Paper for the 4th International Symposium of Critical Discourse Analysis, Language, Social Life and Critical Thought, Athens, 14-16 December, 1995.

Second Draft, March 1996 

Opinions and Ideologies in Editorials

Teun A. van Dijk 

vandijk?discourse-in-society.org (replace ? by @)

 

1. Introduction 

Many phenomena of everyday life tend to be ignored by scholars. Thus, historically, among many types of discourse, everyday conversations were studied systematically much later than less mundane genres such as poetry, drama and novels. The same was true for such pervasive everyday texts as news reports in the media which began to be studied systematically and from a discourse analytical point of view during the last decade (see, e.g., Fairclough, 1995; Fowler, 1991; van Dijk, 1985, 1988a, 1988b, 1991).

Editorials are no exception. Each day we find them, usually at the same page and at the same location, in our daily newspapers. For those people who read them, they help to make up their mind about the events of the world, even if often by critical opposition against them. Given this prominent function of editorials in the expression and construction of public opinion, one would expect a vast scholarly literature on them. Nothing is less true: There are virtually no book-length studies, and rather few substantial articles, on the structures, strategies and social functions of editorials. They are taken for granted as so many of the ordinary types of text and talk in society and culture (but for some discourse analytical studies, see, e.g., Bolivar, 1994; Bonnafous & Fiala, 1984; Hackett & Zhao, 1994; Love & Morrison, 1989; Tirkkonen-Condit, 1987; van Dijk, 1989, 1992, 1995a).

Most work on editorials is practical and anecdotal, and written by (former) journalists (Fischer & Fischer, 1990; Rystrom, 1994). Scholarly studies are often journalistic, and about how the press dealt with a specific historical issue (mostly in the USA), such as the civil rights movement (Bagdikian, 1968), foreign policy (Chang, 1989; Myers, 1982a, 1982b), bias and the editorial endorsement of presidential candidates in the USA (Fedler, Smith, & Counts, 1985; Merron & Gaddy, 1986; St. Dizier, 1985, 1986); political leadership (Sinclair, 1982). There is also some psychological and other work on the influence of editorials on the readers (see, e.g., Alvarado, 1990; Cacioppo, Petty, & Sidera, 1982; Fasold, Yamada, Robinson, & Barish, 1990; Gruner, 1989; Krueger & Fox, 1991).

Against the backdrop of this lack of explicit theorizing about editorials, this paper can only contribute to our insight of one prominent and characteristic feature of editorials, viz., the formulation of opinions and the expression of ideologies. After a brief account of the notions of 'opinion' and 'ideology', we shall therefore focus on some of these discursive properties of the manifestation of evaluative beliefs of newspaper editors.

The approach of this paper fits into a broader framework in which the relations between discourse and socially shared mental representations are being investigated, with special attention to the discursive manifestation of ideologies. In the present stage of this long-term project, especially the cognitive and discursive aspects of these mental representations are being investigated (van Dijk, 1995a).

It is however obvious that opinions and ideologies also have prominent social, political and cultural functions. Thus, in our case opinions and ideologies are expressed in the socially and culturally relevant genre of newspaper editorials. They play a role in the formation and change of public opinion, in setting the political agenda, and in influencing social debate, decision making and other forms of social and political action. When expressed in editorials, opinions and ideologies are being produced by journalists and other writers, who both as professionals and as other social group members (e.g., men, whites, conservatives, etc.) exhibit their shared social representations, and participate in the complex processes of newspaper production and reception as well as in intergroup interaction and institutional reproduction.

To close the discourse-cognition-society circle (or rather triangle), we assume however that these various sociocultural functions, structures and processes are in turn based on socially shared mental representations: In the same way as the mind is part of society, also society is again 'part of' the mind. Social actors manage their social actions and act within social contexts as a function of their personal as well as socially shared interpretations or representations of their social environment. Mind and society mutually presuppose each other, and theories need to explain these interdependencies, without however reducing the one to the other.

2. Opinions 

The concept of 'opinion' belongs to the set of those concepts that are commonly used in everyday life, but whose theoretical analysis is far from straightforward. Whether in philosophy, psychology or the social sciences, opinions have defied many scholars, and we are therefore able to summarize only some of their characteristics, viz., those we need in order to understand their 'linguistic' manifestations in discourse.

It is not surprising that most work on opinion is being done in social psychology, the study of mass communication and political science, viz., on what is generally known as 'public opinion', for which there is even a special journal, viz., the Public Opinion Quarterly , after its first influential use by well-known journalist Walter Lippmann (1922). In most of this work, the notion of opinion is not theoretically analyzed, although Lippman already had a rather sophisticated cognitive approach, e.g. in terms of prejudices defined as schematic 'pictures in our heads'.

Opinions and attitudes are usually collapsed in such studies, as are personal and social opinions (see the critique of this reduction of social attitudes to individual ones in Jaspars & Fraser, 1984). Much social psychological work is on 'attitude change' as a function of some communicator or message 'variable'. Virtually no studies deal with the detailed mental representations of opinions. That is, public opinion in most of this work barely goes beyond the level of practical, commonsense definitions (for some of the more detailed studies of (public) opinion, see e.g., Bennett, 1980; Bogart, 1985; Crespi, 1989; Eagly & Chaiken, 1993; Erikson, Luttbeg, & Tedin, 1988; Fraser & Gaskell, 1990; Holloway & George, 1986; Kinder & Sears, 1985; Milburn, 1991; Noelle-Neumann, 1984; Oskamp, 1991; Price, 1992; Taylor-Gooby, 1985; Zaller, 1992).

Our conception of opinions is multidisciplinary: As such, we define them as a special form of mental representations; their acquisition, usages and functions are social, and their expression often discursive. That is, any satisfactory theory of opinions should, in our opinion, be located in the triangle that relates cognition , society and discourse . Any reduction of one angle of the triangle to any other will result, not only geometrically, into a simple direct line or to one point, that is, a simplistic theoretical representation. Thus, we need the triple approach because as to their nature (location, structure) opinions are mental, whereas much of their acquisition, uses and functions are social, and their expression and reproduction often (though not always) discursive. That is, opinions are mostly formed within contexts of social interaction in general, and through text and talk in particular.

Opinions as mental representations 

A first theoretical decision we take is to locate opinions where also our common sense and everyday language use locate them, viz., in the mind . Thus, 'having an opinion about X' will here first be taken to mean something like 'having some mental representation of X.'

Given our knowledge of contemporary cognitive psychology, this may seem quite straightforward, even if for some strange reason most cognitive psychologists leave the study of opinions to their colleagues in social psychology. However, some scholars prefer to reduce opinions to their manifestations in text or talk. This would imply that people do not 'have' opinions before they start talking about them, and that opinions only are context-bound and constructed as part of discourse itself (see especially the work in discursive and rhetorical psychology, e.g., Billig, 1991; Harre & Stearns, 1995; Potter & Wetherell, 1987, 1988).

In our view, this reduction of mental representations to their discursive expression or construction is not only counter-intuitive, but also does not account for the facts. Indeed, it does not explain

- that people may have opinions without expressing them (in discourse or otherwise);

- that they may have an opinion A but express an opinion B for well-known contextual reasons of politeness, face-keeping, impression management or social norms;

- that they may give many different expressions to the 'same' opinion;

- that they may have the 'same' opinion on different occasions;

- and that they may share the 'same' opinion with others.

On a strict conception of opinions as discourse, thus, we would have as many opinions as we have opinion-discourses, and no way to account for the cross-situational or socially shared identity of opinions. Indeed, since all discourse is contextual, the very notion of shared, social opinions would be impossible, or at most reduced to the ways groups of people talk about an issue (in many different ways).

For these and other reasons, therefore, we analytically distinguish between opinions as mental representations, on the one hand, and their discursive manifestations (acquisition, uses, etc.), on the other hand.

Also, we distinguish between an opinion and the many ways it may actually be used, applied or expressed in various social contexts. That is, opinions, and especially socially shared ones, are relatively context-free, although of course the way they are being used and expressed, especially in talk, are contextually variable. In this way, we account both for the relative stability of social opinions and attitudes, as well as for their contextual and individual uses and variations. Below we first deal with the mental representations of opinions, and then focus on their social uses in editorials.

So far we have made no distinction between 'mental' and 'cognitive', as if all processes and representations of the mind are cognitive. If we assume that the mind may also feature non-cognitive processes or representations, such as emotion or affect, then such a reduction of the mental to the cognitive is not-warranted. It is undeniably the case that opinions may be associated with emotions. However, although opinions will be defined as evaluative beliefs, this does not imply that evaluations are, as such, emotive or affective. Thus journalists or readers may feel angry when writing or reading an editorial, but such a 'state of mind' (or rather: state of the body) is not part of the opinions expressed in the editorial: People may express or form an opinion without being 'emotional'.

In the rest of this paper we shall further ignore the relations between opinions and emotions, and only take a cognitive approach. One reason for this decision is that emotions are usually defined as properties of (the arousal of) individual persons, whereas we are interested especially in opinions as forms of social cognition. Thus, when we say that opinions are mental representations this means that they are dealt with as cognitive representations, but without excluding the role of emotions in their formation, change and use (for discussion, see Frijda, 1987; Ortony, Clore & Collins, 1988).

