'Ideology' in the News: A Critique of Critical Discourse Analysis

(1)

‘Ideology’ in the News:

A Critique of Critical Discourse

Analysis

A thesis presented in part fulfilment of the requirements for the

degree of ‘MA Linguistics of European Languages: English’ from

the University of Amsterdam in the 2013/14 academic year.

Jack Martin Bamber

10618945

jackbamber00100@hotmail.co.uk

(2)

1 | P a g e

Abstract

This project looks at the field of Critical Discourse Analysis and some of the findings thereof. It identifies that Critical Discourse Analysis is a very varied and dynamic field which uses a variety of methodologies to look for ideological uses of language. Some of these methodologies are more objective than others. This project tests some of the findings of Critical Discourse Analysis to see if they hold up empirically, by looking for correlations between the occurrences of three linguistic features that are said to have ideological functions and their actual uses in a corpus of 10 online newspaper articles which come from politically and demographically different sources. The articles were taken from two British online news-sources, the Daily Mail and The Guardian newspapers, and were written about the January 1st removal of migration restrictions for Romanian and Bulgarian citizens coming to the UK. The project concludes by saying that the use of these linguistic features is not as clear-cut as some Critical Discourse theorists have said it is.

(4)

3 | P a g e

1. Introduction

As of the 1st_{of January, 2014, Bulgarian and Romanian citizens were given the same rights to live and}

work in the United Kingdom as any other EU citizen. Prior to this date the UK government had placed restrictions on their rights to do so. In particular, both potential employers and the employee

themselves had to apply for work permits to be able to legally work in the UK. Within the current climate of persistent, pan-European economic depression, many country’s governments are increasingly concerned about the freedom of movement that EU membership allows and the competition for resources it can create between locals and migrants, so these restrictions were not particularly unusual. Several other countries, i.e. Austria, Belgium, France, Germany, Luxembourg, Malta, Spain & The Netherlands, had also placed comparable restrictions on the rights of Romanian and Bulgarian citizens, which too ended on January 1st_{, 2014 (BBC, 2014).}

The UK has always been the recipient of large numbers of migrants, not least because of the country’s imperial history. There are many neighborhoods which have, for example, a substantial Indian, Pakistani or Bangladeshi heritage (see Bradford, West Yorkshire), or a West Indian and Caribbean one (Brixton, London), or an Asian heritage (which can be seen in one of several

‘Chinatowns’ like those in London or Manchester). These groups entered the UK in the post second world war period from many of the British Empire’s former colonies. This brought a great change to the cultural makeup of the United Kingdom that some did not find favourable. One Conservative MP, Enoch Powell, gave his infamous 1968 ‘Rivers of Blood’ (Powell, 2007) speech, which shocked and divided the country with his talk of the UK being “a nation busily engaged in heaping up its own funeral pyre” by admitting 50,000 “dependents” per year. His vision of the future of the country was grim to say the least. He was “filled with foreboding. Like the Roman, [he] seem[ed] to see “the River Tiber foaming with much blood””, such were his perceptions of the negative impact of migration on British society. Indeed, around that time there were many instances of rioting that were said to have a racial basis.

(5)

4 | P a g e

The communities that Powell was referring to have now existed for generations and as such have been a part of the cultural makeup of the UK for decades. For this reason race riots like those in the 60’s, 70’s & 80’s tend to happen less often now, but there are always new populations incoming towards the UK and the concerns that Powell held over the direction that UK society was heading in continue to be held by some people. More recently, there have been an increasing number of immigrants from Eastern Europe. These migrants do not travel to the UK due any old, colonial ties, but due to EU membership which allows free travel within member states. This started with the inclusion of Poland in 2004 and continues today in 2014 with Bulgaria and Romania. Critics of this situation include Nigel Farage of the United Kingdom Independence Party (UKIP) and Nicholas ‘Nick’ Griffin of the British National Party (BNP), two fringe but well-known UK political parties. Griffin has been put on record as saying (among other, far more blatant things) that “the country is full” and that “the British people are going to be a minority in our own homeland in 60 years” (Daily Mail, 2009) (For a closer analysis of BNP literature see Richardson & Wodak, 2009b:258-265). Farage, a self-styled ‘Eurosceptic’, has said that migration has left the UK “unrecognisable” and “alien” (Hope, 2014).

In the UK, the deadline for the removal of immigration restrictions became a catalyst for the discussion of all kinds of attitudes towards immigration and other prevalent social policy issues of the time. Just as was the case for the Polish arrivals before them, and the many other nationalities before them. The Powells, Farages and Griffins of the world tend to be in the minority and their opinions are seen as controversial, at least in the mainstream UK political discourse. After all, UKIP and the BNP are both fairly fringe parties and Powell was sacked from his position in the

Conservative party for his comments (Although UKIP are rapidly gaining ground, coming first in the European parliament elections in the UK, it remains to be seen how well they will perform in a general election). However, in many ways the story remains the same as it ever was; people are concerned about migration, preserving the ‘British culture’, overpopulation & (un)employment. There is now the added factor of Britain’s continued EU membership, but mostly the themes are

(6)

5 | P a g e

quite similar. Despite the fringe nature of the politicians, who go so far as to publicly voice their concerns to the fullest extent, there are sections of the population who may very well agree with some of what these politicians say but not all of it. True racists do exist of course, but most people hold more moderate opinions and/or don’t want to be branded with the social stigma of publicly identifying with racists, xenophobes and Nazi sympathizers. The average Briton may reasonably be expected to be concerned with some of the points mentioned previously, but there is a large gulf between that and becoming a full member of the English Defense League, an anti-islam

organization. That is to say that the concerns that the politicians take and distil into their speeches does come from somewhere and someone, but that ‘someone’ is likely to be somewhat more subtle in expressing these concerns. These attitudes towards migration, both pro and con, were well represented in the newspapers of the time leading up to the January 1st_{restriction removal and,}

because of this, these articles are rich in ideological and political content.

Within linguistics, Critical Discourse Analysis is an interdisciplinary sociolinguistic approach to studying the construction and use of ideological language and communication. Over the years, it has identified many linguistic features that it claims serve an ideological purpose. The newspaper articles about this topic from around this period should therefore be the perfect source of material to see how these linguistic features are used and to test the findings of Critical Discourse Analysis. This is what this project will endeavor to do. It will do this by performing a surface-level, semiotic analysis of a selection of articles from two UK newspapers with differing demographics and political leanings in order to identify the overall positive or negative ‘tone’ of the article. It will then look for

correlations between the tone of the article, the demographic makeup of the new-source, and the occurrence of certain linguistic constructions which have been identified as being potentially

ideologically charged by researchers operating under a Critical Discourse Analysis approach. This will show two things: 1) whether the language used by the two newspapers is different in terms of what features they use, and the amount that these features occur. And 2) the level of objectivity that the findings of Critical Discourse Analysis have. That is, whether the linguistic features that are identified

(7)

6 | P a g e

as having an ideological function actually do in any objective sense, either in absolute terms or in correlations between a text’s tone and content. To begin, though, it will be necessary to define what Critical Discourse Analysis is.

