Ideology Symposium Oxford, September 2004.



Ideology and Discourse Analysis


Teun A. van Dijk

Universitat Pompeu Fabra, Barcelona


Second version. November 2004




In this paper I discuss some of the issues raised by a discourse analytical approach to the study of ideology. Since people acquire, express and reproduce their ideologies largely by text or talk, a discourse analytical study of ideology is most relevant.

Although the focus in this paper is on the discourse-ideology interface, its theoretical framework is multidisciplinary, articulated by the fundamental triangulation of discourse, cognition and society (van Dijk, 1998a). This framework is critical of the traditional approaches to ideology -- especially in the social sciences and philosophy -- that fail to adequately theorize about the sociocognitive nature and structures of ideologies and their discursive reproduction. This paper, however, shall not extensively deal with these classical approaches to ideology (among many historical studies of ideologies, see, e.g., Larrain, 1979)


Theory of ideology


Before I deal with an analysis of the relations between ideology and discourse, it is necessary to briefly summarize the theoretical framework in which these relationships are to be made explicit (for details, see Van Dijk, 1998a).


Defining ideology


The first assumption is that, whatever else ideologies are, they are primarily some kind of ‘ideas’, that is, belief systems. This implies, among other things, that ideologies, as such, do not contain the ideological practices or societal structures (e.g., churches or political parties) that are based on them. It also implies that a theory of ideology needs a cognitive component that is able to properly account for the notions of ‘belief’ and ‘belief system,’ for instance as these are dealt with in contemporary cognitive science.

Secondly, just as there are no private languages, there are no private, personal ideologies. Hence these belief systems are socially shared by the members of a collectivity of social actors. However, not any collectivity develops or needs an ideology, and it will be argued that this is only the case for some kinds of group -- typically so in relation to other groups – and not for instance for communities, such as cultural, national or linguistic communities. In other words, ideologies consist of social representations that define the social identity of a group, that is, its shared beliefs about its fundamental conditions and ways of existence and reproduction. Different types of ideologies are defined by the kind of groups that ‘have’ an ideology, such as social movements, political parties, professions, or churches, among others.

Thirdly, ideologies are not any kind of socially shared beliefs, such as sociocultural knowledge or social attitudes, but more fundamental or axiomatic. They control and organize other socially shared beliefs. Thus, a racist ideology may control attitudes about immigration, a feminist ideology may control attitudes about abortion or glass ceilings on the job or knowledge about gender inequality in society, and a social ideology may favor a more important role of the State in public affairs. Hence, ideologies are foundational social beliefs of a rather general and abstract nature. One of their cognitive functions is to provide (ideological) coherence to the beliefs of a group and thus facilitate their acquisition and use in everyday situations. Among other things, ideologies also specify what general cultural values (freedom, equality, justice, etc.) are relevant for the group.

Fourthly, as the sociocognitive foundation of social groups, ideologies are gradually acquired and (sometimes) changed through life or a life period, and hence need to be relatively stable. One does not become a pacifist, feminist, racist or socialist overnight, nor does one change one’s basic ideological outlook in a few days. Many experiences and discourses are usually necessary to acquire or change ideologies. The often observed variability of ideological opinions of group members, thus, should be accounted for at the personal or contextual level, and is no ground to reject the notion of a shared, stable group ideology (Converse, 1964).

Also the reverse is true: If ideologies can be gradually developed by (members of) a group, they also gradually disintegrate, e.g. when members no longer believe in a cause and ‘leave’ the group, when group grievances have been attended to, or under a host of other social and political conditions, as e.g., was the case for the pacifist and anti-nuclear movements of the 1970s (Curran, 2003; Oberschall, 1993; Van der Pligt, 1992).

Sometimes, ideologies become shared so widely that they seem to have become part of the generally accepted attitudes of an entire community, as obvious beliefs or opinion, or common sense. Thus, much of what today are widely accepted as social or human rights, such as many forms of gender equality, were and are ideological beliefs of the feminist or socialist movements. In that sense, and by definition, these beliefs thus lose their ideological nature as soon as they become part of the Common Ground. It is only in this sense that I would agree to use the phrase ‘the end of ideology’, since obviously our contemporary world is replete with ideologies, especially so by those who deny they are ideologies.


What ideologies are not.


These are quite general properties of ideologies, but already define them in a way that is different from several other approaches. Thus, ideologies are not personal beliefs of individual people; they are not necessarily ‘negative’ (there are racist as well as antiracist ideologies, communist and anticommunist ones); they are not some kind of ‘false consciousness’ (whatever that is exactly); they are not necessarily dominant, but may also define resistance and opposition; they are not the same as discourses or other social practices that express, reproduce or enact them; and they are not the same as any other socially shared beliefs or belief systems (among the many studies of ideology that defend one or more of these theses, see, e.g. Billig, 1982;  Eagleton, 1991; Larrain, 1979; Thompson, 1984; this article however is intended as a debate with other approaches, so I shall not detail this overall criticism of traditional approaches).


The social functions of ideologies


Ideologies, thus defined, have many cognitive and social functions. First of all, as explained above, they organize and ground the social representations shared by the members of (ideological) groups. Secondly, they are the ultimate basis of the discourses and other social practices of the members of social groups as group members. Thirdly, they allow members to organize and coordinate their (joint) actions and interactions in view of the goals and interests of the group as a whole. Finally, they function as the part of the sociocognitive interface between social structures (conditions, etc.) of groups on the one hand, and their discourses and other social practices on the other hand.

Some ideologies may thus function to legitimate domination, but also to articulate resistance in relationships of power, as is the case for feminist or pacifist ideologies. Other ideologies function as the basis of the ‘guidelines’ of professional behavior – for instance for journalists or scientists.

I shall say very little on the many social and political functions of ideologies, especially since these have received ample attention in classical theories (see the references given above; for the role of liberalism and socialism in politics, see, e.g. Freeden, 1996). It should be emphasized however that the cognitive structures of ideologies need to be tailored to their societal functions for (the members of) groups.



Lacunae and open questions


Although this theory accounts for many general properties of ideologies, there are still important lacunae and open questions.


The structures of ideologies


The most important one is the precise cognitive nature of ideologies: what are their precise contents and structures? If socialism, feminism and neoliberalism are ideologies, what exactly do they look like? This very general question is equally difficult to resolve as the fundamental question about the precise structures of knowledge.

On the basis of the ways ideologies organize social attitudes, such as those on immigration or abortion, and in view of their social functions, I have hypothesized a general ideological schema consisting of a handful of basic categories defining the self-identity of groups (identity criteria, typical activities, aims, norms and values, related groups and basic resources – or lack of resources). These explain many properties of ideologies, but I am not sure they apply to all ideologies equally well, such as religious ideologies, political ideologies or professional ideologies, among others. What does seem the case though is that in order to be able to be acquired and used, ideologies need some kind of organization, as is also the case for other belief systems. Thus, it is unlikely that they are merely long, unordered sets or mere lists of beliefs (the notion of organization of cognitive structures, e.g., in terms of schemata, is a major characteristics of modern cognitive psychology; see, among many other studies, e.g. Anderson, 1980; Schank & Abelson, 1977).

The assumption about the organized nature of ideologies does not imply that they are in any way consistent. They are not logical systems, but socio-psycho-logical ones. So they may very well be heterogeneous or inconsistent, especially in their first, more or less spontaneous stages, although various ideologues (writers, leaders, teachers, preachers, etc.) may try to enhance the coherence by explicit manifestoes, catechisms, theories, and so on. Thus, although ideologies organize other social beliefs of groups, this does not mean either that these other social beliefs are consistent, as we also know from the well-known racist beliefs that hold that immigrants are lazy and do not want to work, and at the same time that they take away our jobs (see also the discussions about the nature of political cognition, e.g., in Lau & Sears, 1986; see also Feldman, 1988). We also know that people use several strategies to eliminate or disregard inconsistencies between ideological beliefs and the ‘facts’ with which they are confronted. (Zanna, Klossen, & Darley, 1976).