Opinions as evaluative beliefs 

Cutting a long and complex philosophical and psychological discussion short, we thus assume that opinions are a type of belief , as also knowledge, attitudes, ideologies and other mental representations consist of beliefs. Although widely used in everyday life, we don't know exactly what beliefs are, but shall take them as the basic building blocks of mental representations, as well as the units of thought and judgment. Both opinions and beliefs are typically said to be 'about' something, viz., some aspect of real or imagined states of affairs, something that is true or false, or something about which we like or dislike.

Although a typical mental notion, also the notion of 'belief' is not commonly theorized very deeply in cognitive, social and political psychology (see, however, e.g., Abelson, 1973; Bar-Tal, 1990; Bar-Tal & Kruglanski, 1988; Iyengar & McGuire, 1993; Fraser & Gaskell, 1990; Kruglanski, 1989; Wegman, 1981). In many studies, beliefs, belief systems, opinions and attitudes are not carefully defined and distinguished. The focus is often on what people have beliefs about . The precise nature, the structures or internal representations of beliefs have received scant attention, as is the case for opinions and attitudes.

In philosophy, epistemology, and philosophical logic, both knowledge and beliefs tend to be abstractly associated with propositions, and hence with what is defined (or asserted) as true or false, and analyzed in terms of a predicate and a number of arguments, possibly combined with a doxastic or other modal operator ('It is believed that...' or 'It is necessary that...'). The precise mental nature of such knowledge and beliefs is usually ignored here (Cohen, 1992; Graham, 1993; Hintikka, 1962; Kornblith, 1994; Lehrer, 1990; Moravcsik, 1990; Searle, 1995).

At this point, let us begin by the common sense notion used in everyday language and thought, in which a distinction is made between knowledge and opinions . Following suggestions from contemporary epistemology, knowledge here consists of true justified beliefs, viz., beliefs that satisfy socioculturally variable but shared truth criteria (Kornblith, 1994). In other words, beliefs constitute knowledge if they are true and based on sufficient and acceptable evidence. Depending on the kind of knowledge we are talking about, this may be everyday, commonsense evidence, or scientific proof.

In everyday discourse, someone's beliefs are usually called 'knowledge' only if the speaker also thinks these beliefs are true. In other words, the notion of knowledge is relative to socioculturally shared commonsense or scientific truth criteria or verification instances. In other --more pragmatic-- words, knowledge is being ascribed or attributed to others by those who (e.g. as observers or speakers) evaluate the beliefs of others relative to their own beliefs. As we shall see later, these conditions involve the notion of social or shared cognition , and hence a social and cultural dimension, in the theory of knowledge and beliefs.

Opinions , on the other hand, are generally defined as those beliefs that do not pass the test criteria of knowledge. That is, they may be beliefs some people have but which (for all we know) are false , or more generally beliefs about which we or others may disagree . When distinguished from knowledge, such opinions may also simply be called 'beliefs'. However, we shall use 'beliefs' only to generally refer to any kind of (true, false or disputed) belief.

Next, we shall further restrict the notion of opinion to those beliefs that have an evaluative dimension, viz., those beliefs that imply that something is good or bad, right or wrong. Obviously, such an evaluative dimension is also relative, viz., relative to the person or groups of people who 'find' that something is good/bad or right/wrong. Whereas the criteria used to establish true beliefs, viz., knowledge, are epistemic, the criteria underlying these evaluative beliefs are norms and values .

Thus, the belief 'that Reagan was a bad president' is an opinion if the belief is the result of a judgment in which value-criteria for the quality of presidents are being applied. We especially use the notion of opinion when the application of such values to a particular case is or could be disputed, that is, when there are 'differences of opinion' (Billig, 1991). This further property of opinions also suggests that opinions, just like knowledge, may have a social or cultural basis. This also distinguishes opinions from beliefs that happen to be false. The belief that Reagan was president in 1995 is (for all we know) false, but it does not seem to involve an evaluative dimension, and we would usually call such a belief an 'erroneous' belief or 'error' rather than an opinion. Like (justified and true) knowledge, these false beliefs are also factual , in the sense that we assume that the same epistemic criteria apply in order to accept or reject them as true or false.

Thus, whereas both true and false beliefs are defined here as factual, opinions are defined here as evaluative. The difference lies in the kind of social evaluation criteria being used and the social uses and functions of factual beliefs and opinions: Opinions thus seem to deal more with the evaluators and their norms and values, whereas factual beliefs rather focus on the 'objective' or intersubjectively accepted (observable, non controversial) properties of things, people and events.

As may be expected, the distinction between (true or false) factual beliefs and evaluative beliefs is not sharp. This is, first of all, the case because the criteria on which the distinction is based are not always mutually exclusive. That John is a thief may be a true belief, and hence knowledge, if John was convicted of theft by a court of law, and if such a conviction is taken as the sole criterion of truth in this case. Such a belief may be quite matter-of-factlike, viz., if we have no positive or negative opinion about his 'being' a thief. In more commonsense language we would say: If we could not care less whether John is a thief or not, this is knowledge rather than an opinion . This is one of the reasons why knowledge is often called a form of 'cold' cognition, and opinions a form of 'hot' cognition.

We may also have the opinion that John is a thief (whether or not this is true), if we believe he appropriated something that did not belong to him, and if we think that that action was wrong . The mental process of evaluating people, actions or events relative to some personal or social norms or values will here be called, as usual, a judgment (Arkes & Hammond, 1986; Martin & Tesser, 1992). That is, unlike knowledge, opinions are the result of a (value) judgment. To have a comparable term for the establishment of truth on the basis of truth criteria or verification instances, we shall use the term assessment. Thus, in order to know, we may find out (verify) whether Reagan was a president in 1995 (e.g., by checking the media or history books), but in order to find that Reagan was a bad president we would rather use value-based arguments than to go to the archives. The respective adjectives used in relation to knowledge and opinions are epistemic and doxastic , respectively.

Since judgments are usually socially acceptable if only we defend them in terms of our norms and values, which may be quite personal, opinions were also described above as a property of the relation between persons and the things being evaluated (a relation sometimes, but rather inappropriately, called a 'propositional attitude' in philosophy). In other words, the evaluation basis of judgments is personal or internal (even if this basis may be shared with other people). To assess the truth of a belief, however, we usually need to refer to 'outside' evidence or proof: Simply referring to our own point of view will not usually count as sufficient evidence, unless the beliefs are based on personal experiences, as is typically the case in storytelling.

Despite these approximate distinctions, much of our knowledge (e.g., about theft) may be imbued with evaluative elements, so that the distinction between factual and evaluative beliefs is not always clear-cut. Thus, many predicates whose use is normally based on knowledge criteria, may also be used in a more evaluative sense. For instance, we may use the predicate 'to be the capital of' epistemically in 'Amsterdam is the capital of the Netherlands' and doxastically in 'Amsterdam is the capital of drugs'. As we shall see, this also brings in contextual and social aspects of beliefs, to which we shall turn below.

Whatever the basis of judgment, however, as soon as a belief presupposes norms and values, we shall here call it an opinion. This also means that the 'same' belief may be an opinion in one context (for one person) and knowledge in another context. For instance, the belief 'smoking is bad for your health' may be a negative opinion shared by many people, especially when being used to prevent others from smoking; in another, e.g., medical, context, it may be an epistemic conclusion from scientific proof, that is a (true) factual belief, that is, knowledge.

Incidentally, this last example does not imply that all scientific beliefs are necessarily factual beliefs, even if there is a strong and widespread ideology that this is the case, or at least should be the case. Nor do we hold the skeptical opposite opinion, viz., that all beliefs (scientific or otherwise) are always merely opinions because all assessments and their truth criteria eventually also imply personal or social norms and values, perspectives or points of view. In other words, for all practical purposes, the assessment of beliefs as knowledge maybe relative to socially and culturally shared truth criteria, but these will simply be accepted as a (practically) sufficient basis of truth. This theory of truth also allows what we think it should allow, viz., that truth criteria undergo historical change, e.g., as a consequence of social changes, scientific research, technological advances and experiences as well as their consequences on our knowledge (and knowledge criteria) in everyday life.

Social beliefs 

As has been suggested several times above, beliefs may be shared by more than one person. Indeed, the very notion of knowledge implies, as we defined it, that other people share the same true belief: Beliefs (about other things than personal experiences) held by one person may well (later) turn out to be true, but are usually not (yet) accepted as such, and hence treated as 'mere beliefs'. Henceforth, we shall use the notion of knowledge only in this sense of socially shared true beliefs, and mark true personal beliefs (e.g., those of personal experiences) specifically as such, viz., as 'personal knowledge'. This is not merely a philosophical decision, but also a cognitive one: Knowledge will be henceforth dealt with only as a special type of cognition, viz., as social cognition . Its mental representation is therefore also a form of social representation (Farr & Moscovici, 1984; Breakwell & Canter, 1993; Fraser & Gaskell, 1990).

Technically, social knowledge is usually located in 'semantic' Long Term Memory, which we shall simply call Social Memory , because such social representations are shared by social groups and communities and acquired, used and validated essentially in social contexts. Personal knowledge, e.g., in the form of personal experiences, on the other hand, will be thought to be represented in Episodic Memory (or Personal Memory)(Tulving, 1983). Such personal experiences may of course be shared with others, typically so in storytelling, but this need not result in the kind of 'common, shared beliefs' we here call knowledge, and which require person-independent and event-independent knowledge of groups of people.