2. Critical Discourse Analysis

Critical Discourse Analysis (henceforth CDA) is a movement whose early introduction and

development is often attributed to Norman Fairclough (Blommaert & Bulcaen, 2000:454) and the publication of his 1989 book, Language and Power (Fairclough, 1989). It came out of the Critical Linguistics subfield which was being developed by the University of Lancaster faculty in the 1970’s. The name alone offers some suggestions as to what the approach entails, but the terms ‘Critical’ and ‘Discourse’ are often given somewhat technical definitions by the various researchers who currently operate in this field. Fairclough continues to be a prominent name, but other prolific researchers such as Teun Van Dijk and Ruth Wodak may define the terms differently and their theories have evolved over time, which of course can subtly effect the specifics what a critical discourse analysis involves. CDA is said to not be a single, monolithic framework, but rather a heterogeneous collection of methodologies (see section 2.3) that are related in the approach they take. In the words of Van Dijk:

“CDA does not characterize a school, a field or a subdiscipline of discourse analysis, but rather an explicitly critical approach, position or stance of studying text and talk.”

Van Dijk (1995:17, Original Emphasis)

In general, the underlying theme of CDA is that it is a sociolinguistic-orientated approach to language analysis which sees language as occurring in a social context, by participants who are goal-orientated (that is to say that their language is used for doing things) and who operate within existing social structures. In the words of Fairclough himself, “CDA oscillates (…) between a focus on structures and a focus on the strategies of the social agents, i.e, the ways in which they try to achieve out-comes or

(8)

7 | P a g e

objectives within existing structures and practices” (Fairclough, 2010:223). Therefore we can say that CDA is the study of communicative strategies within a particular field or mode of practice, which is to say within a particular ‘discourse’. In the following introductory sections I shall outline what previous researchers have meant by ‘discourse’. I shall also say what is meant by ‘critical’ as this is where most CDA researchers differ the most. Finally, I shall then explain the specific steps that are usually a part of a CDA and finally the findings that CDA has produced over the years. These three points are important for understanding the theoretical background of CDA and so the motivations behind the construction of the methodology that will be used in this paper to answer the two research questions posed in section 1. When describing the terms and methodologies that are used by CDA theorists I shall try to cover the main approaches but, as has been mentioned previously, CDA a fairly heterogeneous field.

2.1. What is Discourse?

According to Richardson (2007) the concept of discourse is very broadly defined and used in many different ways, but there are two main types. The first is a structuralist definition. This definition sees discourse as being another level of language structure which is ‘above’ sentence level. So, continuing the pattern from morpheme to lexeme to local syntactic structures on a sentence level, discourse is therefore how multiple sentences are structured. It is a kind of text-level syntax. For a structuralist the study of discourse could be seeing how a narrative is produced and develops throughout a text, or how cohesion and reference operates between sentences. Richardson gives the following example from conversation analyst Harvey Sacks (1972):

The baby cried.

The mommy picked it up.

Here, we can see how ‘it’ references the ‘baby’ from the previous sentence, and so links them narratively. The order of the sentence suggests a causal relationship between the crying and the

(9)

8 | P a g e

picking up, which is to say that the mommy picked up the baby because it was crying. If these sentences were presented to us the other way round we might reasonably think that the baby cried because of being picked up. This is therefore an example of a very basic, two-line discourse (if we take a structuralist definition).

A second definition of discourse is the functionalist one. This definition is influenced by the concept of language in use; that language is used for doing things. Functionalist discourse says that

“’meaning’ and ‘doing’ are linked to the concept of its usage” (Richardson, 2007:24), meaning that the way that language is used, its context, is of vital importance for understanding the semantics (‘meaning’) or speech acts (‘doing’) that are contained in a given instance of language. Using a functionalist definition of discourse, an important way that people understand the previous ‘baby’ example is by applying knowledge from beyond the boundaries of the text. By this I mean that readers have a wealth of social knowledge that they bring with them to any communicative event, and their interlocutors are aware of this fact. So the author of the text would know (or at least strongly expect) that the reader would have similar concepts of common socio-cultural practices such as mother-child relationships, due to the fact that the author and reader live within the same (or broadly similar) cultures.

These concepts of common-ground and shared knowledge have been talked about at great length in the work of Herbert Clark. Although he is not a CDA researcher specifically (he is instead more of a general psychologist/psycholinguist) his ideas about the nature of language use from his 1996 book Using Language are very useful for understanding just how context-bound language use can be. In this book, Clark (1996:92-121) shows how group membership can be indicated by numerous factors such as physical appearance, manner of dress, occupation and essentially any possible factor by which people can organize themselves into distinct groups. This group membership brings with it specific knowledge and also communicative and behavioural expectations. This could be a computer science student having a highly developed lexicon of technical programming terms, or an inner-city

(10)

9 | P a g e

youth having a richer non-standard slang vocabulary than an elderly middle-class person. Of course, these group identifiers are not only linguistic in nature, so our expectations of how to act as a receptionist in a 5 star hotel versus acting as a bartender in a seedy dive bar encompass all manner of behaviors. The point is that people have a huge amount of social knowledge that they bring to every communicative exchange which effects the linguistic strategies they use.

CDA, insofar as it views language as occurring in a social context as illustrated by extracts from Clark such as the one previously given, therefore usually takes a functionalist perspective on discourse. Richardson (2007), writing about analyzing news sources, says that “in written and broadcast journalism, meaning is constantly tied to context” (Richardson, 2007:24) as they are full of reference to previous events, other current events, and so make use of assumed shared knowledge between author and audience. However, definitions of discourse that are used in CDA often include more than just the effect of social context on the interpretation of texts. They also include the effect of social context on the production of texts. In many ways, language use involves choosing from a number of options. This can be in terms of what words to use, what sentence structures to use, whether to use a formal or informal register and so on. In other ways, people’s stylistic choices can be restrained by context. Certain types of language have, either explicitly or implicitly, codified rules that one has to follow. When writing a letter, for example, one follows a particular pattern which dictates using special constructions such as beginning with “Dear [name],” in English or “Beste [name],” in Dutch, and signing off in English with “Yours Sincerely” or “Yours Faithfully” and in Dutch with “Met vriendelijk groet” or similar. Likewise, a primetime television interviewer must follow a particular format in order to elicit responses from the interviewee and make the interview a success. At the same time, the topics of conversation must adhere to cultural expectations about what is acceptable to talk about on mainstream national television. Such effects of discourse on language could be listed ad infinitum.