Ideologies only as ‘foundational’ beliefs of a group?


Another point I am not yet sure about, is whether ideologies should be identified only in terms of foundational group beliefs, as I have done above, or should be conceived of more broadly as consisting of all ideological group beliefs, that is, including the more specific group knowledge and attitudes.

I prefer the first ‘axiomatic’ option, first of all, because it rules out the possibility that mere personal opinions or a single group attitude (say about nuclear energy) would as such be called an ‘ideology’. Secondly, by limiting ideologies to fundamental beliefs, we allow variations or changes of less fundamental beliefs within the ‘same’ ideology – much in the same way as personal and regional variants exist of the ‘same’ language. Instead of ‘foundational’ beliefs, we may also speak of ‘core’ beliefs, whatever theoretical metaphor is more useful. In the latter case, more specific attitudes based on such core ideological beliefs need to be described as more peripheral.


Are ideologies ‘known’ by all of their members?


Thirdly, although ideologies by definition are socially shared, obviously not all members of groups ‘know’ these ideologies equally well. As is the case for natural languages, there are differences of ‘expertise’ in a group. Members are able to speak or act on the basis of the acquired ideology, but are not always able to formulate its beliefs explicitly. On the other hand, there are experts, teachers, leaders and other ‘ideologues’ who teach, explain, inculcate and explicitly reproduce the group ideologies (for differences between ‘experts’ and ‘novices’ in political cognition, see, e.g., Fiske & Kinder, 1981).

In the same way, we might assume that not all members identify with an ideological group in the same way, and equally strongly (Vallacher & Wegner, 1989). This suggests that individuals may be ‘more or less’ members of ideological groups – and that the notion of an ideological group is defined a fuzzy set of social actors. Variations of expertise and identification need to be assumed to account for the empirical facts and to provide the necessary flexibility to the theory.





The social basis of ideologies: What kind of collectivity?


Finally, one of the more difficult problems concerns the exact social basis of ideologies. It has been assumed that they are properties of ‘social groups’ and that these groups may be fuzzy sets. But obviously not all social collectivities are ‘ideological groups’. The passengers of a bus are not, nor are the professors of some university. Hence, a number of social criteria about permanence, continuity, social practices, interests, relations to other groups, and so on, need to be satisfied, including the fundamental basis of group identification: a feeling of group belonging that is typically expressed by the pronoun we.

Some social groups may only or primarily be defined in terms of their shared ideologies, social representations and the discourses and other social practices based on them, as is the case for feminists and racists. Other groups, such as political organizations or professional groups, may not only share a (professional, political) ideology, but may be further organized by explicit membership, membership cards, meetings, institutions, organizations, and so on.

In this perspective it seems relevant to distinguish between (various kinds of) social groups, on the one hand, and cultural communities, on the other hand. The first have ideologies – related to their goals and interests in relation to other groups – whereas the latter have other general beliefs, such as knowledge, norms and values – which need not be related to those of other cultural communities. Thus, the speakers of English are a cultural (linguistic) community, whereas the teachers of English are a social (professional) group. The first have, as such, no ideology, whereas the second may well have one.

Further theoretical work needs to be done on the kind of collectivities that share ideologies. Provisionally, I called them ‘groups’ and distinguished them from communities. They should also be distinguished from social categories, such as those of gender or ethnicity: women or black people do not have ideologies, but feminists or antiracists do. But feminists and pacifists are a different kind of collectivity than the usually more organized professional or political collectivities. Racists or conservatives hardly seem to form a ‘group’ in the sense of an organized collectivity of people, as is the case for a racist party. They are more like ‘communities of belief’ than groups that coordinate their actions. However, feminists, pacifists, racists and conservatives not only are collectivities that share ideological beliefs. They also act upon them, and talk among themselves and with others as a function of their ideologies. They may more or less explicitly identify as such, and defend their views and others who have these views. Sometimes, as is the case for racists, they may neither realize nor avow they are so – which also show that labels of ideologies need not be self-attributed.

In other words, ideological collectivities are also communities of practice, and communities of discourse. They may or may not organize themselves as political parties or organizations. It is not the party or the club as such that are ideological as organizations, but the collectivity of people who are their members.

For these reasons I decided to provisionally adopt the term ‘ideological group’, that is, a collectivity of people defined primarily by their shared ideology and the social practices based on them, whether or not these are organized or institutionalized. Other groups, such as professionals, may first organize themselves, e.g., to promote or protect their interests, and develop (professional) ideologies to sustain such activities.

We see that a broader theory of social organization, e.g., in different kinds of collectivities, is closely related to a sociocognitive theory about the kind of beliefs or social representations of these collectivities. It is also for this reason that I distinguish between epistemic or linguistic communities, on the one hand, and ideological groups or organizations, on the other hand. But we also see that a further typology of ‘ideological groups’ may be necessary, e.g., in terms of their organization, permanence, joint actions, as well as the nature of the ideologies themselves: a pacifist social movement as a different kind of ideology than for instance the religion shared by the members of a church.


Ideology and discourse processing


Ideologies have been defined as foundational beliefs that underlie the shared social representations of specific kinds of social groups. These representations are in turn the basis of discourse and other social practices. It has also been assumed that ideologies are largely expressed and acquired by discourse, that is, by spoken or written communicative interaction. When group members explain, motivate or legitimate their (group-based) actions, they typically do so in terms of ideological discourse.

It is however one thing to assume that ideologies are ‘at the basis’ of discourse, and quite another to provide a detailed theory of the actual (cognitive) processes involved in the production or understanding of such ‘biased’ discourse. Indeed, how exactly do we ‘know’ a racist, or sexist, or neoliberal discourse when we read or hear it? To answer this question, I shall again summarize some basic assumptions, partly rooted in the contemporary theory of discourse processing in cognitive psychology, partly based on new assumptions extending such a theory (for details, see Van Dijk & Kintsch, 1983; Van Oostendorp & Goldman, 1999).




Language use in general, and discourse production and comprehension in particular depend on, and influence, the relevant properties of the communicative situation as interpreted by language users. These subjective ‘definitions of the situation’ or ‘contexts’ are represented as specific models in episodic memory: so called context models (Van Dijk, 1999. These context models control many aspects of discourse processing and make sure that a discourse is socially appropriate. As is the case for all subjective mental models, context models may be ideologically ‘biased’ by underlying attitudes that are themselves ideological. Biased context models may have biased discourses as a result, for instance in a more or less polite tone or lexical choice. Thus, the way some men speak to or about women obviously will (also) depend on the way they represent women in general, and a female interlocutor in particular, as we know from machismo in text and talk. The same is true for the comprehension of discourse, which also depends on the way speakers are being perceived as ideologically biased or not.




The meaning or ‘content’ of discourse is controlled by subjective interpretations of language users of the situation or events the discourse is about, that is by their mental models (Johnson-Laird, 1983; Van Dijk & Kintsch, 1983). People understand a discourse if they are able to construct a model for it. Thus, news on the war in Iraq is typically produced and understood on the basis of the subjective models of writers and readers about this war. As is the case for context models, also these ‘event models’ may be ideologically biased, again on the basis of underlying, socially shared attitudes and ideologies. Ideologically biased event models typically give rise to ideological discourses, in which events or actors are described more or less negatively or positively, depending on the ideological bias of the mental model. This is especially the case for all discourses about specific events and actions, such as news reports, editorials, opinion articles and everyday stories about personal experience.




