especially gold medals.

Fig. 8.4
figure 4

Concordance lines of ‘British women’ and ‘Britain’s women’ on COR-OLYMP

Full size image

In the concordance lines, we find a number of positive associations highlighting physical strength (‘beat’ in line 4), persistence (‘secured their first Olympic medal’ in line 7) and power (‘had the race under control’ in line 9)—attributes that are more commonly associated with men and rarely with women (Pearce, 2008). Frequent references to ‘British’ and ‘Britain’s’ when describing women provide yet another evidence for the robustness of the concept of nationalism in the context of global sports events. The example also shows that women are more likely to be positively represented, if they win the highest award, that is, a gold medal. The frequent collocations of ‘gold’ and ‘first’ (see Table 8.3) provide further support for this claim. To test this assumption, we will now look briefly at the collocation profiles of two female athletes Jessica Ennis, who won a gold medal in the women’s heptathlon, and Rebecca Adlington, who received two bronze medals in swimming. For reasons of space, we focus on collocations surrounding the surnames of the athletes only.

The first fact to note is that ‘Ennis’ (as the surname of Jessica Ennis) appears 745 times as opposed to ‘Adlington’, which occurs 384 times. This suggests that Jessica Ennis as a gold winner received almost double amount of attention in the media than the double bronze medallist Rebecca Adlington. Table 8.4 shows the 30 strongest associations with both terms.

Table 8.4 Collocations of ‘Ennis’ and ‘Adlington’ in COR-OLYMP
Full size table

A number of interesting patterns emerge from the lists. Jessica Ennis is mentioned together with other male gold medallists, whereas Rebecca Adlington is not. Ennis is also described in superlatives as ‘best’ or ‘greatest’, which reflects her status as a champion. We also find references to ‘Britain’s’ and ‘British’, suggesting that her sporting success was framed in the national terms—Ennis as incorporation of the nation’s success. Interestingly, the two modifiers do not appear in the vicinity of ‘Adlington’ implying that bronze medallists do not ‘enjoy’ the same status. The collocations occurring with ‘Adlington’ point to her sport discipline and we do not find here any instances of positive associations. Interesting is the grammatical collocation ‘herself’, which occurs five times in the vicinity of ‘Adlington’. As a reflexive pronoun, herself ‘reflects back’ on the subject and is the object of the verb to which the action of the verb is ‘done’. It can also be used as an intensifier. Studying the context in which ‘herself’ collocates with ‘Adlington’ points to a rather negative discourse expressing either an apologetic stance and desperation, as the examples of expanded concordances below show:

By now people were starting to panic, not least Adlington herself. Sensing that the situation demanded it, she tried to make a move. But she just could not summon up the change of pace she needed.

But she cruised past Adlington, who then found herself trying to hang on to bronze.

But look just a little harder and you will find that Adlington, as she said herself, has ‘swum faster than this all year’. ‘I would have liked the time to have been a bit quicker, I’m not going to lie’, Adlington said. ‘I’ve done that time all year and I don’t know what happened. Everything kind of caught up with me’.

Although the collocation profiles of the two athletes differ considerably, there is some similarity. Striking is the use of the term ‘girl’ as a collocation of ‘Ennis’ and ‘Miss’ in references to ‘Adlington’. Referring to an adult woman as a girl can be seen as patronising and trivialising. Previous corpus research has noted the tendency of the terms ‘woman’ and ‘girl’ being used synonymously, whereas this does not seem to apply to ‘man’ and ‘boy’ (Taylor, 2013). The representation above echoes this tendency and supports a claim suggested by the linguist Dwight Bolinger (1980, p. 100) that ‘a female never grows up’. Also, the use of ‘Miss’ in references to Rebecca Adlington shows that women are likely to be described by pointing explicitly to their marital status—something which has also been identified as a common feature in descriptions of women in the media.

The above analysis has revealed a number of notable discursive patterns and tendencies of media reporting during the London Olympics. The frequent use of items such as ‘British’ and ‘Britain’ and decline in regional terms exemplify how global sports events are used to boost national unity and the concept of nation-state. The persistent association with ‘British’ and ‘women’ provides further evidence for the dominance of national bias, which frames female success predominantly as a symbol of ‘national’ success. In terms of gendered discourses, we can note a pattern of ‘unusually’ positive representations of women achieving almost parity with me. However, as the analysis of collocations profiles of Jessica Ennis and Rebecca Adlington exemplary shows, only those women who achieve the highest award are likely to be positively represented. The analysis also demonstrates that despite the generally positive and less gendered representations, the media still tend to adhere to certain trivialising and patronising tendencies when referring to adult women. The use of ‘girl’ or ‘Miss’ illustrate this pattern.