(11)

10 | P a g e

Discourse can therefore be characterized as “text in context” (Wodak & Weiss, 2005) or, more explicitly, as “processes of text production, distribution and consumption” (Fairclough, 1995:2). It is “the social activity of meaning-making through (written/spoken) language, bodily expressions and/or sounds” (Forchtner, 2011:59 Original parenthesis). The process of analysis involves looking at a selection of “discourse fragment[s]” (texts) within a particular “discourse strand” (topic) (Jäger, 2001). A discourse fragment might be a single news broadcast within the discourse strand of ‘news reporting about the Greek economic collapse’, which in turn is part of larger discourses like ‘news reporting’.

2.2. What does it mean to be ‘Critical’?

Having defined ‘discourse’ and so defined CDA’s object of analysis, the other key term that requires definition is ‘critical’. The way that this term is defined effects what form the analysis might take and the conclusions that the researcher might draw from the data. The notion of being ‘critical’ is also what separates Critical Discourse Analysis from other branches of regular Discourse Analysis, so it is obviously important to understand what CDA researchers mean by it. Fairclough (2010:235) views CDA as “a form of critical social science geared to a better understanding of the nature and sources of social wrongs”. In another paper, he states that CDA “is not just descriptive, it is also normative. It addresses social wrongs in their discursive aspects and possible ways of righting or mitigating them” (Chiapello & Fairclough, 2010).

However, as with many other aspects of the heterogeneous CDA approach this need to actually provide solutions to problems is not always a part of a CDA analysis. In Fairclough’s own CDA of the marketization of the UK university system (Fairclough, 2010:91-125) he calls his approach ‘critical’ because it determines causal relationships in “(a) discursive practices and events, and (b) wider social and cultural structures” in order to “investigate how such practices, events and texts arise out of and are ideologically shaped by relations of power and struggles over power”. No need to propose viable solutions to social inequalities is mentioned in this paper. Indeed, the introduction to the first

(12)

11 | P a g e

edition of the same book (Fairclough, 1995) does not mention ‘righting wrongs’ as part of CDA’s modus operandi. Instead it is described as follows:

“the aim [of CDA] is to map three separate forms of analysis onto one another: analysis of (spoken or written) language texts, analysis of discourse practice (process of text production, distribution and consumption) and analysis of discursive events as instances of sociological practice.”

Fairclough (1995:2)

So it would seem that the need to propose working solutions to social problems is a relatively new addition to the CDA canon.

The other aspect of a CDA which makes it said to make it a ‘critical’ discipline is the interpretive nature of the analysis. Although CDA analysts may or may not make viable proposals for how to correct societal ills, CDA certainly does take a more “overt moral and political position with regard to the social problem analysed” (Richardson, 2007:2) than other branches of regular DA such as

Conversation Analysis. It takes this position because CDA also has a strong interpretive, qualitative element that can be absent from some methods of content analysis which are more (or even purely) quantitative. By doing this, CDA aims to “situate what is written or said in the context in which it occurs, rather than just summarizing patterns or regularities in texts; and argue that textual meaning is constructed through interaction between producer, text and consumer” (Richardson, 2007:15). CDA is therefore interpretive because language in general is interpretive.

For an illustration of this we can return to Clark (1996). Language, as a method of communication, is always a joint activity because communication necessarily requires, by definition, at least two participants. With only one participant there would be nothing being communicated and so no communication, no speech acts, are taking place. This extends to discourse, which for Clark is not particularly different compared to any other form of joint action, it is rather “simply a joint activity in which conversational language plays a prominent role” (Clark, 1996:50). From this, he continues on

(13)

12 | P a g e

to say how more formally linguistic features (that is, what is said and how it is said) are found alongside less conventional communicative features, but that both of these are vital to the

understanding of a discourse. He gives the example of a simple exchange at the checkout of a store, where Stone is the assistant and Clark is the customer:

Stone: I’ll be right there. Clark: Okay.

Clark: These two things over here. Stone: Twelve seventy-seven. Clark: Twelve seventy-seven.

Part of a drugstore transaction dialogue (Clark, 1996:51)

“To know what “I’ll be right there” meant, we need to know that I had just caught her eye and was waiting to be served. To know what Stone’s “Twelve seventy-seven” meant, we need to know that

she had just rung up my two items on the cash register” (Clark, 1996:51).

So we can see that a quantitative, content analytical approach is not enough to understand how language is used. The results need interpretation based on speaker motivations and position within the discourse and wider society. This is in-keeping with the idea that a CDA researcher is trying to find features in the text that are not immediately obvious in the text itself:

“The analyst will, or should, then draw out features in the text not normally obvious to the casual reader. The ideology, buried, or somewhat concealed, in the text will become clear.”

Machin & Mayr (2012:207)

Rather than looking at how language does things in a general sense, like in the way illustrated by the example from Clark, CDA is specifically “critical of how unequal language can do ideological work”,

(14)

13 | P a g e

where “ideologies are representation of aspects of the world which contribute to establishing and maintaining relations of power, dominance and exploitation” (O’Halloran, 2011:445). This is the way that ‘critical’ seems to be understood by most CDA researchers.

2.3. The Methodologies of Critical Discourse Analysis

I have already explained that CDA takes an interpretive perspective towards data, and that this is what makes it a ‘critical’ discipline in the eyes of some, although there are others who insist that it is the creation of plans to solve social problems which is the ’critical’ aspect. However this is just a part of what a CDA involves. In this section, I will describe some common methodologies that CDA theorists have used.

Reisigl (2013) has defined CDA as taking one of two methodological approaches, which he calls the Oldenberg approach (Reisigl, 2013:10) and the Duisberg approach (Reisigl, 2013:9). The Duisberg approach takes into account a group of texts and looks at the position of those texts within the institutional framework of an organization and broader discourse structures and how this manifests linguistically. The Oldenberg approach is a highly detailed micro-analysis of a smaller number of texts, or even a single text.

An example of a Duisberg-type approach can be seen in KhosraviNik (2010:11-13) as part of his CDA of representations of immigrants in UK newspapers. He establishes a difference in use between the words ‘emigrant/emigrate’ and ‘immigrant/immigrate’, near-synonyms in terms of use (a migrant can be said to be ‘immigrating’ into a host country or ‘emigrating’ form their native country interchangeably) which nevertheless have differing uses which he says reflects the attitudes of society that are being reproduced by the authors. ‘Immigrant/immigrate’ are being used negatively and ‘emigrant/emigrate’ are being used with more positive connotations. Suitably, ‘migrant/migrate’ falls between these two poles. To establish this as a trend it is obviously necessary to do some quantification, which is indeed what he does. Instances of the target words were counted, and then

(15)

14 | P a g e

the contexts in which the words were used were judged to be positive, negative, or neither. He looks at a large number of texts for just a few features in order to prove the use of these lexical items empirically.