Whereas context and event models are personal and subjective, members also share more general, social beliefs, such as knowledge, attitudes and ideologies. These general beliefs control the construction of specific models and hence, indirectly, also the production and understanding of discourse. Most fundamental is the shared knowledge of various types, levels or scopes of communities (culture, nation, city, organization). Because this knowledge is by definition acquired and shared by all competent members of these communities, it is normally presupposed. In this sense, discourse meanings are like icebergs of which only part of the non-presupposed meanings are explicitly expressed. Since communities have been assumed to be non-ideological for their own members, their knowledge is also non-ideological within the community. This means that even for different ideological groups and speakers of a community, its knowledge is taken for granted. It is this shared, presupposed, taken for granted knowledge that makes discourse and communication and mutual understanding possible, also across ideological group boundaries. In other words, according to this theory, all other socially shared beliefs, and hence also ideologies of groups, are based on, and presuppose the general knowledge of the community. The same is true for the construction of mental models and the discourses based on them: the interpretation of meaning and reference is a process in which various kinds of knowledge are activated and formed. Of course, members of other communities may describe the taken for granted knowledge of a community as ‘mere belief’ (superstition, etc.) and hence also as ideological. And what may be described as ‘religious belief’ today, may have been presupposed as ‘knowledge’ in an earlier phase of a cultural community. Note finally that the notion of knowledge as it is used here is defined at the level of the community, and not in terms of individual beliefs.

That is, just as is the case for ideologies, individual members of a community may know more than others, for instance as a result of different education. That is, within a community, knowledge may be stratified or differentiated, with various types of expertise. The word ‘gene’ may be used in everyday language and in the newspaper or other public discourse, but it may be assumed that less people know what ‘genes’ are than people who know what cars and televisions are. In other words, as a macro level description of shared beliefs by a group, the concept of knowledge is an abstraction and an idealization, as is the case for the concept of a natural ‘language’ shared by a community, and which is actually ‘known’ and used in quite different ways by the members of the community.

One might venture a ‘minimalist’ hypothesis for knowledge as one might do so for language – in terms of the everyday beliefs all competent members share, whatever their education. However, this again requires a definition of cultural ‘competence’ or ‘normalcy’ which requires a similar kind of analysis. Another option is to use an average base level, which might be defined for to all members of the community who have finished obligatory schooling.

In other words, the ‘shared’ nature of knowledge in a community needs to be further defined, both cognitively and socially. For the moment, thus, we take the practical decision that knowledge are the beliefs of a community that are presupposed in its public discourses directed at the community at large, as is the case for most discourses of the mass media.


Group beliefs


On the other hand, within communities groups may form on the basis of different aims, goals, interests or practices. These groups also share beliefs, such as knowledge, attitudes and ideologies. Thus, racists may share the belief about white superiority, a belief that they may take for granted, and hence define as ‘knowledge’. Groups may also have more complex evaluative belief complexes, such as attitudes about immigration, abortion or euthanasia (Van Dijk, 1984, 1987, 1993).

Group beliefs are characteristically ideological, in the sense that they are controlled and organized by underlying ideologies. They control the context and event models of their members when these speak as group members, and thus indirectly the discourse structures controlled by these biased models. In some types of discourse, general group beliefs may influence discourse directly, and not via specific mental models. This is for instance the case in political propaganda, sermons, and other ideological discourses that feature general beliefs of a group.

Ideological group beliefs take different forms, depending on their social functions. Some beliefs may be expressed in order to influence social policy or promote a cause, as may be the case for feminists, antiracists or pacifists. Other beliefs focus more on the norms and values of everyday practices of group members, as is typically the case of professional groups, such as the methods of research of scholars or the medical practices of doctors. Although the functions may differ, I shall assume these practices have the same ideological basis. Thus, the way a macho man may treat a woman is controlled much in the same way as a traditional doctor may treat his or her patients, e.g. presupposing ignorance. And feminists may get organized much in the same way as professionals do. In others words, although we may distinguish different kinds of social groups and different kinds of ideologies with different functions, I shall assume that we also need a general theory of ideology and its functions that abstracts from such differences.


Strategic processing


On the basis of the ideologically biased models and socially shared beliefs discussed above language users strategically produce and understand talk and text, on line, word by word, sentence by sentence, turn by turn (Van Dijk & Kintsch, 1983). All variable phonological, lexical or syntactic forms may thus be controlled by the underlying representations, as is also the case for the local and global meanings and the actions engaged in by the language users. Intonation, pronouns, nominalizations, topic choice and change, level of specificity or precision of action or actor description, implicitness, turn taking, interruptions, politeness, arguments and fallacies, narrative structures, style or rhetorical figures, among a host of other discourse structures may thus strategically ‘index’ the ideology of the speaker or writer. As we shall see below, the general strategy controlling these various structures and moves is based on the underlying ingroup-outgroup polarization of ideologies: Our good things and Their bad things will tend to be emphasized, as is the case for the mitigation of Our bad things and Their good things.


From ideology to discourse


From this brief summary of some of the underlying sociocognitive processes underlying the production and comprehension of ideological discourse, we see that the relation between ideologies and discourse is complex and often quite indirect. Such discourse may depend on ideologically biased contexts, on the ideological way participants interpret events as subjective mental models, or more directly on general group beliefs that are ideologically controlled.

This theory also allows that given specific contextual conditions, speakers may of course hide or dissimulate their ideological opinions. Indeed, feminists, antiracists or pacifists do not always show their opinions, even in situations where this would be relevant and appropriate. That is, discourse is not always ideologically transparent, and discourse analysis does not always allow us to infer what people’s ideological beliefs are. This always depends on the definition of the communicative situation by the participants, that is, on context. In other words, our concept of ideology is non-deterministic: Members do not necessarily and always express or enact the beliefs of the groups they identify with. Also ideological discourse is always personally and contextually variable.

This is strategically true in international negotiations and bargaining situations in which the suspension of explicit ideological statements may be relevant. This does not mean that ideologies are increasingly less important in a globalized world, but only that in some contexts they are not being manifest so that conflicts can be resolved more easily.

Obviously, this also has considerable influence on the empirical methodologies for the study of ideology. In ingroup talk, ideological beliefs may be presupposed, and in talk with outgroup members ideological beliefs may be censored or modified, e.g. in ‘politically correct’ discourse. In both cases, the relation between ideology and discourse needs special, indirect or other unobtrusive methods to be studied empirically.


Ideological structures and strategies of discourse


If ideologies are acquired, expressed, enacted and reproduced by discourse, this must happen through a number of discursive structures and strategies. For instance, the pronoun we is one of these structures, typically used to deictically refer to the ingroup of the current speaker. In theory, and depending on context, any variable structure of discourse may be ideologically ‘marked’. Specific intonation, stress or volume in the expression of a word or phrase may be interpreted as sexist or racist. Preference for specific topics may express a macho or a neoliberal ideology, and so on.

It should be stressed that ideologies may only influence the contextually variable structures of discourse. Obviously the obligatory, grammatical structures cannot be ideologically marked because they are the same for all speakers of the language and in that sense ideologically neutral. However, there may be some debate on whether some general grammatical rules are really ideologically innocent, as is the case for expressions of gender. Some variable structures are more ideologically ‘sensitive’ than others. For instance meanings are more prone to ideological marking than syntactic structures, because ideologies are belief systems and beliefs characteristically tend to be formulated as meanings of discourse. Syntactic structures and rhetorical figures such as metaphors, hyperboles or euphemisms are used to emphasize or de-emphasize ideological meanings, but as formal structures they have no ideological meaning. Thus, there is no specific racist or antiracist type of hyperbole, pronominalization, or intonation – although there are preferences for racist or sexist metaphors – only the meanings that are modified by them.

We have assumed that ideological discourse structures are organized by the constraints of the context models, but also as a function of the structures of the underlying ideologies and the social representations and models controlled by them. Thus, if ideologies are organized by well-known ingroup-outgroup polarization, then we may expect such a polarization also to be ‘coded’ in talk and text. This may happen, as suggested, by pronouns such as us and them, but also by possessives and demonstratives such as our people and those people, respectively.