Since social knowledge is usually context-free , it will usually also be abstract and general , for the simple reason that it need to be applicable, by all group members, in many different social situations and communicative events. An exception to this principle forms the shared social knowledge of historical facts , e.g., the news events we read about in the paper, see on television or know from our history textbooks: These may be socially shared but are at the same time specific and concrete.

At the same time as we may have both specific and general socially shared knowledge, we may probably also have both specific and general personal knowledge. Personal experiences may embody knowledge about specific events, but also general knowledge about our personal experiences: I may know that my neighbors came to dinner yesterday, which is knowledge about a specific events, but obviously I also have more general personal knowledge about my neighbors.

The same reasoning applies to opinions . People may have personal opinions, but they may also share them with others, that is, with other members of a group. Thus, I may have a typical personal opinion about my neighbor, but opinions about e.g. social issues such as abortion, nuclear energy or unemployment tend to be shared by other people. According to the definition of opinion, social opinions must be based on socially shared evaluation instances, on the sociocultural norms and values shared by a group or community, and thus are measured and evaluated relative to a moral order (Eisenberg, Reykowski, & Staub, Rokeach, 1973, 1979).

Complex structures of such social opinions (for instance about an issue such as abortion) will here be called attitudes . That is, unlike much attitude research in social psychology (Eagly & Chaiken, 1993), we carefully distinguish between --specific, personal-- evaluative beliefs (opinions) and social, general and abstract evaluative beliefs shared by the members of a group, viz., attitudes (Jaspars & Fraser, 1984). The notion of 'attitude' will here only be used in this latter, social sense. This implies that in our theoretical language it does not make sense to say that individual persons have an 'attitude' about this apple, their neighbor or about the war in Bosnia. They either have personal opinions about them, or, in case of socially known events or issues, they may share (social) opinions or attitudes with others. In sum, together with knowledge, attitudes are part of social cognition and are mentally represented as social representations.

Social representations of knowledge and attitudes do not consist of arbitrary collections of beliefs. Most contemporary psychologists assume that they must be structured in some way, if only because such structures facilitate learning, use and change. Thus, for knowledge notions such as scripts and frames were introduced to account for the internal organization of knowledge, e.g. of stereotypical events such as going to the movies, or for our conventional knowledge about houses, respectively (Schank & Abelson, 1977). Such structures may be seen as schemata , consisting of a hierarchically ordered and linked number of categories. Somewhat less successfully, such attempts at a more sophisticated organization of social memory have also been made for attitudes (Abelson, 1976). So far, however, no generally accepted representation format for attitudes has been proposed in the literature, and we shall further ignore this important issue in this paper (see also, Fiske & Taylor, 1991; Himmelweit & Gaskell, 1990; Wyer & Srull, 1984, 1989).

Models 

Personal experiences, including personal opinions, as represented in Episodic memory, are currently usually represented in terms of mental models (Johnson-Laird, 1983; van Dijk & Kintsch, 1983; van Dijk, 1987; van Oostendorp & Zwaan, 1994). Since this notion is also crucial as a basis for the production and comprehension of discourse, we need to briefly introduce it after the previous notions of social representations.

Contrary to socially shared representations, models are personal and subjective. They represent the individual ways people make sense of their environment, and how they interpret events and discourse. Models represent what we usually call our 'experiences'. Thus, each event, when understood and represented by an individual, will be represented as such a model. The same is true for discourse comprehension: Understanding a text not only involves the construction of a textual representation in memory, but also the construction (or updating) of a mental, episodic model. We say that we have (more or less) 'understood' a text if we are able to construct a (partial) model for this text. Similarly, during discourse production, people activate or form a new model and use this as a basis for their text or talk.

In discourse, thus, a model is a subjective representation of what the text is (to be) 'about'. This offers also a useful cognitive counterpart of the notion of reference and referent, and a solid basis for the subjective and intersubjective definition of coherence: A text is meaningful and coherent (for some language users) if it has a mental model. In other words, it is not the 'world' or the 'events' themselves that make discourse meaningful, coherent or indeed true or false, but the model people construct of such events. All meaning and interpretation is therefore relative to our models. We see that models help explain many of the familiar problems of the philosophy of language, epistemology and the theory of discourse.

Relevant and crucial for our purposes here is that models not only represent true or false (factual) beliefs, but also opinions . Thus, whereas groups represent their opinions in shared mental representations (much in the same way as they do with grammars), individuals represent their personal evaluative beliefs, e.g. about specific events, in their models.

At present we ignore the internal structures of such models. If models routinely are constructed to understand events or discourses about events, they must have a strategically applicable schematic form. Such a form may be thought of as consisting of a number of relevant categories, such as Setting (Time, Location, Circumstances), Participants (Agents, etc.), Events or Actions, as well as their evaluative modifiers, representing personal opinions. That is, each time we witness or participate in an event or read or hear about it, we apply such a schematic form to it in order to understand it. Since event representation is of course routinely related to the ways we talk or write about events, it is not surprising that the semantic ('case') structure of sentences (or of stories) is in several ways similar to the structures of the 'underlying' models used to communicate about such events.

Obviously, such a fast, strategic process may not be perfect, and we may therefore also misinterpret events, and construe 'wrong' or 'biased' models. Such biases may also result from (biased, group based) knowledge and opinions, prejudices and stereotypes, e.g., when the 'facts' do not square with our general representations about them. As we shall see later, editorials are typically based on such personally or socially biased models of editors.

Models have another fundamental role in mental processing: They represent the interface between social representations of groups, on the one hand, and individual experiences, social practices and discourse, on the other hand. Thus, social members may activate, instantiate and integrate elements of socially shared knowledge and attitudes as (specific) beliefs in their models. For instance, general knowledge and attitudes about civil wars may be instantiated in a model representing a person's knowledge and opinions about the civil war in Bosnia. This relationship also works conversely: Once constructed a series of models about the same or similar events, people are able to infer more general, and more abstract factual or evaluative beliefs from them. That is, one way people learn is by model generalization.

Of course, such processes take place in a social context, and need social 'normalization': Once generalized, models still need to be compared with the knowledge shared and expressed by other social members. This is especially the case for 'accepted' knowledge. For general opinions the social criteria may be less strict (although they also apply): We may accept to share our opinions only with a small number of people, and not with everybody, but will generally expect or claim that our knowledge is more generally shared. Hence the conventional distinction between 'subjective' and 'objective' beliefs.

This is also one of the ways that prejudices are being formed and changed, viz., as non-warranted overgeneralizations of models of personal experiences (or of models of events talked about by others). Of course, knowledge and attitudes (including prejudices) may also be formed directly, that is, without the generalization of personal models, e.g., when people accept social representations directly from other group members, e.g., through persuasive discourse. Propaganda and manipulation are well-known examples in which knowledge and attitudes may be influenced without the interface of personal experiences. The same is true for editorials, as we shall see below.

The briefly summarized theoretical framework discussed so far, is able for the first time to describe and explain some of the processes involved in editorial influence, manipulation and the 'manufacture of consent': No theory of discourse understanding and social influence is complete without an account of the role of models and their relations to shared social representations.

Context models 

Essential in our analysis of editorials, as well as for the study of discourse in general, is the assumption that language users not only construct models of events they talk about , but also of the communicative events in which they participate. Such models will be called context models (van Dijk, 1996a). Such models represent the knowledge and opinions speakers and writers (as well as recipients) have about themselves and each other, in various communicative and social roles, as well as about the setting, circumstances, intentions, goals, purposes, and other properties of the context. That is, also contexts are subjectively represented by language users. Contrary to standard conceptions in social psychology and much discourse and conversation analysis, it is crucial to emphasize that therefore it is not the context itself that influence our discourse or the interpretation of discourse, but our subjective models of the context.

Thus, whereas event models as discussed above provide the (subjective) knowledge we have about events we talk or write about, that is, form the basis of discursive semantics, context models provide all necessary information for the contextually variable or pragmatic information that influences the interaction strategies (e.g., of politeness, self-presentation, etc.), the style, rhetoric and the surface structures of text and talk. That is, as we shall see for editorials, many of the properties of opinion discourse are not merely based on the opinions we have about some fact or issue, but also, and crucially, on the contextual constraints on how to formulate such opinions, as represented in context models. Indeed, as is well known, editorial writers thus represent and express their identity as journalists, elites, middle class, etc., as well as their assumptions about the expectations, knowledge and attitudes of the readers. A theory of context models allows us to finally describe and explain the relations between these properties of the social context on the one hand and discourse meaning and structure, on the other hand.

Ideologies 

Finally, on the social side, we introduce the notion of ideology as the basis of evaluative social representations. That is, an ideology, as defined here, is a fairly abstract system of evaluative beliefs, typically shared by a social group, that underlies the attitudes of a group (van Dijk, 1995a). Thus, a racist ideology may be assumed to organize (and provide the general beliefs for) more specific attitudes about e.g., immigration, integration, housing, work, education, etc. of immigrants and minorities (van Dijk, 1996c).

This conception of ideology is more specific than most traditional approaches to ideology, which either vaguely define them as false beliefs, as false consciousness, as belief systems, as common sense or even reduce them to 'ideological' social practices or discourses (of the huge literature on ideology, see e.g., the following books for a history and discussion of this notion: Billig, 1982, 1991; Eagleton, 1991; Larrain, 1979; Thompson, 1984, 1990).