Overall, the results seem to point to salient discursive patterns regarding identity constructions during a global sport event. However, the analysis focused on a single case and limited number of examples. Thus, the identified patterns need to be verified by comparing, for example, female with male representations or representations of sportspeople of various ethnic backgrounds. Also, to see whether such discourses are typical for global sports events, a comparison with other sports events would be necessary.

Conclusions

The above analysis serves as a case study to illustrate how the CADS approach can be effectively used to study media constructions of global sports events. By scrutinising frequency lists and studying collocation profiles of selected lexical items denoting national and gender identity, we were able to identify prominent discourses around identity that the British press foregrounded during the London Olympics 2012. The comparative CADS approach has also allowed us to assess the saliency of such discourses: comparing the media constructions of the London Olympics with media reporting during the same time but one year before and one year after has demonstrated the prominence of nationalistic bias and positive female representations, though the latter appeared to serve more nationalistic tendencies rather than gender equality (Wensig & Bruce, 2003). For reasons of space, the analysis focused on selected analytical tools and procedures. Issues involved in corpus creation, data sampling, and the use of statistical tests were not considered and readers are encouraged to consult, for example, McEnery and Hardie (2012) and Baker (2006) to find out more about these important aspects.

All in all, one of the major benefits of the CADS approach is its quantitative basis. Examining large amounts of textual data enables the researchers to produce empirical results that are more generalisable than those obtained from studying a few texts. It also fosters a greater distance to the data and increases objectivity of research in that it can help reduce some of the cognitive biases such as the primacy effect or confirmation bias (Baker, 2006, pp. 10–12). Moreover, corpus analysis enables us to see frequent and repeated patterns that may not be immediately visible to naked eye perusal or simply run counter to our intuition leaving room for serendipitous effects (Partington, 2014). Repeated associations are important because they help reveal discourses that are continually and systematically disseminated, thus, increasing our understanding of the recurrent or dominant ways of talking. It is precisely the ability of a corpus approach that it allows us ‘to see which choices are privileged, giving evidence for mainstream, popular or entrenched ways of thinking’ (Baker et al., 2013, p. 25). Also, by combining the quantitative corpus tools with qualitative discourse-analytical techniques, the CADS approach permits us to study our data comprehensively. By retrieving frequencies or keywords, we can identify general salient patterns in our corpus. These can be subsequently examined in more depth using qualitative techniques, for example, close readings of selected stretches of discourse via concordance lines. In this way, we may discover much more subtle and nuanced meanings that a list of keywords or collocations cannot reveal.

Despite its numerous benefits, a corpus methodology such as CADS has some limitations, especially when adopting the approach to study events and their mediatisation. Firstly, CADS is mostly concerned with texts and words, and leaves out the visual type of data. This could be a problem when examining media constructions of events, for which the visual component is central. A combination of CADS with a multimodal analysis would benefit future corpus-based research, though arguably this presents methodological and analytical challenges. Also, CADS is primarily interested in studying press data. Given the growing importance of social media in disseminating events of all kinds, future research needs to give more attention to the online mediatisation of events. Finally, corpus-based analyses of media often assume that frequent discourses or representations are influential and widely shared in society. Whilst this may be true, corpus-based results are rarely tested against views and attitudes of the wider public. The issue of reception and production of media contents is indeed rather neglected. Thus, another useful direction for future corpus-based research would be integration and triangulation of corpus methods with other qualitative and quantitative techniques used in social sciences, for example, ethnography or surveys. This would allow the researchers to validate some of the insights derived from corpus-based analyses, which, in turn, could offer invaluable insights into how discourses that underlie media representations of events really ‘work’ in a given discourse community.

Notes

  1. 1.

    This example refers to Linford Christie, whose win in the 1988 Olympic Games was contested on the grounds that he may have used performance enhancing drugs. He was eventually cleared of any wrongdoing.

  2. 2.

    There has been a great deal of debate regarding the status of Corpus Linguistics. Some researchers see CL as a new theoretical approach to language, whereas others argue that CL is more of a methodological package (see Partington et al., 2013). The author of this chapter agrees with the view expressed by Partington et al. (2013, p. 7) that the use of corpora and corpus tools is not in itself a new theoretical advance in Linguistics, but it certainly is a new methodological paradigm that has the potential to revise some of the existing theories or concepts and lead to new ones.

  3. 3.

    When comparing frequency results obtained from corpora of unequal sizes, it is important to normalise the raw frequencies to a common base, for example, per million words. Normalised frequencies (NF) are calculated using the following simple formula: NF = (number of examples of the word in the whole corpus ÷ size of corpus) × (base of normalisation). Here, the normalised frequencies (NF) were calculated using 100,000 as the base of normalisation. This base of normalisation was used throughout.