Thus, we assume that ideological discourse is generally organized by a general strategy of positive self-presentation (boasting) and negative other-presentation (derogation). This strategy may operate at all levels, generally in such a way that our good things are emphasized and our bad things de-emphasized, and the opposite for the Others – whose bad things will be enhanced, and whose good things will be mitigated, hidden or forgotten.

This general polarizing principle when applied to discourse affects both forms and meanings. Thus, we may enhance the negative properties of terrorists by reporting gruesome acts of them (a question of meaning or content), but then do so at great length, on the front page, with big headlines, with gruesome pictures, repeatedly so, and so on, which are formal characteristics. We may also do this by syntactic means, for instance by reporting their gruesome acts as being accomplished by active, responsible agents, that is by referring to them in first, topical positions of clauses and sentences, and not as implicit agents or in passive sentences in which agents are de-emphasized.

In other words, there are many discursive ways to enhance or mitigate our/their good/bad things, and hence to mark discourse ideologically. In critical discourse analysis (CDA) this is more or less standard theory. We may summarize some of the ways discourses code for underlying ideologies in Table 1, emphasizing that no discourse structures uniquely code for one communicative and interactional function – that is, all these structures may also be used for other reasons and functions. The table can feature only a small selection of discourse structures. It is organized by discourse levels, such as those of global and local meanings, lexicon, syntax, sound structures, formats, rhetorical structures and interactional structures – that is, by Forms, Meanings and Actions. At each of these levels we may find codification of underlying ideologies, and this generally happens by emphasizing or de-emphasizing in many ways the good or bad properties of Our own group and the good or bad properties of the outgroup (for critical studies of ideology in discourse, see, e.g., Caldas-Coulthard & Coulthard, 1996; Fairclough, 1995; Fowler, 1991; Kress & Hodge, 1979; Van Dijk, 1995, 1998a, 1998b, 1999; 2001, 2004; Wodak, 1989; Wodak & Meyer, 2001).

Table 1




·         CONTEXT: Speaker speaks as a member of a social group; and/or addresses recipient as group member; ideologically biased context models: subj. representations of communicative event and its participants as members of categories or groups.



Overall strategy: positive presentation/action of Us, negative presentation/action of Them

·         Emphasize Our good things, and Their bad things, and

De-emphasize Our bad things, and Their good things



o        Topics (semantic macrostructures)

§         Select/Change positive/negative topics about Us/Them.

o        Local meanings and coherence

§         Positive/Negative Meanings for Us/Them are

·         Manifestation: Explicit vs. Implicit

·         Precision: Precise vs. Vague

·         Granularity: Detailed/fine vs. Broad, rough

·         Level: General vs. Specific, detailed

·         Modality: We/They Must/Should…

·         Evidentiality: We have the truth vs. They are misguided

·         Local coherence: based on biased models

·         Disclaimers (denying Our bad things): ‘We are not racists, but..’

o        Lexicon: Select Positive/Negative terms for Us/Them (e.g., ‘terrorist’ vs. ‘freedom fighter’)



o        Syntax: (De)emphasize Positive/Negative Agency of Us/Them

§         Cleft vs. non-cleft sentences (“It is X who…”)

§         Active vs. Passives (“USA invades Iraq” vs. “Iraq invaded by USA”)

§         Full clauses/propositions vs. nominalizations (“The invasion of Iraq”).

o        Sound structures: Intonation, etc. (de)emphasizing Our/Their Good/Bad things

o        Format (schema, superstructure: overall form)

Positive/Negative meanings for Us/Them in

·         First, dominant categories (e.g., Headlines, Titles, Summaries, Conclusions) vs. last, non-dominant categories.

·         Argumentation structures, topoi (stereotypical arguments, e.g., ‘For their own good’)

o        Fallacies that falsely conclude Our/Their Good/Bad things, e.g. overgeneralizations, authority, etc.


o        Rhetorical structures

Emphasizing or de-emphasizing Our/Their Good/Bad things by

§          Forms: Repetition

§          Meanings: Comparisons, metaphors, metonymies, irony; euphemisms, hyperboles, number games, etc.


o        Speech acts, communicative acts, and interaction

§         Speech acts that presuppose Our/Their Good/Bad things: promises, accusations, etc.

§         Interaction strategies that imply Our/Their Good/Bad things: Cooperation, agreement

Problems of ideological discourse analysis




Within the framework of ideological discourse analysis presented above, there are a number of issues and problems that need special attention.

The first problem I would like to raise may be labeled with the controversial concept of ‘intentionality’, understood here in the action-theoretical sense, and not in the philosophical sense of ‘aboutness’, e.g. of propositions (Brand, 1984). Both as ideological participants as well as analysts we may ask ourselves whether specific discourse features, such as passive sentences or nominalizations are ‘intentional’ aspects of ideological discourse, or whether such structures are largely automatized and hence hardly consciously controlled. Participants may know such situations when they are criticized for a sexist or racist word and then defend themselves by affirming ‘that they did not mean it that way’, or that such was ‘not their intention’.

Also for these reasons, many tenets in interaction analysis exclude intentions from their objects of research, e.g., with the further argument that intentions cannot be directly observed, and secondly that what socially counts is how discourses are understood (“heard”) by participants, whatever the intentions of the speaker (see, e.g., Duranti, 2001). This is also a principle of some legal and political treatments of social actions and discourse – whether or not one intends a sexist, racist or other offensive remark, one is bound to know the possible interpretations and hence the consequences of one’s discourse. In other words, what counts are (foreseeable) social consequences, and not (good or bad) intentions.

Whereas socially and legally such may be a legitimate position, theoretically it poses problems. Indeed, there is no a priori ground why subjective interpretations of recipients should be favored above the equally subjective intentions of speakers. Indeed, none of them are ‘observable’, and both are equally vague notions. In cognitive terms, both are (fragments of) mental models of some kind. Speakers may ‘defend’ their ‘good’ intentions by referring to other discourses and acts, or by citing sources who can vouch for their intentions. Recipients are strengthened in their interpretations if they are not the only ones who interpret a discourse in a specific way. If speakers and recipients are ideological opponents, such a situation may typically lead to a conflict – some of which even wind up in court, for instance when someone publishes a racist article or makes a racist remark in a public speech.

Theoretically, thus, it should be borne in mind that as such words, phrases, topics or intonations, are not ideologically biased. It is their specific use in specific communicative situations that make them so – as we know also from the use of obvious racist words as nigger by, say, a member of the KKK, or by a African American leader, respectively. Part of this context is not only who is speaking to whom and in what role, but also the intentions of the speaker, whether or not these are made explicit in discourse. Of course, by their frequent ideological use, words may have strong associations with ideological meanings, as is the case for “free market”. However, it is always the text and context, and hence the uses of words, that are ideological: the word “free market” may be used as frequently in a neo-liberal as in an anti-neo-liberal treatise, one position in favor, the other against.

Thus, if the underlying mental models or social representations of speakers are not controlled by some ideology, then by definition also the intentions and the mental model of the context, and hence the discourse cannot be ideologically biased. Recipients who do not know the speaker (and her social representations, ideologies) may thus very well misunderstand or misinterpret the speaker. Often such misunderstandings may be resolved by further questions, explanations or negotiation (“What do you mean…?”).

In other words, whether or not we adopt a social or legal position in which ‘good intentions’ do not count in discourse, but only how discourse comes across and is interpreted, it needs to be stressed that theoretically we need both intentions and interpretations as part of ideological communication. Only then are we able to explain ideological disputes, problems, conflicts and other properties of ideological interaction. Intentions are no more or less mysterious than interpretations – they are two of a kind, namely subjective mental models of participants. And only then are we able to address more detailed questions such as which properties of discourse can be consciously controlled, and which not, or less so. Thus, choice of overall topics is obviously more ‘intentional’ than the detailed syntactic structure or intonation of a sentence. Selection of words falls in between – lexicalization is largely automatic given underlying mental models and the lexicon as a basis, but often specific words are chosen deliberately, and depending on genre and context quite well controlled, especially in written communication. There is no doubt that in an important political speech of a president or presidential candidate practically each word is chosen as a function of its ideologically and communicative presuppositions and implications. That is, when overall communicative control is strict, also ideological discourse expression will become more conscious. In some contexts, on the other hand, both discourse control and ideological control will be largely automatized.