A more Oldenberg-type approach can be seen in Fairclough’s (1993) study of the ‘marketization’ of universities that was mentioned previously. Here, Fairclough looks at only 4 texts that he has found which relate to university activities. By only looking at a small data set he is able to go into a lot more detail about the uses of the language and study it on more levels. Compared to KhrosavNik, who looked at language choice on a lexical level, Fairclough is able to get into the specifics of how certain grammatical constructions are used due to not being burdened by sifting through a large data set. Likewise, Capdevilla & Callaghan (2008) take a single speech from Michael Howard (ex-leader of the UK Conservative Party) as their data, and looked at how he used tropes and stereotypes in his argumentation as rhetorical devices.

Finally, a Discourse-Historical Approach has recently been proposed that aims to do all of the above and more by tying the text(s) not only to wider discourse practices, but looking at how these discourse practices have changed and how they have been renegotiated diachronically. An example of this kind of methodology can be seen in the work of Wodak & Richardson, specifically in their studies of how right-wing discourses on worker’s rights, migration, and ‘nativism’ have changed throughout the years in Austria and the UK (Richardson & Wodak, 2009a & 2009b).

As we can see, “CDA presents a diverse picture” methodologically speaking (Blommaert & Bulcaen, 2000:450). There are a variety of approaches that a researcher can take and not all of them are as rigorously empirical as they perhaps could be. As Fairclough said himself regarding his own previously discussed ‘University Marketization’ study, his findings “can hardly [be] said to be properly representative” (Fairclough, 1993:15) due to only using 4 texts as examples of his

theorizing. For this reason, some researchers frame their research as an attempt to dissect different perspectives on a particular issue, whereas others are more quantitatively-minded and so look for

(16)

15 | P a g e

objective patterns. The plurality of methodologies has led to an equally diverse range of findings. It also raises some questions about the validity of some of the findings of CDA if not all researchers are ‘on the same page’ when it comes to their methodological approaches. I will now discuss the former before moving on to the latter.

2.4. The Findings of Critical Discourse Analysis

Over the years, the CDA literature has produced a large amount of linguistic features that are said by its researchers to be ideologically charged. This list is already of a substantial size, but it is also growing all the time as researchers identify new features and branch out from traditional linguistic features and into the realms of “other semiotic dimensions” (Van Dijk, 1995:18). In fact there are so many features, and ideology in language is seemingly being found to be so prevalent, that “A

completed list of linguistic devices relevant for CDA cannot be given” (Wodak & Meyr, 2008:28). Part of the reason for this large and currently unknown number of features is that these linguistic

devices, the object of CDA’s attention, are described as “forms of interaction which are in principle subject to speaker control” (Wodak & Meyr, 2001:26). If each use of language is a choice, then we can talk not only about what features are present (and why) but also about what features aren’t present (and why). A CDA is not only interested in the form and content of language, but of the context and use of language. A CDA researcher not only comments about what language an author has used, but what they haven’t used and what they could have reasonably used instead, leading to a potentially endless set of CDA-relevant language-devices.

This has not stopped researchers from attempting to give their own taxonomies. The way that the features are organized varies greatly from researcher to researcher, but they generally share common features. For instance, Van Dijk (2007:125-126) presents his collection of “Some expressions of ideology in discourse” as being arrangeable under three headings: “Meaning”, “Form”, and “Action”. The features all serve to in some way assume or create a distinction between ‘Us’ and ‘Them’ groups in racist/nationalist/xenophobic discourse. Under the “Meaning” heading are

(17)

16 | P a g e

headings like “Topics”, “Local meanings and coherence” and “Lexicon”. Under these sub-headings are further lists of features. So the author’s use of modality, for instance, is an ideological device which is a part of the “local meaning” of the text and contributes to the overall “meaning”. Use of passive sentences is an ideological device which is part of syntax, and so a part of the “form” of the text, etc.

Jäger (2001:54-56) also attempts to create a detailed framework of features which he presents as an example of how to perform a typical Critical Discourse Analysis. His framework uses more traditional linguistic categories than Van Dijk’s but the content is largely similar. Instead of ‘form’, ‘meaning’ and ‘action’, Jäger simply breaks his analysis into a “structure” and “fine” analysis. The ‘structure’ analysis includes descriptive data about the text’s demographics (its readership and circulation, etc.), the overall theme of the text and where it fits into the larger discourse. The ‘fine’ analysis is a list of linguistic features arranged by categories such as “Rhetorical means” and “Text ‘surface’” features, and the “Ideological statements” that the text is making either explicitly or implicitly.

By combining the results of these two frameworks with the contents of other books about CDA such as Michin & Mayr (2012) and Richardson (2007) we can create the following list of features. This list is not exhaustive but it is likely that it will never be possible to produce such a comprehensive list as the research is ever growing and branching into new domains. However, it is illuminating to see the sheer scale of linguo-communicative features that we are talking about.

Linguistic Features That Have Been Found To Have Ideological Uses According To Critical Discourse Analysis:

Semiotics:

Image content (what is the image of, and how does this represent the topic being discussed?) (E.g. Feng & O’Halloran, 2012) and

(18)

17 | P a g e

Saliency (What is visually more noticeable and therefore what is being highlighted as most important) e.g.:

Color: brightness, hue, saturation

Text size, font, boldness and italicization. (‘Graphic Layout’, Jäger, 2001)

Iconography/Iconicity (What do visual elements like logos represent and what meanings do they bring to the text?)

Lexicon:

Structural Oppositions (opposing terms for ‘us’ vs. ‘them’) (Van Dijk, 2007)

Lexical Connotations (‘terrorist’ or ‘freedom fighter’) (Richardson, 2007. KhrosaviNik, 2010) Over/Underlexicalisation & Lexical Absence (are more or less words than necessary used? e.g. women doctors) (Richardson, 2007)

Quotatives/reference (i.e. how are individuals or groups of people quoted or referred to in the text?) (Machin & Mayr, 2012)

Modality (Should, would, could) (Machin & Mayr, 2012)

Rhetoric:

Argument structure (e.g. are the claims made in the most salient features of the text different to those made in the main body?)

Use of rhetorical devices e.g.: Repetition

Metaphor, Simile and other Comparisons (X is (like) Y) (Jäger, 2001) Idioms/sayings/stereotypical arguments (Jäger, 2001)

Syntactic/Information Structure:

‘Given’ and ‘new’ information (What knowledge does the author presuppose?) (Machin & Mayr, 2012)

Agency/passivity (‘Police shot the protestors’ vs ‘The protestors were shot’) (Schroder, 2002)

(19)

18 | P a g e

(Non-)Cleft sentences (Van Dijk, 2007)

Intertextuality:

How does the text position itself in relation to other texts on the same subject?