While they are evaluative, ideologies also embody the sociocultural norms and values that are relevant for each social group. Thus, the sociocultural value of Equality may be more relevant for minorities and women, and hence have a prominent place in feminist and anti-racist ideologies, than in the ideologies of corporate managers. Similarly, the same value, e.g., Freedom, may have a different function for managers (freedom of the market) than it has for women, youths, or immigrants (for whom freedom from oppression or control may be more relevant).

Ideologies are commonly assumed to represent, socially, the major interests of a group. Sometimes the notion of ideology is therefore restricted to the study of 'dominant' ideologies that function as the legitimation of dominant group power and interests. We shall however not reduce the notion in this way, and therefore also use it to describe the ideologies of other groups, e.g., dominated groups (e.g., women, minorities) or professional groups (doctors, journalists) (see also Abercrombie, Hill & turner, 1980, 1990).

An analysis of the cognitive representation of such interests may involve such fixed categories as Membership, Activities, Goals, Norms and Values, Position and Resources of a group. Such a schematic representation will often take the form of a schema a group has about itself and its relations to other groups (e.g. racists with respect to minorities, men with respect to women, or vice versa, and so on). In their social and economic context, this means that ideologies are often patterned after the social conflict and struggle they help to manage for group members: Ideologies are often polarized, e.g., when representing Us vs. Them. Unfortunately, the further internal properties and representation formats of ideologies are as yet largely unknown. We may however assume that such a format should satisfy the social constraints of the acquisition, uses and functions of ideologies in society, and for social groups, as well as their cognitive functions in the management of social representations, e.g. in the organization of attitudes.

We now have a first impression of the cognitive framework that underlies the processes and the structures of discourse in general, and editorials in particular. Thus, we distinguished between personal and social representations. Personal representations of knowledge and opinions are coded in event and context models, which eventually will be used (constructed) to produce or understand editorials. At their basis we find the social representations of knowledge, attitudes and ideologies, as shared by groups. These social representations are related to models by processes of instantiation in model formation, and by processes of generalization and abstraction of models in the formation and change of social representations.

Obviously, each social practice or discourse will be a function of both, viz., of the possibly variable individual representations in models, on the one hand, and the social representations editorial writers share with members of other groups of which they may be a member (other editors, other men, other middle class people, other U.S. citizens, other white people, other conservatives, and so on). Precisely because editors (like all language users) may be a member of different groups, some of which have conflicting interests, and hence conflicting ideologies, we may also expect their opinions (as represented in their models) to reflect such contradictions. For our analysis, especially these (conflicting) social identities and their mental representations will be relevant in a critical study of editorials (for a social and political analysis of editors, their allegiances and group memberships, see, e.g. Altschull, 1984; Bagdikian, 1983; Herman & Chomsky, 1988; Lee & Solomon, 1990; Lichter, Rothman & Lichter, 1990; Paletz & Entman, 1981; van Dijk, 1995b).

Representation formats for opinions 

The notion of representation, used for beliefs in general and for opinions, knowledge and attitudes in particular, presupposes some kind of representation 'format' or 'language'. That is, we need to know how to represent such beliefs in memory. This is a tricky issue which we shall avoid in this paper. Traditionally, beliefs are often represented as the propositions of some kind of abstract cognitive language ('mentalese'), and this also explains why we usually talk about 'true' beliefs, although the notion of truth should perhaps be applied to sentences (expressing belief propositions) rather than to mental representations, which may be said to 'correspond' (or not) to the events they represent. There are many other ways to talk about mental representations and the 'world' of facts, and hence about truth, and truth criteria. The same holds for the representation languages we may use to describe mental representations.

Because of a lack of practical alternatives, everyday language or some form of formal language (e.g. predicate logic) is commonly used. This is no problem as long as we know that the 'mind' does not 'think' in some natural or formal language. That is, representation languages are merely conventions that allow us to intelligibly speak about mental representations, which for all we know are 'coded' in some form of neural network of the brain. This also means that the current language used to describe memory and mental representations is abstract (and non-committal) with respect to the 'real' processes and structures of the mind-brain. For the same reason, current connectionist approaches to the mind claim to be closer to the neural network structure of the brain. This issue and its many complications are however beyond the scope of this paper (see, e.g., Baumgartner & Payr, 1995; Clancey, Smoliar & Stefik, 1994; Rumelhart & McClelland, 1986). Moreover, this recent work offers at present no insight into the nature, structure and functions of opinions or into the processes of their discursive expression.

Given a propositional or natural language to describe social representations, often in combination with various kinds of network or tree-diagram, opinions will simply be distinguished from knowledge by the presence of evaluative predicates, viz., predicates whose use implies norms or values. We have seen that this is a problem, because the same predicate may be used as forming a true belief in one context, and as forming an evaluative belief or opinion in another. We resolve this problem by separating different mental representations in the mind, viz., as knowledge and at attitudes. That is, if a proposition occurs in an attitude, it is by definition the result of the application of norms or values, and hence an evaluative belief or opinion. The 'same' proposition may also occur as part of social knowledge. For instance, the belief 'Many Turks migrated to the Netherlands' may well represent merely a fact for many people, but it may also have a negative implication for many other people, viz., those who resent such immigration. As a practical distinction of representation, thus, factual beliefs and evaluative beliefs may be preceded by the propositional functions 'It is (generally) known that', and by 'It is a bad (good, etc.) thing that', or 'I (or We) find that...', respectively. With these few tentative remarks, we however leave the discussion on representation formats alone.

Opinion discourse 

Editorials are a genre that may be characterized both as a special type of media discourse, as well as belonging to the large class of opinion discourses. Opinions may be expressed by language users in many types of discourse, first of all in everyday conversations, argumentations and any other discourse in which (dis)agreement is expressed or persuasion enacted (Antaki, 1994; Billig, 1987; Pomerantz, 1984; van Eemeren, et al., 1987; Wegman, 1994).

Of course much fewer discourse types typically have as their typical main function the expression and persuasive communication of opinions. In the class of opinion discourses, for instance, we find such diverse genres as letters to the editor, complaint letters, political party propaganda, racist pamphlets, and so on. The functional typicality of such genres is both contextually defined in terms of the intentions and purposes of the speaker or writer, as well as in the expectations of the recipients. Thus, typically, readers of newspapers expect editorials to formulate opinions about recent news events.

Discursive vs. mental opinions 

We have stressed above that we should distinguish between opinions as mental representations, and the ways such opinions are formulated in text and talk. One of the many reasons for that argument is that we need to be able to say that the 'same' opinion can be expressed by the same person on different occasions, or that the 'same' opinion can be shared by several people, or even whole groups and cultures. If we reduce opinions to their actual formulation in discourse, we would have as many opinions as we have textual formulations of them.

The fact that, empirically, we often study opinions through their expressions or formulations in discourse, does not imply that these two dimensions, viz., the discursive and the cognitive, should be conflated. Indeed, we may not only infer from discourse what people's opinions are, but also from other social practices, as when racist opinions may be inferred from discriminatory actions.

Such inference processes are not simple or straightforward: Precisely one of the reasons why we think it is necessary to distinguish opinion discourse from the actual opinions, is the fact that people may not express their opinions, or may indeed express different ones, e.g. for reasons of politeness, face-keeping and impression management, or in general under the constraints of social norms and values. This means, among other things, that we need a complex methodology of unobtrusive observation, analyses of discourse in different contexts, and so on, in order to infer opinions from opinion discourse. In sum, when analyzing opinion discourse, we need to be aware of the fact that its meaning is not always a direct expression of the opinions of (the event models of the) speakers or writers, but that various transformations (deletion, change, substitution, etc.) may take place for the reasons represented in people's mental context models.

It could be argued, as it is done in discursive psychology that, at least in a social context, (mental) opinions, even if these should exist, are of no (social) consequence (Billig, 1991). What is relevant is how people express themselves, and it is merely their manifest opinions and their formulations that are available to others. There is one point in this argument that is obviously sound: We also need to study, in detail, the ways opinions are expressed and persuasively communicated, as well as the possible social effects of these opinions discourses. However, in order to describe and explain the fact that despite contextual variations, the same people (or groups) may express the same or similar opinions in different contexts, we need to postulate some underlying invariant, viz., an opinion that is not the same as its variable expression in text or talk. Most social phenomena, such as public opinion, prejudices and stereotypes, and many other socially shared representations, cannot simply be reduced to their contextually variable expression. We also need to account for continuity, permanence and coherence, both of opinions themselves, as well as for their 'typical' expression in discourse and other social practices. There is nothing mysterious about that: A similar distinction is made between the (shared) knowledge about the fact that Amsterdam is the capital of the Netherlands, and the ways this knowledge may be variably expressed and functioning in different languages, styles or contexts.

Of course, as we have done, we do make a distinction between the more general or abstract opinions people have and share, and the locally emerging or construed opinions being uttered in specific contexts . This is especially true for opinions about elements of the present context itself, e.g. opinions of the speaker about the recipient, about the nature of the conversation, as well as about ongoing events, and so on. Also, while talking, arguing or negotiating, people may of course 'change their minds', so that a new opinion may well be construed during ongoing discourse. This indeed, accounts for one way people actually form new opinions. But the same analysis implies that the other speaker may have kept his or her opinion after having persuaded the first, or that the speaker who just formed a new opinion is likely to defend that opinion in the next conversation.