Ideological (over-) interpretation


Discourse structures have many cognitive, interactional and social functions. None of them are exclusively ideological. For instance, passive sentences and nominalizations hide or background agency. This may have an ideological function in an otherwise ideologically biased text and context – for instance to mitigate our agency and responsibility for negative actions, thus implementing locally the overall ideological discourse strategy of positive self-presentation of ingroups.

However, we should be careful not to over-interpret discourse data. Often passive sentences and nominalizations are used when agents are unknown, when they have just been mentioned and should not be repeated, or when the current focus is on other participants – such as the victims of violent actions rather than on the actors. This means that such data should never be described in isolation, but in relation to the text (co-text) as a whole and in relation to the context – who is speaking to whom, when, and with what intention. Most importantly, also theoretically, is to realize that discourse is not just to express or reproduce ideologies. People do many other things with words at the same time.




The theory of ideological discourse processing outlined above has an explicit component that accounts for contextualization, defined in terms of subjective context models of participants. These dynamically updated models represent what of the communicative situation is ongoingly relevant for each participant at each moment of text or talk. We have seen that context models control discourse production and comprehension. We have also assumed that these context models may themselves be ideologically biased, e.g., when speakers represent and evaluate their interlocutors in terms of racist, sexist or other ideologies. Ideological context models usually lead to ideological discourses or discourse interpretations. We have also seen though that the expression of ideological allegiances may be (partly) controlled. In relation to the issue of intentionality discussed above, this means that we may have discourses that are produced in an ideological context but barely show traces or markers of such a context (including the ideologies of the speaker) – or at least no explicit expressions that are interpreted as ideological by many recipients. But other recipients may very well be able to understand ‘coded’ ideological expressions – expressions that for other recipients may remain deniable. It is therefore crucial for all ideological discourse analysis, to make explicit what exactly the social situation is and how it is being represented by the participants in their context models. If a recipient, based on previous experiences, defines a speaker as a male chauvinist, then much of what he says will be ‘heard’ as an expression of male chauvinism whether or not there are contextualization cues that warrant such an interpretation.

As external observers or analysts of such interactions this may posit the problem of (over-) interpretation discussed above: what to do if we cannot find overt expressions of an ideology? In my view, we should not so much be concerned with what analysts may or may not do, but rather with what participants actually do. So, if recipients interpret discourses as ideological on contextual grounds, even if there are no overt ideological markers, then we must account for such ‘over-interpretations’ in the theory. That is, the mental models recipients build when interpreting discourse may also be construed on the basis of inferences about ideological intentions of speakers as inferred from previous experiences, hearsay or other reliable information about a speaker. Indeed, this is the way we ‘hear’ speakers in the first place: based on our earlier, general ‘impressions’ (models) of that speaker. There is no doubt that this also may lead to prejudiced ideological interpretations of discourses – especially when individual speakers are judged not on the basis of earlier discourses or other social practices, but on the basis of their group membership. Obviously, the theory should also account for such ideological prejudices.


Ideology, knowledge and presupposition


The distinction between ‘mere’ ideology and ‘true’ knowledge has been part of the discussion about ideology since the invention of that notion by Destutt de Tracy more than 200 years ago. Although the distinction has many implications that cannot be discussed in this paper, a few remarks relevant for ideological discourse analysis should be made here.

First of all, I define knowledge as the beliefs certified and shared by a (knowledge) community, where certification takes place with the historically variable (epistemic) criteria or ‘methods’ of that community (e.g., observation, direct experience, reliable sources, inference, experiments and other ‘methods’). In other words, knowledge is not ‘justified true belief’, as the classical definition in epistemology has it, but accepted beliefs in a community. In other words, our definition is pragmatic, rather than semantic, and related to a consensus theory of knowledge restricted to specific knowledge communities. This makes knowledge relative and intersubjective, on the one hand, namely relative to a community and its members, but on the other hand ‘objective’, namely within the community, on the grounds of certification socioculturally accepted in that community.

Secondly, it should be repeated that -- contrary to some critical conceptions of knowledge -- our theory of ideology does not imply that all knowledge is ideological. On the contrary, it is the general knowledge shared by a community (the Common Ground) that forms the foundation of all social representations of a community, and hence also of the ideologies of the various social groups in such a community. Ideological struggle and debate presupposes such general knowledge. For instance, to have different ideological positions about immigration, members need to know what immigration is in the first place.

However, as defined above, ideologies are the basis of the social representations of groups, including their knowledge. That is, group knowledge may well be ideological based. Thus, if racists claim to ‘know’ that blacks are intellectually inferior, then such ‘knowledge’ is obviously ideological: it may be knowledge for them, but it is mere, prejudiced belief, for others, today – and it might have been a consensual belief, and hence knowledge, two centuries ago.

Note though that such is also the case for those beliefs many people may now hold to be true, such as feminist knowledge about gender inequalities, as long as there are groups who believe that this knowledge is ‘merely feminist’ belief. Only when across all groups in a culture some belief is shared and presupposed in all discourse, then we may speak of general, cultural knowledge – until there are social groups (scholars, social movements) who challenge such consensual beliefs.

In ideological discourse analysis, then, we may want to examine texts with propositions that are asserted or presupposed (to be true, to be shared, or to be taken for granted), and hence presented as knowledge. However, context analysis in that case may reveal that speakers assert or presuppose such knowledge as group members, in which case the knowledge may well be ideological, and judged to be ‘mere belief’ or prejudice by members of other groups. Often, group members know that the beliefs they hold are not (yet, anymore) shared by other groups, and in that case they will assert and explicitly (try) to certify such beliefs as knowledge in communication with members of other groups. Incidentally, this is not only the case for prejudiced beliefs, religious beliefs, superstitions, but also for many scientific or other ‘new’ beliefs.

Hence, a good empirical test to distinguish general cultural knowledge from ‘group knowledge’ and ideologies, is presupposition in all public discourses of all groups in a culture and at a given moment. Of course, it may turn out that later such knowledge may be ideologically based anyway, but in that case it is no longer generally presupposed.

From this discussion it must be concluded that at least under the definition of knowledge and ideology presented here, it is not true to assume that all knowledge is ideological. Again, such beliefs may be ideological for some observers or critical analysts, that is, from an outside point of view of another epistemic community, but if they are generally accepted by the members of a community, such beliefs by definition are called, presupposed, used and counted as knowledge.


The analysis of ideology as ‘critical’ analysis


The theory of ideology and discourse as presented here, in which ideologies are defined in general terms, and not merely as dominant or as legitimation for inequality, may seem to have lost its ‘critical’ edge. Nothing is less true, however. On the contrary, it offers a much more explicit framework for critical study. Ideologies, as defined, may be used as a foundation for discourses and social practices both by dominant as well as by dominated groups, both for oppression as well as for resistance. This is why we may speak both of racist as well as of antiracist ideologies, of sexist and of feminist ideologies, of neoliberal and of socialist ideologies. In other words, we do not follow the old distinction of Mannheim (1936) between ideologies and utopias. The basic structures and functions of ideologies are the same in these cases, namely to self-represent the group and the membership and identification of its members, to organize their social practices or struggle, and to promote the interests of the group and its members with respect to other groups.

It follows that it is theoretically inconsistent and unproductive to reserve the notion of ideology and ideological critique only for dominant ideologies. Indeed, it is a characteristic of ideological discourse to attribute ideologies only to our opponents and ‘truth’ to ourselves. However, in the same way as we need a general notion of power, which may also include forms of resistance or counter-power, we need a general notion of ideology. Critical discourse analysis may then very well focus especially on the dominant groups and their ideologies.