Communicative Acts (Van Dijk, 2007):

Use of interactional strategies that presuppose an Us and Them (e.g. assuming agreement between author and reader)

Use of interactional strategies that create an Us and Them (e.g. creating agreement between author and reader)

Disclaimers (We’re not racist (but…))

2.5. The Objectivity of Critical Discourse Analysis

I have so far explained that CDA is an approach which takes a more-or-less qualitative stance. I then continued on to show the list of features that CDA theorists claim can serve an ideological purpose is in fact vast and ever-growing. With these two points in mind, we might ask the question ‘is this what authors are really doing when they use these features?’ The list of features that can be used

ideologically includes a lot of things, and some of these features are actually quite common in general language. CDA is an interpretive discipline, but so is poetry and literature analysis. Just as there is not a ‘correct’ interpretation (in literature studies they might say ‘reading’) of a poem, we might say there is no ‘correct’ reading of ideology in language. However, as with literature studies, there are some readings/interpretations that are more persuasive than others and these are the ones that tend to have the most convincing evidence. I would say that KhosraviNik’s (2010) quantifying of lexical items combined with the consistent use of these items in the texts is more convincing than a detailed analysis of a single text because it has a greater amount of empirical proof. It is not just one use of one feature that we should be interested in when talking about how Language works (although seeing what is communicatively possible is nevertheless interesting), but the systematic use or the systematic interpretation of such features. In other words, idiosyncratic language is interesting but tells us little of wider communicative practice without testing it in other

(20)

19 | P a g e

situations. KhosraviNik’s data collection also contributed to a more convincing argument because it was very inclusive. By this I mean that it included all articles that fit within certain selection criteria. It did not selectively pick articles that the researcher presumed to have particularly ideological language and work from there. I consider the failure of some CDA researchers to be inclusive of all data to be a weak-point for their argumentation. Consider this:

“It would rather be like you paining your garden shed red as someone gave you some paint they had left over and you had no money to buy the green paint you wanted, and in fact hated the red. Later

a multimodal discourse analyst turns up and interprets your ideology on the basis of the hue and saturation of the red you have used.”

Machin & Mayr (2012:212)

This humorous example has a serious point; people do not always act ideologically, and certainly they do not always do so consciously. This is something which has been overlooked by some CDA researchers and has led to some vague and unconvincing claims akin to the use of color hue and saturation in the quote. The current state of affairs is that not all of the linguistic devices that were presented in section 2.4 have the same amount of evidence in their favor and the methodologies used to justify their inclusion may not be the same. This, combined with CDA’s unashamedly political approach, means that we can ask some serious questions about the validity of CDA’s findings.

Several of the big names in CDA have come to its defense in the wake of criticisms like these. The politicized nature of CDA might be used by some as a criticism but Van Dijk (1995:19, Original parenthesis) counters this by saying that CDA “is often seen as “political” (biased) and hence as “unscientific” (subjective) by scholars who think that their “objective” critical work does not imply a stance and hence sociopolitical position, viz., a conservative one that serves to sustain the status quo”. By this he says that any stance taken by an academic researcher is inherently political, even if their stance is to not take a stance. This means that although some academic research might claim to

(21)

20 | P a g e

be less political (and so more objective), this is not true. We can therefore separate the ‘politics’ of CDA from its ‘objectivity’, according to Van Dijk.

Likewise, Wodak & Meyr (2008:32) state that “Severe ‘objectivity’ cannot be reached by means of discourse analysis” because every act of qualitative analysis potentially embeds the researcher’s own beliefs & ideologies, and so prejudices the analysis “towards the analysts’ preconceptions”. In this way, the objectivity problems of CDA are extended to all research that is mainly qualitative. Certainly ‘true’ objectivity (what is presumably meant by “severe objectivity”) is a standard that not every type of research can achieve and the use of such a term opens the proverbial ‘can of worms’ that is the philosophical debate about what it even means to be truly objective. Philosophers have argued about this for centuries. However, even qualitative methodologies have measures in place to stop research from becoming a free-for-all of baseless claims. CDA researchers tend to agree that greater objectivity (if not “severe objectivity”) is something that should be aimed for.

Although they point out that CDA is a “program [...] in its infancy” (Van Dijk, 1995:24), CDA researchers agree that they should indeed try to aim for clarity and not superficial, ad-hoc descriptions of language use where the researcher projects their own reading as much as they interpret what is actually there. To do this requires the establishment of a more rigorous and standardized methodological toolbox in order to have comparable findings across all the areas that CDA researchers operate in. The “lofty aims” of CDA “can be effectively realized only when CDA is a thorough theoretical and descriptive enterprise that eschews fuzziness, impressionism and

superficiality” (Van Dijk, 1995:24).

So overall the objectivity of CDA is in a very unclear place at the present time; “How CDA validates and grounds its own critical standards is […] not easy to answer” (Forchtner, 2011:60). On the one hand, researchers dismiss some alleged problems of CDA as being problems that are not particular to CDA, or by saying that they are not even problems at all. On the other hand, CDA is constantly striving for some amount of standardization in the face of a very diverse range of research. There is a

(22)

21 | P a g e

continuum stretching from those who are content to make modest claims that a certain construction is being used in a certain way in the observed context only, to the likes of Fowler (1991) and claims like this:

“In Lanugage and Control, we claimed that nominalization was, inherently, potentially mystifactory; that it permitted habits of concealment”

Fowler (1991:80, Original emphasis)

Look at the way the argument is presented and we can see why the objectivity of such claims is unclear. Nominalization (the act of using nouns or noun phrases in place of verbs, replacing processes with entities) is said here to have inherent properties. Namely, inherent “mystificatory” properties that conceal facts or ideas. Yet it is only an inherent ability to be potentially mystifactory, to permit habits of concealment. The way it is worded here implies the presence of an inherent quality in nominalizations and nominalization use that is also totally optional. This claim is both categorical in its use of language while simultaneously being incredibly vague, which leaves us wondering if it is really an inherent property at all and what is the true nature of this property. Speaking about the same phenomenon with an equal amount of (un)certainty, Schroder (2002:105) has this to say:

“Syntactic transformations, particularly those labeled ‘passivization’ and ‘nominalization,’ can be considered ideologically problematic because they may obscure agency”

Schroder (2002:105, Original emphasis)

Again, nominalization can be considered problematic because of what it may do. It follows that there are therefore uses which are problematic and those which are not, but if we are only talking about possibilities of use then it is not the nominalization itself which is “ideologically problematic”, but the person who is using it. I will return to Fowler and Schroder and to the uses of nominalization later in this paper.

(23)

22 | P a g e

The one and only way to check the objectivity of the findings is, ultimately, to actually test them. It was alluded to in section 1 of this paper that this is part of what this project shall aim to do; to test some of the findings of CDA and see if the linguistic features that they have identified do, in fact, serve an ideological function in and of themselves. I will describe the methodology that will be used to do so presently.