In sum, in the discursive expression, formation, and change of opinions, we constantly find a dynamic interplay between continuity and change, and both need to be accounted for. Moreover, we should not forget that 'changing one's opinion' is not a discursive process, but a mental one, and so is the result of that process. Finally, we already emphasized that what is being changed here are generally personal opinions, as represented in models, about fairly concrete events, people or actions. More fundamental, socially shared opinions (e.g. about nuclear energy or abortion) are hardly formed or changed during one particular discourse. On the contrary, types of discourse may precisely be classified according to the same opinions being expressed by them (as is the case for party propaganda, ecological pamphlets, or indeed editorials).

Organizing and supporting opinions: Arguments 

Opinion discourse is usually not merely a list of variable expressions of underlying opinions. Such opinion expressions tend to be structured in specific ways. The most prominent feature of opinion discourses is that opinions are supported by sequences of arguments: Opinion discourse is argumentative. Although this paper cannot possibly summarize all properties of argumentative structures that have been studied since classical rhetoric until the broad field of argumentation today, we need to make a few remarks about them that are relevant for our study of editorials (for details, see e.g., Billig, 1987; van Eemeren & Grootendorst, 1992; van Eemeren, et al., 1991, 1993).

One of the fundamental considerations that explain the very notion of argumentation in the first place is the communicative constraint on the expression of opinions, viz., that recipients expect or demand that opinions are made plausible, defended, supported or otherwise 'backed up'. Especially since opinions may be personal, or may vary from group to group, often requires that people are supposed to show why they have such an opinion anyway. This is especially the case for all controversial opinions, or where there is a conflict-of-opinion. In contexts where identical opinions are presupposed (e.g. in 'like-minded' groups), or in cases of total consensus or common sense opinions, less or no argumentation will be needed to support opinions.

Secondly, not only opinion expressions but also the arguments adduced to support them, are based on underlying mental representations, which may be as varied as the opinions themselves. Thus, they may include other opinions and factual beliefs of (specific) models, for instance when dealing with causes and consequences of fact, or they may be items of more general, socially shared knowledge, attitudes, ideologies, norms and values. Of course, while shared, social representations will generally be very effective as underlying beliefs that may support opinions, and this will be especially the case for undisputed beliefs. Indeed, as is the case for discourse in general, general beliefs often need not be expressed at all, but may be presupposed. In argumentation this will be the case for tacit (implied) premises.

This close relationship between opinion expressions and arguments on the one hand, and their mental basis, on the other, might suggest that thinking itself is a form of arguing or even rhetorical in nature (Billig, 1987). As is the case for all discourse, we may of course have various forms of mental 'preformulation' of arguments, as we also may have such preformulations for storytelling or giving lectures. This does not imply, however, that our mental processes themselves are 'argumentative', 'rhetorical', or for that matter 'narrative' or 'expository'. Again, we should be careful to make an analytical distinction between mental representations and processes on the one hand, and structures of discourse on the other. Processes of thought are of a different nature than discursive argumentation, which is inherently contextual and social, and taking place in forms of interaction and communication. Even if we take a cognitive approach to discourse structures and meanings, we should distinguish between mental representations of events and opinions and mental representations of (linguistic) meanings and discourse structures.

Of course, context models allow us to represent (plan or recall) argumentative talk, but context models are just that, mental models, and not themselves forms of interaction and communication. Similarly, making inferences from propositions, or searching beliefs in support of a belief, are cognitive processes that cannot be reduced to the rhetorical or other structures of text and talk. This does not mean, of course, that such discursive structures may not be used as heuristic devices in the construction of theories of thinking. Instead of reducing discourse to mental representations or vice versa, it is especially relevant to study their complex relationships.

Discursive formulations of opinions 

In the rest of this paper, we shall focus on the discursive formulation of opinions rather than on their argumentative support. Our point is that besides an appropriate cognitive theory of opinions, we also need an explicit discursive one, viz., a theory that shows at what levels and by what grammatical and discursive structures opinions tend to be expressed in a contextually adequate way.

The task of formulating discursive principles for the expression of opinions is daunting because first of all, as we have seen, the very distinction between factual and evaluative opinions is not very precise. Thus, predicates like 'thief' may appear both in descriptive, factual statements, as well as in evaluative expressions, for instance in accusations. Since language users mostly are able to infer whether a factual or an opinion statement was made, they apparently have the competence to infer from text and context structures what kind of statement was made. This means, in our case, that each opinion statement need to be studied, first, in relation to other statements or structures in the text (e.g., some form of argumentation may suggest that an opinion rather than a factual statement was made or intended).

Secondly, context and hence context models are important: Given our knowledge of the speaker, the situation, the medium, the genre, and so on, we may strategically infer that opinions rather than factual beliefs are being expressed, or vice versa. Thus, as suggested before, in party propaganda, a racist leaflet, from a sexist man talking about women, or indeed in an editorial, opinions are being expected, rather than (only) factual statements. But such contextual constraints may be neither sufficient nor necessary: Some opinion statements are such in all or most contexts, and also in 'evaluative contexts', speakers or writers will of course also make factual statements. Moreover, although intentions and purposes are part of the context, we seldom have access to them (unless again through discourse). In sum, for the identification of opinion statements, complex textual and contextual analysis may be involved: A simple identification of 'evaluative' words (such as good, bad, honest or beautiful ) often won't be enough to establish whether an opinion is being expressed.

It may be provisionally assumed that most opinions tend to be formulated as part of the 'content' or 'meaning' of discourse, that is, as properties of semantic representations . Indeed, if opinions are represented as part of mental models of events, then the selection of the 'information' from that model that is included in the 'content' of discourse will most likely also feature semantic representations of such opinions. That is, lexical items and their semantic combinations in propositions and sentences, are the first level or dimension of discourse to look for formulations of opinions.

The same is probably true for more complex semantic structures, such as relations of local or global coherence, discourse topics, and so on. That is, the evaluative part of a sequence may not be constructed in an isolated proposition, but in a specific sequence. Indeed, relating events with their causes and consequences (or rather their model representations) that define text coherence may as such be taken as evaluative. Opinions may also be involved in implications, entailments and presuppositions: Mentioning something explicitly or not may be an important move in expressing one's opinion. Similar remarks hold for overall topics: Since topicality implies importance and prominence, topical or thematic functions may themselves imply evaluations. In sum, semantics is a major field for the 'realization' of underlying opinions.

However, things are not that simple. Since interpretation is contextual and not limited to discursive meanings, we know that also context models feature opinions, and these may influence many properties of discourse, from speech acts and interactive strategies (e.g. of politeness, impression management and persuasion) to variably style and rhetoric. That is, very often opinions are not so much express by what is being said, but rather how it is being said. We may call our loved ones with the most awful of even negative words, but in the appropriate contexts and with the right intonation, it may be inferred that the negative description was in fact a playful compliment. The same is true for the rhetorical irony of an editorial.

In other words, stress, volume, intonation at the level of sound structures of discourse, syntactic and other variation of the level of sentence structure, as well as other dimensions of style and rhetoric also provide clues about whether or not some expression was meant to be understood as factual or evaluative. Whether for instance a specific intonation by itself may be said to express an opinion, or whether this is only the case when such an intonation is related to specific meanings, may be a matter of contention, but generally surface structures will only be strategic means in the interpretation of meanings, actions or functions of discourse. Yet, even this brief argument shows that for an analysis of opinion in discourse, we cannot limit ourselves to semantics.

Editorials 

Editorials are public, mass communicated types of opinion discourse par excellence. After opinion programs on TV like talk shows, and together with the Op-Ed articles (i.e. opinion articles that are placed at the Op osite page of the Ed itorial page in much of the U.S. press) of columnists and other writers, they are probably the widest circulated opinion discourses of society, whether or not all readers of the newspaper read them daily. Their influence may not so much be based on massive popular influence, as rather on their influence on the elites. We may assume, for instance, that members of parliament or Congress, cabinet ministers, corporate managers, and other leaders follow the opinions of the most respected newspapers. Indeed, much critical media research suggests that general opinions of newspapers cannot be fully inconsistent with those of other elite institutions, and that processes of influence here are mutual.

A next point to be stressed, therefore, is the fact that editorial opinion is generally institutional, and not personal. Even when written by a single editor (editorials are seldom signed), editorials count as the opinion of 'the' newspaper. This means that they will generally be shared among several editors, or between editors and management, or between editors and other social groups they belong to. Important, therefore, is the realization that whatever specific (model based) opinions about specific events are being formulated, they will tend to be derived from social representations, rather than from the personal experiences or opinions of an editor.

This sociocognitive foundation of editorials also shows in their structures: First person pronouns and stories about personal experiences will be quite rare in them. On the contrary, editorials are 'impersonal', focus on public (news) events, and support general (social, economic, cultural or political) opinions, usually shared by other elites. Also other elements of style will mark this institutional, public, more or less formal properties of the context, e.g., in the selection of lexical items, syntactic structures and modes of argumentation. Despite such contextual constraints, there will be vast stylistic differences between e.g., an editorial in the British tabloid The Sun and those in newspapers 'of record' such as the New

York Times, Le Monde or El Pais. 

Most routinized discourse types have more or less conventional schematic structures, a schema consisting on a number of canonical categories which define the functions of the respective 'parts' of the text. No such conventional structure of editorials has thus far been proposed in any theoretical and empirical detail. We only know that like other types of press discourse, they have headlines, although these may generally be brief and consist of one word. However, given their institutional functions as editorial 'comments' on todays news events, we may at least expect the following schematic categories:

1. Summary of the event.
2. Evaluation of the event --especially of actions and actors.
3. Pragmatic Conclusion (Recommendation, advice, warning).