The more explicit theory of ideology and its relations to discourse presented here also allows a much more detailed analysis of the processes at work in the acquisition, uses and changes of ideologies in discourse. Adequate CDA requires precise theoretical instruments, not vague, traditional notions (such as ‘false consciousness’). Its critical appropriateness depends on the precision of its analysis, the selection of its objects of analysis and critique, its aims, and the political and ethical position of the scholars who engage in it.


Ideology vs. Discourse


There are directions of research, such as discursive psychology, that reduce mental structures (such as attitudes) to discourse structures, e.g., with the argument that discourses are observable and social, whereas minds are not (see, e.g., Edwards & Potter, 1992). This implies that in such a perspective ideologies cannot be defined in terms of some kind of mental representation, but only in terms of the discourse structures than express or enact them. Also in other work on ideology, there is a tendency to identify ideology with discourse or other social practices.

Although the theory presented here obviously accounts for ideological discourse, it does not reduce ideologies to discourse – nor does it do so with other cognitive notions such as knowledge, opinions, attitudes or norms and values. These cognitive notions are of another theoretical order – they are used in a theory of mind – than a theory of discourse, which is a theory of social interaction.

One of the arguments against a reduction of ideologies or other socially shared beliefs can be expressed, enacted or produced not only by discourse, but also by other social practices, such as domination, empowerment, discrimination, oppression, exclusion, resistance, opposition, dissent, and so on. Furthermore, people may ‘have’ ideologies without actually expressing them or acting upon them all the time. Indeed, people may sometimes NOT act (not work, not vote, etc.) when expected to do so, out of ideological grounds – such as going on strike. Reduction of ideology to discourse does not allow an ideological analysis of other social practices. It is true that ideologies are generally explicitly expressed, acquired and reproduced by discourse, but not exclusively so. Many ideologies, such as sexism, are being acquired and practiced by imitation of the actions of others, not only by discursive explanations or teachers.

The argument that seems to favor an exclusively discursive approach to ideology, namely that discourses are more ‘social and ‘observable’ than minds, is a fallacy. Firstly, we have emphasized that ideologies are shared beliefs of groups, that is, they are both socially and mentally shared. Just like languages, thus, ideologies are social, also because their functions are social. Secondly, discourses are usually defined as abstract structures of form, meaning and interaction – and not (merely) as the sound waves or graphical/visual/electronic marks or muscle movements that realize them physically. Lexical items, syntactic structures, meanings and interactions cannot be ‘observed’ directly, but are interpreted, and hence either abstract object of linguistic theory, or mental constructs of language users. In other words, unless one denies the cognitive nature of meaning and understanding, any empirical definition of discourse implies cognitive notions of some kind.

In sum, to reject cognitive notions such as knowledge, beliefs, opinions and ideologies, among others on grounds of ‘observability’ is a behaviorist (in terms of contemporary proponents of such views, perhaps better called ‘interactionist’) fallacy. A sound theory of ideological discourse explicitly relates ideologies and discourses, but as different kinds of theoretical or empirical objects.


The discursive acquisition of ideologies


Although ideologies may be acquired by group socialization in many social practices, e.g., at home, at school, at the job or in the bar, the primary sources and medium of ideological ‘learning’ are talk and text. That is, ideologies usually are not merely acquired by merely imitating the actions of other group members. Such processes of observation and participation usually are usually accompanied by reasons and explanations (e.g., ‘We do not admit X, because people like us do not mingle with people like X’), which may imply (tacit or explicit) self-attributions of superiority and other-attributions of inferiority or ‘difference’.

Thus, from childhood onwards kids gradually learn some basic elements of ideologies of gender, ethnicity, class, religion, politics, and so on, by growing up and participating and education in the respective social groups of which they are members. Each of these groups has more or less explicit, more or less formalized and institutionalized ways to teach ideologies to new members, e.g., special meetings, schooling, catechism, textbooks, propaganda leaflets, sermons, and so on, that is, various kinds of discourse.

Whereas some discourse genres may apply ideologies more or less implicitly, didactic ideological discourse is much more explicit, namely by formulating the general contents of the ideological schema of the group: who we are (where we come from, what we look like, who can be a member of our group, etc); what we do and stand for, what our norms and values are, who are friends and enemies are, and what our power resources are, etc.). It provides reasons and arguments in terms of general norms and values, and in view of the interests of the group and its members; about what is good and bad, just or unjust. It gives examples, and provides images of venerated gods and leaders or other exemplary people. It tells stories about heroes and villains.

Ideological discourse thus gradually develops the general ideological framework of the group. This is a slow process. Young kids as yet have barely ideological notions. Most ideologies are chosen or learned more explicitly as from adolescence or early adulthood – typically when people are students, and need to give a broader meaning to their actions and aims as well to the world in which they live.

This is not the place to detail the process of ideological learning, but if we assume that such acquisition largely takes place through discourse, then we may surmise that this happens following the reverse process of that of ideological discourse production sketched above. People interpret discourses as mental models. Ideological markers such as the one listed in Table 1 orient recipients in the way ingroup and outgroup actors are represented in such mental models. Both the text as well as the context of repeated communication of the same kind may lead to the generalization and abstraction of mental models to more general, group-based attitudes. And finally, several attitudes in one domain of society may be subsumed by more general ideological propositions.

That is, ideologies may be learned both bottom up, namely by generalizing mental models (experiences, concrete stories) to socially shared and normalized mental representations, or top down by explicit ideological instruction by ideologies of various kinds (leaders, teachers, priests, etc.). Some ideologies will tend to be learned more explicitly, and hence top down, such as those of scientific methodology, religion, and political ideologies (pacifism, etc.). Others are much more implicit and integrated in the practices of everyday life, such as gender, race and class ideologies. They tend to become explicit in cases of conflict, struggle or resistance. It will be one of the major empirical tasks of ideological discourse analysis to systematically examine the structures and strategies of these different types of ideological discourses and their role in the acquisition and reproduction of ideologies by the members of groups, and hence by groups as a whole. Many of these social and cognitive acquisition processes can be studied by systematic analysis of such didactic texts and contexts.




Let us finally examine an example. Consider the following leader published in the Sunday Telegraph on August 8, 2004:


The Sunday Telegraph

Pinko paper

Last week the left-wing Institute for Public Policy Research, with staggering predictability, produced yet another report attempting to persuade us that the Government is not doing enough to narrow the difference in income between "rich and poor".

The more thoughtful commentators on the Left have begun to see what a blind alley this is. As The Independent's John Rentoul wrote: "Trends in income and wealth distribution in free economies are mostly beyond the power of democratic governments to influence . . . [the Treasury] should tell the IPPR to shove it."

But one newspaper in Britain is still resolutely redistributionist, publishing a leading article praising the IPPR under the subtitle "Rise in British wealth inequality marks a worrying trend". No, it was not the Morning Star. It was the Financial Times.

In truth, this was less surprising than it might seem. The FT has been getting pinker and pinker in its political views over a number of years: after all, it endorsed Neil Kinnock in 1992.

Whether this is what the FT's healthily avaricious readers expect to find is another matter. Its total UK circulation has fallen by a third over the past four years, and its full-paid UK circulation is now barely 100,000. The FT's political stance has helped earn its boss, Marjorie Scardino, a damehood. We somehow doubt that is a consolation to the company's shareholders.

For the author, the readers and the analyst alike, the author of this editorial is an editor of a conservative British newspaper, and as such a member of an ideological group. With this information in their context model this editorial is being construed, interpreted and analyzed as an ideological text.

The overall topic of the text is a critique of an IPPR report on income differences in the UK and of the Financial Times for supporting this report. That is, also the overall topic organizing this editorial is ideological, since the differences between rich and poor is an issue about which the Left and the Right have different attitudes, as also shown in this editorial.