3. Methodology

In the previous sections I have discussed what CDA is, the approaches and kinds of methodologies that are often used, and the findings of a selection of papers and books within the field. I then continued to talk about how, due to the heterogeneity of CDA research, there is a differing amount of evidence and different types of evidence for these features and the strength of the claims that are made in their favor vary wildly. For this reason, this project will aim to study the use of a few of these linguistic features in a number of texts in order to see if they do have some kind of objective ideological qualities or not. I have given an example of the kinds of strong claims that CDA

researchers have made in favor of nominalization, but “At all levels of text we may (…) postulate structures that preferentially affect the structure and content of mental models, as well as the more general and abstract forms of knowledge, beliefs, opinions, attitudes or ideologies that are shared by groups of recipients” (Van Dijk, 1995:23). So in this paper I will look at the real-world manifestations of some of these features.

The introduction to this paper introduced the topic of Romanian and Bulgarian immigration to the UK, a much debated topic that is clearly of a highly political nature, and so the articles that are chosen will be about this topic. The articles will be chosen from two sources which have the likely potential to represent the topic in different ways. I will use a combination of demographic data and a semiotic content analysis to show that there is difference between both the two news sources and between the individual texts themselves. Finally, a selection of linguistic features will be looked at in this set of articles to see if they have any objective ideological function by quantifying their use and

(24)

23 | P a g e

correlating this with the variables which differentiate the two news-sources. If one of these linguistic features occurs significantly more frequently in a certain type of text, or there are some clear

patterns in the distribution of these features, then we might say that it has some kind of objective ideological function. First, I shall provide a justification for why certain linguistic features have been chosen to be analyzed over others. Then I shall continue to discuss the data selection method, followed by the differences between the news sources in terms of demographics and the differences between articles regarding whether each article is ‘for’ or ‘against’ migration (the ‘tone’ of the articles). This will conclude the methodology section and lead on to the presentation of the results.

3.1. Which linguistic features to look for?

In section 2.4, The Findings of Critical Discourse Analysis, I presented an (incomplete) list of linguistic features that have been identified as serving ideological functions by CDA researchers. This list is very long and because of this it is not possible to analyze every feature in a project of this size and scope. It is therefore necessary to reduce this selection to a more manageable number. “How language can be used to mystify responsibility for social action is still a fixture for CDA” (O’Halloran, 2011:448) and for this reason I will be looking at occurrences of three related features that serve to remove agency and responsibility from a sentence and anonymize who it is who is performing the actions that are being talked about. These features will be nominalizations, passive sentences, and use of names. I have already mentioned the way that some researchers make strong claims about the role of nominalization in section 2.5. Here I shall go into further detail about what the features I will be looking at are, what is specifically ‘problematic’ about these types of constructions, and how they will be counted in the analysis of the data set.

3.1.1. Nominalization

Nominalization is the practice of “choosing noun phrases over verbs” (Billig, 2008:785). It is a “linguistic concept that Critical Linguists are vigilant of” (O’Halloran, 2011:448). An example of what

(25)

24 | P a g e

a nominalization is and how it can be problematic is using inflectional morphology to change ‘allege’ to ‘allegations’, which has the following effect:

“reflect on how much information goes unexpressed in a derived nominal, compared with a full clause: compare, for example, ‘allegations’ with the fully spelt-out proposition ‘X has alleged against

Y that Y did A and B [etc.]’”

(Fowler, 1991:80)

The nominalization also has the effect of ‘reifying’ the verb; giving it the status of a concrete object whereby “processes (…) assume the status of things” (Fowler, 991:80). Nouns, unlike verbs, can be quantified, you can have more of it than other people, and it can grow and change. Nouns also contain no agents, no timeframes, and no modality. It is in this way that they can be used to hide agency.

Not all nominalizations need to use derivational affixes, however. For instance you can challenge someone but also issue ‘a challenge’. You can swim and go for ‘a swim’. Therefore for the purposes of this study a nominalization will be counted if it is a verb form that is being used as a noun. If a word can be both a verb and a noun, then it will still be counted as a potential nominalization.

3.1.2. Passive Constructions

In English, we can talk about a difference between an ‘active’ and 'passive’ voice. The active voice follows the usual SVO structure of English and uses all three elements, whereas the passive moves the object into subject position. The subject therefore gets object position and is (optionally) omitted. See the following example:

The president has released the prisoners.

The prisoners have been released [by the president].

(26)

25 | P a g e

Here we can see that the passive voice is formed with the use of a form of ‘to be’ (been) and the past participle of the main verb (released). The grammatical verb does not have to be ‘be’, but can also be “equally well formed with (…) have, get, want, need, see and hear” (Edmonds, 2013:59). This allows the order of constituent parts to be altered and the object of the resulting sentence to be omitted, or included with the use of an optional prepositional phrase.

Even if the president is included in this sentence with the use of a prepositional phrase, the

movement of the agent to the end of the sentence changes what elements are in “the discourse role of Topic”, which is the “most important participant of the discourse at the point when the clause is produced” (Downing & Locke, 2006:253). This means that, as a linear structure, the order of the constituent parts changes what is read first and so slightly effects the understanding of the sentence because it highlights a different part of it.

For this project, a passive sentence will be counted if it fulfills the criterion that the object of the verb occurs in the first position. Due to the role that the linear structure of the sentence has on the interpretation, they will still be counted regardless of whether or not they use the optional

prepositional phrase.

3.1.3. Use of Names

The third and final agency-reducing feature that I shall be looking at is the use of personal names. This feature has been chosen as one possible way of quantifying the use of address terms. When referring to an individual, a reporter has the choice of using a person’s name or referring to them in terms of some sort of societal role. This role can be anything from their job to their nationality to their relationship to the journalist or a combination of these. For instance someone can be referred to as ‘one concerned resident’, ‘a representative of the government’, ‘an unnamed source’, etc. The use of such terms shows the relationship between the speaker and the person in question, and amount of deference and respect that the speaker thinks a person is due (Mavunga, Mutambwa &

(27)

26 | P a g e

Kutsaru, 2014:150-151). We can see how such a device may serve ideological functions in this way. Citing examples from literary sources such as Shakespeare’s Henry IV, Allan (2006) shows how use of informal address terms can be used for this purpose. When some characters address the king informally, they are indexing their relative position on the social hierarchy as being closer to that of the king’s. In English this is usually done lexically due to the lack of variant pronouns that can index social distance/relative position on the social hierarchy such as the Dutch U/Jij, German Sie/Du, and French Vous/Tu. Therefore “Naming or addressing a ruler often involves extreme pomp and

circumstance beset with euphemism” (Allan, 2006:1051), but if neutrality and clarity are goals of a journalist then names should be used over such titles. As well as this, use of address terms can also serve the function of reducing agency. If we consider the examples I have given such as ‘one

concerned resident’ etc. where the person behind the action is not known. Again, if explicit clarity is a journalist’s aim, then naming participants should be used.