Thus, each editorial will have to briefly tell 'what happened', if only to tell what it will be about, and to remind the readers what the event was as it is more fully described in concurrent or earlier news reports. Of course this may be very brief, e.g., a simple nominalization, when the event is already widely known ('yesterday's bomb attack').

Secondly, it needs to be spelled out in the editorial what was good or bad, wrong or right of the event, and especially about the people involved as (responsible) actors.

Thirdly, newspaper readers may expect an answer to the question 'What next?', 'What are we going to do about this?', and such an implicit question may be answered by the third, concluding part of the editorial, that is, by some kind of pragmatic 'coda' (as in many stories): What (e.g. the authorities) should or should not do.

We may also expect that the first part (summarizing the event) will be more or less factual, although the description of the event may itself take place in partly evaluative terms (as in 'Terrorists yesterday bombed...'). The second, evaluative part, will focus on the opinions of the newspaper, and will involve values and underlying ideologies. The third, pragmatic part, about what should/could (not) be done is action-oriented, and therefore based on norms. Of course, all these predictions are purely theoretical. It may very well be that in the actual editorial different categories and modes of expression are mixed.

An example 

Let us now look at a concrete example, and examine in some detail how opinions may typically be formulated in editorials. The function of this example is not primarily to 'test' the theoretical framework explained above, but to show how the (mental) representations of opinions and ideologies are eventually manifesting themselves in discourse of a specific genre. Since such expressions may be fairly indirect, much of our analysis consists in making inferences about opinions that are not explicitly expressed in the text.

Once we have insight into these relationships between text and opinion, we may design experimental and discourse analytical methods to 'test' the cognitive-discursive theory of opinions. Discursive evidence of the theoretical distinctions made above may be found for instance in explicit, common sense distinctions being made in talk and text between different types of beliefs as well as in the argumentative or other discursive structures that organize the expression of such beliefs.

In order to enhance the readability of the analysis, we shall often refrain from repeating the mental properties of the opinions being inferred from the text. It is understood however that when we talk about opinions, we mean 'representations of opinions in event or context models'. And when we talk about knowledge, attitudes and ideologies, these are assumed to be forms of socially shared mental representations. The same is true for the relations between the opinions (of the newspaper) being expressed in this article and the more general social representations which they instantiate.

In other words, we are not simply 'reading' or 'interpreting' this text, but in fact (rather informally) reconstructing the event and context models of the editor(s) of the WP, as well as the preferred models they want to convey through their persuasive editorial discourse. Discourse analysis thus also becomes (and presupposes) cognitive analysis, whereas both discourse structures and cognitive structures may then be 'contextualized' by examining their social and political functions.

Our example is an editorial selected from our large data base of 1993 of the New York Times and the Washington Post , viz., "Explosion in Managua", published in the Washington Post of August 5, 1993. This editorial formulates the Post's opinion about a recent event in Nicaragua, viz., the explosion of an arms cache allegedly built up by the (ex-)Sandinistas. Its analysis will be interesting, since the opinions and ideologies involved may be expected to be related to prevailing mainstream and especially conservative concerns in the USA with Nicaragua, e.g., after the Sandinista victory over Somoza, the Iran-Contra affair, and especially (former) President Reagan's support for the Contras, whom he saw as 'freedom fighters.' More specifically, the U.S. support of the campaign of Violeta Chamorro led to her election as president of Nicaragua. Let us therefore examine some of these opinions and ideologies in more detail. We do this by sequentially analyzing some of the sentences of this editorial (for the full text, see Appendix; note that literal editorial text quoted below in my own text is marked by double quotes, and meanings, special meanings and concepts with single quotes):

[1] VIOLETA CHAMORRO'S gamble on rebuilding Nicaraguan democracy in partnership with the Marxist-oriented Sandinistas, whom she'd defeated in elections, blew up on May 23.
The schematic (superstructural) function of these first three sentences is, as suggested above, to give a Summary of the event this editorial takes as occasioning its opinion, viz., an explosion in an arms cache. This summary is not limited to a brief description of the event itself, but also features a broader Political Context category as expressed in [1], viz., the current attempts of Chamorro to cooperate with the Sandinistas.

This context description is not merely factual, but involves a number of opinions. Thus, the choice of the words "rebuilding democracy" (instead of for instance "rebuilding the country" or "maintaining democracy") presupposes that there was no democracy before Chamorro came to power, an opinion which (together with our general, historical knowledge about Nicaragua) implies the opinion that the Sandinista government was undemocratic. This opinion is consistent with the prevalent U.S. attitude about the Sandinistas. Indeed, this opinion is also marked by the common use of "Marxist-oriented" as a qualifier of "Sandinistas" in the same sentence. This is standard practice in the U.S. media for all non-capitalist countries and governments. Indeed, for comparison, we may observe that Violeta Chamorro is not similarly described as "capitalist-oriented", "market-oriented", "U.S.-oriented" or, to extend the context description, as "who came to power with U.S. support", which could be factually based opinions from the Sandinista perspective. Finally, the use of lexical item "gamble" implies that cooperation with "Marxists" is 'risky', as the metaphor "blew up" also suggests.

Structurally, we see that the negative opinions in this first sentence are expressed or signalled by specific lexical selections, presuppositions and implications, qualifying descriptions and metaphor. We also see that understanding these expressed beliefs as negative opinions also requires historical knowledge of Nicaragua, as well as about the role and position of the U.S. in Nicaraguan affairs, as well as attitudes and ideologies of mainstream U.S. institutions about Marxists in general as well as about the Sandinistas in particular (for details, see e.g., Chomsky, 1985, 1987a, 1987b, 1989, 1992, 1993, 1994).

Theoretically, this analysis suggests how models and social representations are expressed in text. For instance, we find complex and variable discursive evidence for the presence of opinions in the model of the author about the events in Nicaragua, but these opinions are not always expressed explicitly, that is, as complete propositions as they (hypothetically) appear in models. Similarly, we see how specific opinions in models (critique of Chamorro) are related to more general attitudes and ideologies (e.g., anti-communist ones), and that in discourse these relationships typically take the form of (implied) steps in argumentation.

[2] On that day an explosion in Managua brought to general knowledge an arms cache and international-terrorism center that the Sandinistas had set up when they governed and had then hidden from their successors.
The second sentence details this initial summary by specifying the nature of the (literal) explosion after its metaphorical introduction in the first sentence. There are several strategies in the formulation of opinion here. First the use of the predicate "brought to general knowledge" presupposes a complex history of U.S. 'revelations' about the role of the Sandinistas in international terrorism, especially in Latin America. Specifically, the use of this phrase implies that at least some people (and the WP in particular) already knew about this connection all the time, but that not everybody would believe them. That is, the choice of the phrase is part of a strategy of political vindication and positive self-presentation .

At the same time, we here have a nice example of the partisan and relative use of the notion of 'knowledge' discussed earlier in this paper, viz., the distinction between private and public beliefs, and the implication that what we believe to be true is knowledge, and not simply a belief. Also the verb tenses ("had set up", "had hidden") suggest that this is (knowledge about) fact, rather than belief. Note also that a fragment of the context model is being expressed here, viz., the suggestion that the WP already knew something others ("general knowledge") did not yet know. Similarly, also the political function of this suggestion (vindication) can only be explained when the context model is being made explicit in this way.

In other words, the explosion is taken as proof and incontrovertible evidence for (our knowledge about) the terrorist intentions of the Sandinistas, as also the lexical choice and the hyperbole "international terrorism-center" suggests. That is, what is a cache of arms is exaggerated as forming an international terrorist center, thus allegedly proving the earlier accusations of the Reagan administration (and the WP) to be correct after all. We see that the presentation and description of the current event is more than just a summary of the events, but at the same time an argument (viz., presentation of 'evidence') for the implied editorial opinion that Sandinistas were terrorists after all (which in turn implies that U.S. intervention, viz., by supporting the Contras and later Violeta Chamorro) was legitimate after all. In terms of our theoretical framework this means that the discursive 'doing' of vindication, consists in using a current model of events in order to prove that old models were true.

That the Sandinistas still can't be trusted is a further opinion, implied by the phrase "hidden from their successors", which also explains why the Chamorro's gamble to cooperate with such an unreliable group is indeed risky, and (we may expect in the rest of the editorial) to be condemned.

This analysis also shows that in order to make explicit the opinions of an editorial, we need to spell out rather extensive sets of knowledge in the forms of presuppositions and implications, in this case about the recent history of Nicaragua and the intervention of the U.S. in this country as well as about the Cold War and U.S. anticommunism in general.