The headline, expressing part of the main topic, summarizes its critique of the FT by using the adjective and pun ‘pinko’, referring implicitly both to the characteristic color of the pages of the FT, on the one hand, and to the political meaning of being ‘pink’, namely as being influenced by the Left, associated with the color red, on the other hand. By being accused of ‘leftist’ opinions, the FT is thus categorized as a political opponent of the Telegraph, that is, as part of the outgroup, as Them. The alliteration ‘pinko paper’ rhetorically emphasizes the negative characteristic attributed to the FT. Note also that headlines are first and on top, and usually in bigger type than the text – that is, there are also visual markers that emphasize the importance of the global topic of the text.

The body of the editorial provides a description of one of Them, namely the Institute of Public Policy Research (IPPR), characterized primarily as ‘left-wing’, and hence as an ideological opponent, and as being close to the current Labour government. The political implicature of such a description is that if one of Them says or writes something bad, it also applies to the others of Them¸ that is, to the Left in general, and the Labour government of Tony Blair in particular. According to the overall ideological strategy outlined above, the Telegraph – itself ‘with staggering predictability’ -- characterizes the report of the IPPR in very negative terms, namely as being ‘predictable’, thereby implying that the Left always says the same things, using the rhetorical hyperbole ‘staggering’ to emphasize this negative characteristic. The use of the adjective ‘another’ has the same ideological implications as the negative qualification of ‘predictable’. The quotation marks around “rich and poor” not only mark that this is probably the expression of the IPPR report, but at the same time signals the ideological distance of the Telegraph with respect to such a description of class differences in the UK.

The second paragraph makes use of a well-known persuasive move, namely to cite with approval someone of the ideological opponents as an argument for one’s position, a well-known argumentation fallacy, implying that ‘if one of them says so himself, then it must be true’.  This opponent is not just introduced and described as a journalist of the Independent, but also positively described as ‘more thoughtful’, thereby ideologically implying that others on the Left are ‘less thoughtful’, which is a euphemism for ‘stupid’. In other words, the ideological proposition underlying these moves and descriptions is ‘We are intelligent, and They are stupid’, an appraisal that is also expressed in the well-known derogatory alliterative description of the ‘Loony Left’.

The quote from the Independent is a straightforward formulation of a neoliberal attitude about income redistribution, as the value-based characterization ‘free economies’ also suggests; also the expression ‘beyond the power of democratic governments’ politically implies a norm that governments should not interfere in the economy – a basic ideological proposition of a neoliberal ideology. By approvingly citing this fragment, the Telegraph implies that it agrees with this proposition, and hence self-categorizes itself as a mouthpiece of neoliberal ideologies. The metaphor ‘blind alley’ is an explicit negative ideological rejection of income redistribution, in which political policies can be described in types of roads – those which have the future (are open), and those that have not (are closed).

The rest of the editorial then selects the main ideological opponent of the editorial. Generally, the press does not like to criticize the press. However, in this case, the apparent alignment with the Left of an establishment newspaper for the business community, and hence as a potential defender of conservative and neoliberal views, is unacceptable treason for the Telegraph and hence severely criticized. The rhetorical negation “No it was not the Morning Star”, emphasizes that this ‘leftist’ position of the FT is not expected. The description of the FT as “resolutely redistributionist” is a hyperbolic, ideological characterization of one of the propositions of leftist or socialist ideologies, and hence ideologically disqualified and (implicitly) rejected and criticized by the Telegraph.

In line 16, the reference to “surprise” confirms the implication inferred earlier, namely that one would not expect such a ‘leftist’ position of the FT. This is followed by a more general negative description of the FT as becoming “pinker and pinker” over the years, in other words, more and more leftist – which in the eyes of the Telegraph is as such a disqualification. These words also provide the local details of the overall negative reference to the FT in the headline (“Pinko Paper”). We see that there is not a further argument with the FT support of the IPPR report. It suffices that an influential newspaper supports a leftist report of a leftist organization in order to be disqualified. As an argument for the general statement that ‘explains’ this position of the FT, namely that the FT has become a leftist newspaper, the Telegraph cites a case of ‘historical evidence’: that the FT supported, well-known leftist politician Neil Kinnock. By standard argumentation criteria using one example to support a general evaluation would usually be qualified as a fallacy.

The last paragraph finally attributes the criticized, ‘leftist’ position of the FT to bad management, more specifically of its boss Marjorie Scandino, and supported by the argument that the UK circulation “has fallen by a third”, and the further argument that the shareholders won’t like that. In other words, we find ‘sound’ business arguments against the untenable ideological position attributed to the FT: the only arguments the FT – and its shareholders – will understand. Note the seemingly overcomplete reference to ‘full paid UK circulation’, which seems to suggest that the accusation may not hold when the complete circulation (also abroad) is taken into account – a well-known fallacy of manipulating statistics. Apart from aligning itself positively (“healthily avaricious”) with the shareholders of the FT, consistently so with its neoliberal ideology, the Telegraph finally personally attacks Scandino by attributing her damehood to her political position – a personal attack known as the well-known ad hominem fallacy. The political implicature, namely that this is of course a decision of a Labour government, is obvious. By ideologically attacking and disqualifying the FT and its boss, the Telegraph at the same time is aiming at its main opponent: Labour and Blair.

With these few, succinctly analyzed, examples from an editorial, we see how an underlying conservative, neoliberal ideology gets articulated in routine texts in the press. One of the typical ideological propositions of the Left, namely income redistribution, is being criticized, and so are those who defend it – the IPPR, and especially the major newspaper –surprisingly-- supporting it: the Financial Times. The strategies used to disqualify the ideological opponents are all following the schema outlined above – overall negative topic in headline, rhetorical puns (‘pinko’), hyperboles (‘staggering’, ‘resolutely’), metaphors (‘blind alley’), fallacies (authorities, single case evidence, ad hominem, etc.), selecting and praising ideological opponents who agree with one’s position, and a general negative description of the ideological opponent and its positions (‘predictable’, implied ‘stupid’, ‘pinker and pinker’), among other moves.

Obviously, this one text only realizes a tiny fragment of the vast amount of possible ideological markings in discourse. However, this example makes clear how underlying ideologies control large part of the structures of this text, such as its headlines and main topic, its adjectives, its quotations, its person and institution descriptions, and much of its rhetoric. In a brief editorial like this, the proper ideological debate is barely presented – it is only briefly stated and implied that the Telegraph rejects income redistribution and State interference in the market in general, that is any ‘leftist’ ideological position, thereby characterizing and confirming itself as conservative in the UK context, and neoliberal more generally.  The main strategy thus is not to explain why income redistribution is a “blind alley”, but rather to attack influential persons and institutions – and finally the government – that may support such a position.

Note finally that this text is not just any odd text. It is an editorial of a major ‘quality’ newspaper in a major country, possibly read by hundreds of thousands of people. Even when these readers know very little of the details and possible benefits of income redistribution, this editorial will help them make up or reinforce their minds by taking strong position against such a (loony) ‘leftist’ idea. At the same time, the attack on the Financial Times and its boss, means that the ideological struggle is not one of sophisticated ideological debate about the pros and cons of income redistribution, but rather a populist derogation of opponents by all the discursive tricks in the rhetorical bag – such as ad hominem attacks and general negative description of opponents.

Our analysis also shows how ideologies are institutionally co-produced and reproduced by powerful (business) institutions such as newspapers. Their editorials directly express dominant editorial views, which in turn are faithful mouthpieces of the owners. That is, ideological discourses are collective discourses of groups, and in many indirect and subtle ways, reflect the ideological positions of their organizations and their interests.

That such linkages between business, newspaper contents and ideological positions must be analyzed with care may be shown by the very fact that the Telegraph criticizes, namely that other newspapers, such as the Financial Times, sometimes may defend ideologically ‘inconsistent’ policies such as wealth distribution. In the way ideologies may be complex, and combine in sometimes inconsistent clusters, also ideological discourse is not always a direct, coherent and transparent manifestation of underlying ideologies. As suggested before – it all depends on context. Ideological control is not deterministic, but strategic.