Because this project aims to quantify language features, looking into the specific connotations of each address term would not fit into this methodology because it can be so subjective. In terms of hiding agency though, the form of address that is most transparent and the least anonymous is the use of a person’s name. Rather than hiding agency, the use of a person’s name is in fact maximizing agency and is the most identifiable and personal way of referring to someone. This is therefore what I shall be quantifying. In this project a use of someone’s name will be counted on every occasion that somebody’s first name, surname, or both are used.

These three features; nominalizations, passives, and use of names, will be the features that will be studied in this paper. I will now explain the data selection process for choosing the articles that comprise the data set.

(28)

27 | P a g e

3.2. Data Selection

Ten articles were chosen in total. Five of these were taken from the online version of the Daily Mail newspaper, MailOnline, and the other five were taken from the equivalent site for The Guardian newspaper which is stylized as ‘theguardian’. Bibliographic information about the articles is included in the table below. See Appendix I for links to the original articles.

Text Reference Publication Author Publication Date Wordcount DM1 MailOnline Jonathan Petre &

Simon Walters

28th_{DEC 2013} ₁₈₆₂

DM2 MailOnline Hugo Gye 31st_{DEC 2014} ₁₀₅₆

DM3 MailOnline Martin Robinson 30th_{DEC 2013} ₆₅₁

DM4 MailOnline Tim Shipman 30th_{DEC 2013} ₈₆₀

DM5 MailOnline Sanchez Manning 21st_{DEC 2013} ₇₀₉

G1 theguardian Hugh Muir 1st_{JAN 2014} ₁₃₆₁

G2 theguardian Alan Travis 21st_{NOV 2013} ₈₆₃

G3 theguardian Press Association

(anonymous)

30th_{DEC 2013} ₄₀₃

G4 theguardian Paul Quinn 31st_{DEC 2013} ₆₉₀

G5 theguardian Daniel Boffey 28th_{DEC 2013} ₇₈₂

Fig. 1 Bibliographic Information for the Data Set

The Daily Mail articles therefore have a total wordcount of 5138, whereas the Guardian articles have a total wordcount of 4100.

There were a number of factors that were considered during the data selection. This was mainly a balance between trying to be open and as undiscriminatory as possible with regards to the data (and so avoiding the over-selectiveness of previous CDA researchers) while at the same time trying create

(29)

28 | P a g e

criteria that would result in a data set that was a manageable size. The two newspapers were chosen because of their differing target demographics and political leanings, which will allow for a

comparison between the uses of CDA-identified ideological language devices. I will now explain the steps and decisions that were made when selecting this data, before continuing on to describe the form that the analysis will take.

The data was chosen on the basis of whether it was about a specific topic, namely the January 1st

removal of movement restrictions on Romanian and Bulgarian citizens wishing to come live and work in the United Kingdom as described in section 1. The reason for this topic is that it was (and continues to be) a divisive topic in UK political discourse. This should therefore lead to a large amount of ideological language being used, and so it is the perfect place to test the findings of CDA. The articles are also all from the month preceding January 1st_{, 2014, and not after that date.}

Although these events are fairly recent, having taken place at the beginning of the year in which this project is written, there is still enough of a temporal gap between then and now that we are able to greatly benefit from hindsight regarding how these two news outlets handled the events differently. The raw numbers of people who actually did arrive in the UK from Romania and Bulgaria was

practically negligible, with the Daily Mail reporting only 24 Romanians having entered the country as of January 14th_{, 2014 (Robinson, 2014), a whole two weeks after the EU-mandated removal of}

migration restrictions. Hardly the mass influx that the same paper was predicting in the lead-up to January 1st_{. Other outlets reported on how the Daily Mail handled the story during this time, calling}

them out for what were sometimes blatant lies (Greenslade, 2014). With this hindsight, we can see that the Daily Mail was not basing their stories on as rigorous fact-checking as other papers were. Their predictions were grossly overstated and, as I mentioned, their facts could be frankly untrue. On the other hand, The Guardian’s coverage was generally more balanced, or, at the very least, it wasn’t identified as being particularly unfair, biased, or wrong.

(30)

29 | P a g e

The different treatment of the same states of affairs is one dimension that we can use for comparison between the two news sources. We can also go into more depth about how each individual article treats the topic of immigration, and I will do so in section 3.3 where I will discuss how to identify the tone of individual articles. The two news sources also differ in terms of the audience that each paper has, which is what I will discuss presently in section 3.3. As mentioned previously these two dimensions, the demographics and ‘tone’ of the article, will serve as two correlates that I will compare with the number of instances of ‘ideologically’ loaded language to see if any patterns emerge and, therefore, if they can actually said to have inherent ideological qualities.

3.3. Demographic Comparison of the Two News Outlets.

An intuitive and qualitative comparison of the two newspapers is not enough on its own to say objectively that there is a difference between the two news outlets, nor can it say what the scale of difference is. Although one gets a good grasp on the comparative characteristics and virtues of news outlets just by being exposed to them for most of one’s life, there are also facts and figures that can more objectively show these differences. Figures 2 & 3 below are taken from Newsquest, a national marketing agency for UK newspapers, and they show what the average audiences for most of the main UK newspapers are like. In figure 2 below we can see that The Guardian has an average

readership that is in its early 30’s and trends towards the wealthier end of the spectrum. Conversely, the Daily Mail’s readership is older and poorer. It does not have the same working-class appeal of The Sun and the Daily Mail but it is close, certainly closer than the Guardian. The age and economical differences between the two groups suggests that we should see differences in the kinds of social attitudes and ideologies that the papers express.

Interestingly, the online versions of the papers are said to trend towards having similar, or at least more similar, demographics than their print counterparts, demonstrated by figure 3. Those who are likely to use the internet for their news tend to have a more similar socioeconomic status than those who use the printed media. The audiences shift towards this common point but they do not

(31)

30 | P a g e

converge completely, and because of the difference in the origin point (that is, the socioeconomic status of the audience of the original, print version) they do not converge completely. Also, because of brand identities and editorial consistency, we cannot expect all online news sources to share the same demographics. Furthermore, the National Readership Survey (Guardian, 2012) classifies both the print and online versions of The Guardian under ‘Quality Newspapers’ and the Daily Mail counterparts under ‘Tabloid and Mid-Market’. Therefore, the online versions of the Guardian and Daily Mail newspapers sill have suitably different audiences that allow for a comparison of language use. The number of readers (Guardian, 2012) for the online version of the Guardian is 6,410

thousand people and for the Daily Mail this is 6,820 thousand people, so the two sources are comparable in this way and mainly differ in terms of their readership demographics.

(32)

31 | P a g e

Fig. 3 Demographic Positioning of UK Online Newspapers (taken from Newsworks, 2014)

3.4. Identifying the ‘tone’ of The Articles

As mentioned previously, as well as breaking down the data by news source we can also break it down according to how the topic of immigration is presented in each article individually. That is to say that a news article can present the topic of migration in a positive, negative, or neutral way. The process of identifying the tone of the article feels like it should be a very intuitive one, and in many ways it is, but justifying these judgments and formalizing the process is nevertheless important.