[3] In the debris were a guerrilla arsenal with 19 surface-to-air missiles, documents detailing a kidnapping ring directed against Latin millionaires, and hundreds of false passports and identity papers.
Sentence [3] details the alleged contents of the secret Sandinista cache, and its very 'objective' description is primarily intended to support the accusation of 'terrorism' in the previous sentence: The Sandinistas not only appear to have been doing what they were accused of all the time, viz., supporting guerrillas elsewhere in Latin America (and with sophisticated arms at that), but even were engaged in true crime, viz., kidnapping of millionaires. Whereas the first accusation should worry the politicians, the second duly addresses the rich and corporate business. Interestingly, no explicit opinions are expressed here, but the very factuality of this description is the best possible support for the ideological, opinion-based accusation of terrorism. This shows again that opinions don't come alone. They need to be studied in their full contextual and co-textual functionality. Factual statements may thus function as 'objective' support in an otherwise evaluative discourse.
[4] The disclosure showed the final folly of Mrs. Chamorro's effort to bring in Sandinista Humberto Ortega as minister of defense in the name of 'reconciliation.'
After the (evaluative) Summary-of-Events category, sentence [4] opens the explicit Evaluation category of the editorial: Here the evaluative comment by the newspaper on the action of the protagonist ("showed the final folly") is formulated explicitly. In fact, in this case, the comment is not primarily directed at Mrs. Chamorro's present intention to do business with the Sandinistas, but focuses on an earlier phase of this cooperation, viz., to keep Humberto Ortega (brother of former president Ortega) as minister of defense. The evaluation expressed by the lexical entry "folly" is remarkable in the sense that Mrs. Chamorro's cooperation with the Sandinistas is not merely found wrong, inappropriate or politically disadvantageous. Rather, an element of irrationality is implied by the 'foolishness' to want to cooperate with Sandinistas: Those who expect trustworthiness from communists must be literally 'out of their minds'. The same is conveyed by the use of single quotes with "reconciliation". Since this is a positive notion in the domain of moral values, its association with the Sandinistas and peace in Nicaragua is possible only when the notion of 'canceled' or otherwise subverted, e.g. through irony or quotation marks. In other words, quotes may express an opinion in the sense that a positive notion may be suspended or disqualified (or vice versa, a negative one put into doubt).
[5] It also rendered politically untenable any further American support while Sandinistas wield unearned and unaccountable power.
Editorials in the mainstream U.S. press seldom comment on international events or events in other countries without detailing and suggesting U.S. foreign policies. Thus, sentence [5] may be read first of all as a simple account of current policy (no further support), but also legitimizes it as a warranted policy given what has been said before: U.S. support was not merely untenable, but it was right that it was not given. The argumentative support for this opinion is summarized in the adverbial clause ("while the Sandinistas..."). This very clause provides further reasons for the policy, viz., that the Sandinistas wield "unearned" and "unaccountable" power. The selection of these terms implies that the Sandinista power is undemocratic, an opinion which is in line with the overall attitude spelled out in this editorial about the Sandinistas. Historically and politically, this conclusion also suggests that not only the U.S. supported Mrs. Chamorro during the elections, but also that the U.S. should continue to influence the internal political situation in Nicaragua, even if its president found it politically necessary to deal with leaders of the biggest political party of the country, viz., that of the Sandinistas, and even if the Sandinistas, despite such U.S. meddling in the elections had peacefully given up their power.
[6] An angry Senate last month voted down virtually all forms of economic aid to Nicaragua.
The description of the political situation now changes focus entirely to the U.S. political scene, but again the description is not factual, but involves a qualification ("angry") which is an interpretation of the (emotional) state of mind of the Senate's decision, an interpretation that is consistent with the evaluation of the WP of the situation in Nicaragua. In other words, the description and interpretation of political events is consistent with their editorial evaluation. In this case, this means that the WP not only endorses the Senate's decision, but also legitimates and supports it with its own arguments. For the study of the relations between the press and politics this observation also suggests that political events (such as a Senate vote) are not merely described, positively or negatively evaluated, but may also be actively supported by providing further arguments for it for the public at large.
[7] A crisis of governance in Managua had already undermined Mrs. Chamorro's effectiveness and brought renewed armed strife.
The spotlight switches back to Nicaragua and Mrs. Chamorro, who is now described with the evaluative compounds "crisis of governance", and "undermined (...) effectiveness". Besides being very formal, these qualifications are also politically polite. Mrs. Chamorro is not explicit called a bad president, but these diplomatic euphemisms seem to imply just that.

All evaluation, however, is relative, and as soon as attitudes and ideologies are expressed, one should always ask "for whom?". Thus, here the question may be raised, given the historical record, for whom Mrs. Chamorro has become less 'effective'. Given the reactions of the Senate (and the WP), one might conclude that she has become less effective for the U.S.

Note that in the last clause of this sentence ("and brought renewed armed strife") may seem an objective description of the situation. However, in complex ways, also this fact-description presupposes and implies ideologically based opinions. As is the case for the hyperbolic description of rebels (or non-compliant governments) as "terrorists", the description of violence or civil war as "armed strife" is a standard media euphemism. In the same way as overplaying events may imply a rhetorically effective opinion, so is underplaying it. This would be the case when civil war would be the case here, as suggested by the use of "renewed", which presupposes, historically, reference to the earlier civil war set off by the U.S.-backed Contras. However, apart from discontent (e.g., because promises --e.g. land and credits for former Contras-- were not fulfilled) and some skirmishes, no such widespread fighting, let alone civil war was at stake at the moment. If that is true, the argumentative back-up of Mrs. Chamorro's "ineffectiveness" is based not so much on the facts, but on a suggestive reconstruction of the situation in Nicaragua, viz., that a new civil war seems to be going on, which (from a U.S. point of view) in turn would make U.S. intervention, viz., stopping aid, legitimate.

The point of our analysis here is, again, that even seemingly factual statements, when related with the rest of the text and political context, have implications of an evaluative nature. Moreover, such evaluations may be rhetorically 'modulated', viz., through hyperboles, euphemisms or indirect suggestions. Such rhetorical (re)formulations also show that one should indeed carefully distinguish between opinions and their discursive expressions. Although we have no way (here) of knowing, the WP editor(s) may well think Mrs. Chamorro a 'bad president', but because she is the president brought to power by the U.S., this cannot be said explicitly, so that a euphemism may become imperative. On the other hand, the editors may well know that the situation in Nicaragua is not as alarming ("international terrorist network", "armed strife") as suggested, but an objective, let alone a euphemistic account, would be politically inappropriate while being inconsistent with the evaluation by the political elites (Senate, president, etc.) of the situation in Nicaragua. If the WP would not support (and even encourage) anti-Nicaraguan decisions of Congress, it would have chosen a very different way of describing the situation there.

[8] Even many senators who had earlier argued for a negotiated rather than a punitive approach to human rights and property differences with Nicaragua supported the aid cutoff.
Newspaper discourse is not known for its carefully crafted coherence. The same is true for editorials, in which we also find a regular switch of protagonists, scenes or focus, as is the case here between Chamorro and Nicaragua, on the one hand, and the U.S. and Congress, on the other. This structural fact subtly suggests the close intermingling of the political situation in both countries. Thus, sentence [8] goes back to the Senate's decision, and provides a further (factual) support with the presupposition-laden use of "even", which indeed presupposes that one had not expected this decision from the more liberal senators. This in turn implies that the situation in Nicaragua must be really bad, an implied opinion that dovetails with the WP assessment of the situation.

Again, the evaluative words are formal and thus tend to be euphemistic: A "punitive approach" sounds less harsh than more common (and more negative) 'punishment' or 'retaliation'. The same is true on the other side, where "human rights and property differences" seem both formal and vague, because these phrases do not specify who is violating human rights (the Chamorro administration? the Sandinistas? the former Contras?), and whose property (former U.S. property?) is involved. The reaction of "even" the liberal senators however suggest that whatever "differences" (probably "of opinion') are involved, they must be incompatible with U.S. interests.

The evaluation of Mrs. Chamorro's "ineffectiveness" while being supported by these vague descriptions of the ongoing U.S.-Nicaraguan conflict is also being put in a clearer perspective with this sentence: She indeed appears no longer effective in protecting U.S. interests. That is, between the lines, the clauses and the words, the current editorial evaluation ties in with a complex historical and political analysis of the situation in Nicaragua, that is, with more complex U.S. attitudes and (imperialist) ideologies, pertaining e.g., to the right to economically (if not militarily) control the political situation in other countries.

[9] Now the issue moves to the House.

[10] It cannot support aid unless Nicaragua moves firmly to subordinate the military to civilian authority.

The next two sentences not only move back the focus again to the USA, but also from the immediate past (Senate vote) to the immediate future, as is the case in [9]. This change of tense and time, also marks the change from assessment and evaluation to the crucial category of editorial advice, as is also the case here in sentence [10] which is a barely veiled recommendation to the House not to vote for aid for Nicaragua. Instead of formulating such a recommendation with the modal "should not", another normative word is chosen, viz. "cannot", which is even more powerful, because it implies impossibility: They inevitably have to vote down such aid.

The "unless" clause backing up this indirect advice expresses a straightforward condition, already implied by the earlier part of the text: viz., the (Sandinista) military should be put under (civilian) control. The opinion expressed here is expressed loudly and clearly ("firmly"). This is possible because it instantiates a basic general opinion, viz., that the military should be subordinate to the civilian authorities, which is part of the democratic creed. Our historical knowledge however tells us a rather different story, viz., when the U.S. supported the Contras against the then civilian authorities of Nicaragua. That is, another reading of this crucial condition is not that democracy should prevail, but that Marxists should be opposed, which would be an interpretation of this editorial which would be consistent with the historical record as well as with the current situation: Mrs. Chamorro should be forced to become more "effective" in following U.S. interests (especially where "property differences" are involved).

[11] The Sandinistas still control the military and the intelligence operations, regarding these as arms not of state power but of their party's power.

[12] This profound abuse must end.