Against the background of a multidisciplinary theory of ideology, this paper summarizes some of the relations between ideologies and discourse. Defined as socially shared representations of groups, ideologies are the foundations of group attitudes and other beliefs, and thus also control the ‘biased’ personal mental models that underlie the production of ideological discourse. This theory not only accounts for the ways ideological discourses are produced and understood, but also how ideologies themselves are discursively reproduced by groups and acquired by their members. It is stressed that ideologies are not only expressed by discourse – and hence should not be reduced to discourse either – but also may be expressed and enacted by other social practices. Ideological acquisition, legitimation and in general ideological accounts, however, are usually discursive.

It is also shown how ideologies relate to knowledge. Thus the assumption that all knowledge is ideological is rejected as too strict, because general cultural knowledge is presupposed and hence accepted by all ideologically different groups. It is also argued that a general theory of ideology as presented here does not lose its critical edge – on the contrary, a more explicit theory of ideology is better equipped to critically examine ideologies and their discourse practices.

When ideologies are mapped onto discourse, they typically become expressed in terms of their own underlying structures, such as the polarization between positive ingroup description and negative outgroup description. This may take place not only explicitly by propositional means (topics, meanings, etc.), but also by many other discursive moves that emphasize or de-emphasize Our/Their Good/Bad Things, such as headlines and position, sound structures and visuals, lexicalization, syntactic structure, semantic moves such as disclaimers, and a host of rhetorical figures and argumentation moves. Thus, at all levels of text and talk we may thus witness the influence of the ideological ‘bias’ of underlying mental models and social representations based on ideologies. It is however recalled that not all discourse structures are ideologically controlled, and that no discourse structure only has ideological functions. All depends on the context, defined here as subjective mental models (which may also themselves be ideological) representing the relevant properties of communicative situations.

By way of final illustration, an editorial of the Sunday Telegraph (on income redistribution in the UK) is briefly analyzed, and it is shown how, indeed, the underlying conservative, neoliberal ideology of this newspaper controls all levels and many properties of the editorial. At the same time, such an editorial shows how conservative and neoliberal ideologies are being reproduced in society – e.g. by the use of personal attacks, discrediting of opponents (‘Loony Left’, ‘Pinko Paper’), among many other moves – by powerful institutions such as newspapers.



I am indebted to Michael Freeden for inviting me to this interesting symposium as well as for comments on the first version of this paper. I also would like to thank Sophie Duchesne for very useful comments on the sociological foundations of the notion of ideology, as well as Matthew Humphrey, David Kettler, Bo Strath and Ernesto Laclau for other questions and suggestions.





Anderson, J. R. (1980). Cognitive psychology and its implications. San Francisco: W. H. Freeman.


Billig, M. (1982). Ideology and social psychology: Extremism, moderation, and contradiction. New York: St. Martin's Press.


Brand, M. (1984). Intending and acting. Toward a naturalized action theory. Cambridge, Mass.: MIT Press.


Caldas-Coulthard, C. R., & Coulthard, M. (Eds.). (1996). Texts and Practices: Readings in Critical Discourse Analysis. London, England: Routledge.


Converse, P. E. (1964). The nature of belief systems in mass publics. In D.E. Apter (Ed.), Ideology and discontent. New York: Free Press.


Curran, T. F. (2003). Soldiers of peace. Civil War pacifism and the postwar radical peace movement. New York: Fordham University Press.


Duranti, A. (2001). Intentionality. In A. Duranti (Ed.), Key Terms in Language and Culture (pp. 129–131). Malden, MA: Blackwell.


Eagleton, T. (1991). Ideology. An introduction. London: Verso Eds.


Fairclough, N. L. (1995). Critical Discourse Analysis: The Critical Study of Language. Harlow, England: Longman.


Feldman, S. (1988). Structure and Consistency in Public Opinion: The Role of Core Beliefs and Values. American Journal of Political Science, 32, 416-440.


Fiske, S. T., & Kinder, D. R. (1981). Involvement, expertise, and schema use: Evidence from political cognition. In N. Cantor & J. F. Kihlstrom (Eds.), Personality, cognition and social interaction. Hillsdale, NJ: Erlbaum.


Fowler, R. (1991). Language in the news: Discourse and ideology in the British press. London New York: Routledge.


Freeden, M. (1996). Ideologies and political theory. A conceptual approach. Oxford: Clarendon Press.


Kress, G., & Hodge, B. (1979). Language and ideology. London: Routledge & Kegan Paul.


Larraín, J. (1979). The concept of ideology. Athens: University of Georgia Press.


Lau, R. R., & Sears, D. O. (Eds.). (1986). Political cognition. Hillsdale, NJ: Erlbaum.


Mannheim, K. (1936). Ideology and utopia: An introduction to the sociology of knowledge. London New York: K. Paul, Trench, Trubner & Co. Harcourt, Brace and company.


Oberschall, A. (1993). Social movements. Ideologies, interests, and identities. New Brunswick, NJ: Transaction.


Schank, R. C., & Abelson, R. P. (1977). Scripts, plans, goals, and understanding: An inquiry into human knowledge structures. Hillsdale, N.J.: L. Erlbaum


Thompson, J. B. (1984). Studies in the theory of ideology. Berkeley: University of California Press.


Vallacher, R. R., & Wegner, D. M. (1989). Levels of Personal Agency: Individual Variation in Action Identification. Journal of Personality and Social Psychology, 57, 660-671.


Van der Pligt, J. (1992). Nuclear energy and the public. Oxford: Blackwell.


Van Dijk, T. A. (1984). Prejudice in discourse an analysis of ethnic prejudice in cognition and conversation. Amsterdam Philadelphia: J. Benjamins Co..


Van Dijk, T. A. (1987). Communicating racism: Ethnic prejudice in thought and talk. Newbury Park, CA: Sage Publications, Inc.


Van Dijk, T. A. (1993). Elite discourse and racism. Newbury Park, CA, USA: Sage Publications.


Van Dijk, T. A. (1993). Principles of Critical Discourse Analysis. Discourse and Society 4(2), 249-83.


Van Dijk, T. A. (1995). Discourse Semantics and Ideology. Discourse & Society, 6(2), 243-289.


van Dijk, T. (1998a). Ideology: A Multidisciplinary Approach. London: Sage.


Van Dijk, T. A. (1998b). Opinions and ideologies in the press. In A. Bell & P. Garrett, Approaches to media discourse. Oxford: Blackwell.


Van Dijk, T. A. (1999). Context models in discourse processing. In: van Oostendorp, Herre, &  Goldman, Susan R. (Eds.), The construction of mental representations during reading. (pp. 123-148). Mahwah, NJ, USA: Lawrence Erlbaum Associates.


Van Dijk, T. A. (2001). Discourse, ideology and Context. Folia Linguistica XXXV/1-2, 11-40.


Van Dijk, T. A. (2004). Discourse, knowledge and ideology. In Martin Pütz, JoAnne Neff & Teun A. van Dijk (Eds.), Communicating Ideologies. Multidisciplinary Perspectives on Language, Discourse and Social Practice. (pp. 5-38). Frankfurt/Main: Peter Lang.


Van Oostendorp, H., & Goldman, S. R. (Eds.). (1999). The construction of mental representations during reading. Mahwah, NJ: Lawrence Erlbaum.


Wodak, R. (1989). Language, Power, & Ideology: Studies in Political Discourse. Amsterdam: Walter Benjamins.


Wodak, R., & Meyer, M. (Eds.). (2001). Methods of critical discourse analysis. London: Sage


Zanna, M. P., Klosson, E. C., & Darley, J. M. (1976). How television news viewers deal with facts that contradict their beliefs: A consistency and attribution analysis. Journal of Applied Social Psychology 6, 159-176.