First I should explicitly state what I mean by the ‘tone’ of the article. This project aims to find instances of specific language features and correlate this with the ideology of the text. This

difference in ideology might manifest as a result of the different audiences of the text, but it is also possible that each individual text might present the issue in a different way according to what the journalist thinks. Rather than being a result of audience design on an institutional level, the manifestation of ideology could be a result of what the individual author thinks about the topic of

(33)

32 | P a g e

immigration. This is what I mean by ‘tone’: what the author of the article conveys about the topic and whether this is expressed, in one word, as positively, negatively, or neutrally.

To identify the tone of the article, we look to the most salient features of the text and what they tell us. Why the salient features? We look at the salient features because they are the most ‘important’ features of the text when reading the text casually or quickly. They are the parts of the text that are ‘eye-catching’ or ‘attention-grabbing’. They are the largest, boldest, most visible aspects of the text and as such they contain the general theme or ‘gist’ of the article. In other words, this is the principle that the most perceptually salient parts of the text are a distilled version of the full text. Again, this seems intuitive, but there is research and theory that can justify this intuition.

We can look to eye-tracking and attention studies such as Bucher & Schumacher (2006) to see that, when reading online newspaper articles, people spend the first 5 seconds in an “orientation phase” in which “Readers try to identify the medium and their own position within that medium” (Bucher & Schomacher, 2006:359). This is followed by “a clear concentration on elements such as page header, newspaper header, and the picture of the main article” (Bucher & Schumacher, 2006:360). Both this paper and Porta, Ravarelli & Spaghi (2012:414) note a phenomenon which is known in such studies as “banner blindness”. This is where a reader will ignore certain parts of the text which are expected to be filled with advertisements, which a reader is typically not interested in. Instead, they go for the most salient parts of the content they are interested in, which are the most salient parts of the article itself.

As well as eye-tracking studies, we can use pragmatic knowledge of the discourse structure of newspapers to justify which parts of the text are most important. The discourse of newspapers has particular set features that serve certain purposes and readers know this. People know through experience of newspapers that the headline will summarize the story and that the images and associated captions that are used will be related to the story. Journalists are taught about use of features such as lead paragraphs to provide further detail and summarization after the headline. In

(34)

33 | P a g e

fact the roles of such devices are not necessarily known only to journalists, as writing newspaper articles is often included as part of GCSE and A-Level English qualifications in the UK as part of learning about style and genre. This means that many people are taught how to use these features at some point in their mandatory education. The eye-tracking experiments and the discourse structure of news articles back up the intuitive assessment of what are the most salient parts of a newspaper article. These are the features that will be analyzed in order to identify an article’s tone. Specifically I will be looking at the article’s headline, sub-headline(s), lead paragraph, and leading image(s) and caption(s).

The text-based parts of the text can be analyzed through content analysis of what the text is explicitly stating with regards to the migration. The image-based parts of the text will require a different and more semiotically-based approach. It is not enough to just describe what the images are showing, but what that means in terms of how the subjects of the images are being portrayed and how this relates to the story in general. What is literally being shown, the description of the subjects of the image, is called the image’s denotation. What this represents and what it means for the tone of the article is known as the image’s connotation.

“Denotation refers to a word’s definition, while connotation refers to the emotions associated with the word”

Lawton (2011:136)

The above quote from Lawton is describing the difference between denotation and connotation of words, lexical items, but the same definition can be applied to visual mediums as well. For instance, the word ‘rose’, or the image of a rose, has the denotation of a type of flower (or one specific flower from that type, in the case of an image) but also has connotations, emotional associations, with love and romance.

(35)

34 | P a g e

The meaning of images is therefore based on the connotations of the content that is denoted and how these connotations effect our interpretation of the image’s subject(s). In images that focus on the face, such as in photos of politicians, the posture, gesture, and facial expression also have emotive meaning (Feng & O’Halloran, 2012). For instance, a portrait photograph with classic ‘mugshot’ framing (the kind of photo taken of criminals when they are arrested) brings with it all kinds of connotations of criminality and guilt due to its typical use of cataloging criminals. These connotations can be used strategically outside of law-keeping situations to attribute guilt where there may not be any, and so the ‘mugshot’ by itself is “a signifier widely used to exclude or deride certain groups” (Lashar, 2014:85). Images can be used in this way as a kind of “visual rhetoric”, meaning that “it is possible to argue visually […] it is possible for an image to advance and defend standpoints” (Richardson & Wodak, 2009b:251).

4. Results

The results for the quantification of features are presented in the following way: First, the results of the semiotic analysis are presented in summarized form in a table. This is then used to arrange the results of each graph, so that the articles are arranged with ‘negative’ articles on the left, ‘neutral’ in the middle’, and naturally ‘positive’ on the right. The numbers in the graphs are the absolute

numbers of the occurrences of each feature. Below each graph is a table which averages the results, both as a mean average for number of occurrences of a feature in each news-source and as an average per 1000 words.

4.1. Semiotic Analysis

The summarized results of the surface semiotic analysis, the tone of each individual article, follows:

Article: DM1 DM2 DM3 DM4 DM5 G1 G2 G3 G4 G5

Tone: - - - + = = + +

(36)

35 | P a g e

In Fig. 4, the tone of the article is represented with one of three symbols. A negative result is represented with a subtraction symbol (-), positive with an addition symbol (+), and an article which is generally more balanced or presents the topic both positively and negatively is represented with an equals symbol (=).

These results show that the Daily Mail is consistently negative towards the migration, whereas the Guardian is more varied. The Guardian in general represents the migration more positively than the Daily Mail. This data further supports the claim that the two papers have different ideologies, but the addition of three variables (positive, negative, neutral) might show trends and correlations more clearly than just two (Daily Mail vs Guardian).

The analysis was conducted by looking at what is contained in the most visually salient parts of the article, both text and image. The full analyses for each of the articles are contained in Appendix III, arranged from DM1 to 5 and followed by G1 to 5. The analysis presents a copy or description of the Headline(s), Lead Paragraph, Photo and image, followed by and Analysis and Conclusion for each article. I will present a sample of one of the analyses here, in order to illustrate the specifics of what the analysis involved for each article:

Sample Analysis of the Article DM2

Headline(s):

“Roma already in Britain ‘are defecating on people’s doorsteps’ says top Tory council leader as she warns of burden that Romanian and Bulgarian immigrants will place on public services.”

Sub-headline bulletpoints:

“Philippa Roe of Westminster City Council says council taxpayers will face rising bills from Bulgarian and Romanian immigration.

'Ideology' in the News: A Critique of Critical Discourse Analysis