From its earlier indirectness and formal style, the editorial now openly starts to attack the Sandinistas, and adapts its style accordingly, even in its political advice ("profound abuse"). That is, the occasioning event (the disclosure of the arms cache) is now being used to evaluate more generally the situation in Nicaragua, and to make sure that the U.S. Congress uses its economic power to curb the little (military) power the Sandinistas still wield. Thus [11] begins with a seemingly factual description of the current situation, viz., that of Sandinista power, and implicitly condemns such power again by appealing to a violation of democratic norm ('the Army should serve the State and not a party'. In other words, for our analysis, it is interesting to observe that opinions need not at all be formulated explicitly (although they often do, as in the hyperbolic use of "profound abuse" in [12]): It may be sufficient to state-as-fact that someone violates a generally accepted norm.
[13] To end it will require Mrs. Chamorro to turn out those who have betrayed the national trust and to reach to the democratic parties that supported her election but then found themselves frozen out.

[14] A leader who has shown much bravery, she has some substantial achievements in reordering the economy and reducing the effects of a long civil war.

[15] But she has allowed a party to have an army, which it uses to maintain secret capabilities and to run a political police force.

[16] If she cannot make these changes, Nicaraguan democrats must find another way.

After the summary normative recommendation that this "profound abuse must end", the editorial apparently sets out to detail this opinion ("to end it will require...") with some advice to Mrs. Chamorro. However, the WP does not exactly seem to know how to do so either, because no detailed advice is coming forward, other than to "reach to the democratic parties" (implying again the opinion that the Sandinistas are not), and more ominously that if Mrs. Chamorro fails, "democrats must find another way" in [16]. Since Mrs. Chamorro is president and heads the administration, one wonders which democratic alternatives these 'democratic' parties might have. Again, knowledge of history provides for a rather more cynical interpretation of the 'democratic' stature of the parties that supported Mrs. Chamorro during the elections, some of which were responsible for the vicious Contra war which involved the brutal killings of innocent women and children, and the destruction of schools, health centers, and so on. On the other hand, the Sandinistas are of course negatively portrayed as having "betrayed the national trust", and are therefore explicitly evaluated again as unreliable, as all Marxists.

Note that in these evaluative propositions and their stylistic and rhetorical expression, we also find the familiar polarization, viz., between good (democratic, trust, etc.) US and undemocratic (betrayed, terrorist, etc.) THEM. The ingroup is of course first 'We-Americans', represented here by the U.S. Congress (described as doing well when cutting aid), and implicitly of course 'we-at-the-WP'. Then, in a next circle of those belonging to US are our friend and allies, viz., democratic (i.e. pro-U.S.) forces elsewhere, viz., Chamorro and the 'democratic' parties that supported her. However, our friends, especially when not "effective" enough to defend our interests (e.g. not rigorously oppressing the Sandinistas), may be criticized, although this will happen in a euphemistic and subdued, formal style. On the other hand, the description of the outgroup, viz., the Sandinistas, no words are minced, and they are described as violating all values we stand for (democracy, non-violence, subordination of the military, etc.).

The strategy to give advice to an ally or good friend who seems to be losing her "effectiveness" needs to be careful though. We may therefore expect the usual disclaimers, as in [14] detailing all the good things Mrs. Chamorro has done for her country ("bravery", "achievements", "reducing effects of the civil war") although slightly mitigated with "some" (achievements). However, since the editorial's aim is not to praise Mrs. Chamorro, such praise mainly has a strategic function, viz., to formulate criticism, usually introduced with but , as is also the case here (sentence [15]). That is, she is blamed for what is assessed as "a party having an army", which is a repetition of the accusation formulated in [11]. Indeed, as we have seen above, negative descriptions of the Other (here a violation of democratic principles), may be enhanced by repeating them.

Conditionals such as [16] may be read as threats. In this case, such an interpretation is hardly exaggerated, because U.S. Congress is precisely acting upon the negative compliance with the condition, and will cut aid if Mrs. Chamorro does not toe the line. Moreover, "Nicaraguan democrats" are now addressed directly, and implicitly advised to take things in their own hands, which also implies that Mrs. Chamorro is now no longer part of we-the-democrats, but compromised while not acting forcefully enough against the Sandinistas.

For our analysis, these observations show again that as soon as policies and advice to allies are involved opinions about who should do what are part of the normative structure underlying the editorial. This normative structure applied to the situation in the U.S., in Nicaragua and U.S.-Nicaraguan relationships provides the concrete advice being given in the editorial. The support for such recommendation is a structure of attitudes and values, viz., who is Good and who is Bad, and why.

[17] President Chamorro has welcomed international help in clamping down on any terrorist connections; discovery of fraudulently obtained Nicaraguan passports at the Brooklyn home of a World Trade Center bombing suspect naturally sharpened American interest.

[18] But cooperation with Interpol and the FBI does not go to the basic problem of civilian control of the military.

[19] Not at American bidding but to serve Nicaraguan democracy, this is where Managua must turn.

These final sentences partly repeat (as the second clause of [18]) what was said before, and partly spell out earlier propositions. Thus, [17] states that Mrs. Chamorro has done something well, viz., she opposes our common enemy, viz., terrorists, but as we may expect in such critique, [18] suggests that international cooperation is not enough ("does not go to the basic problem"), which is an explicit opinion, based on the value principle that all important jobs must be done well.

Sentence [17] finally brings in one missing clue in the argumentation: Why would the WP and the U.S. Congress be bothered about the situation in Nicaragua in the first place. The explicit use of "interest" here shows the 'connection', viz., between (fraudulent) Nicaraguan passports and the WTC bombing. It is not claimed, but given the earlier part of the editorial, strongly suggested, that these fraudulent passports might have come from the secrete Sandinista "international-terrorism center". We now also understand why this cache, which might be seen primarily as serving the Sandinistas themselves, or maybe guerrillas in other Central American countries, is associated with "international terrorism", since it is not the Sandinistas that can legitimately be accused of threatening the U.S., but as supporters of "terrorism", and hence of bombings in the U.S., a clear connection is suggested, and hence U.S. intervention legitimated.

That the WP editors feel they are doing just that, viz., legitimating, supporting and arguing for another form of U.S. intervention in Nicaragua is confirmed by the last sentence, which is the democratic disclaimer that protects them against an accusation of condoning intervention and imperialism: Chamorro should act not for the U.S. but for Nicaraguan democracy. That is, in this way the editorial of the WP saves its democratic face, by the altruistic claim that of course foreign presidents should not act because the U.S. tells them so. Such a final disclaimer is especially necessary after the earlier support of a congressional decision to cut aid, to put pressure on the Nicaraguan administration, and to encourage other "democrats" in Nicaragua to take over from a less "effective" Mrs. Chamorro.

These readings are no arbitrary interpretations. Each proposition commented upon may be connected to others that form a complex pattern of opinions, attitudes and ideologies, viz., attitudes and ideologies about democracy, Marxism, international aid, Congressional action, the subordination of the military, and so on. Some of the sentences of this editorial express these opinions openly and explicitly. This is especially the case for the negative description of the outgroup, viz., the Sandinistas. For the characterization of some members of our own group, viz., Mrs. Chamorro, viz., our erstwhile ally, a more indirect and implicit style and rhetoric is chosen in the expression of the (otherwise) clear opinions and moral recommendations.

Concluding remark 

Thus, we conclude first that many opinions of editorials are not expressed explicitly, and that the degree of explicitness depends not only on the nature of the opinions themselves, but on the context (or rather context model), viz., the role and political position of the Washington Post as a conservative newspaper, its relation with Congress, and hence its relation to a U.S. ally and (female) president. This alone allows for very complex relationships between ideologies and concrete opinions, whose textual formulation must be managed with care, so as not to offend Congress or an allied, conservative foreign president.

Secondly, we have found that opinions may not only be expressed implicitly, but also be implied indirectly by specific factual statements. Such statements may as such be true, but have negative consequences, and presuppose negative opinions of those who are involved. Indeed, mentioning some 'facts' (e.g., about Sandinista control over the army) and not other 'facts' (earlier U.S. intervention in the elections, U.S, support of Chamorro, the backing of the murderous Contra war by the "democrats" appealed to here, etc. etc.), is one way of expressing opinions. Or more generally, what this analysis shows is how a model and its opinions is undergoing several processes (selection, emphasis, de-emphasis, etc.) before its propositions are 'fed' to the semantic representation construed during text production, and how such processes are a function of the (model of the context), featuring, e.g., the identification of the WP with a conservative view of Nicaragua and U.S. foreign policy.

In other words, what this editorial is doing is expressing the stylistic, rhetorical and argumentative devices that strongly suggest readers to construct a preferred model of the current event (the finding of the cache) in particular, and of the situation in Nicaragua, in general. In such a preferred model we thus find a focus on some preferred facts, viz., positive ones in which we are involved, and negative ones in which the Others are involved. In this case, the anti-marxist crusade against the Sandinistas may seem ludicrous, and is therefore not emphasized, although not absent ("Marxist"). However, more effective, is to explicitly associate the Sandinistas with international terrorism, and hence clearly evaluate them as a threat to the U.S.A.

Once this preferred model is strategically expressed and construed by the readers, together with the implicit and explicit opinions that mark who is good and bad, and what actions were good and bad, another model is being constructed, viz., a contextual one, in which the WP is seen as legitimating a decision of the Senate, as recommending one to the House, as well as to Mrs. Chamorro. That is, having described and evaluated the situation in the rest of the text, the WP now takes the contextual, pragmatic step and draws practical conclusions about what should be done -- It not only expresses an opinion or conveys a biased event model, but engages itself in an illocutionary act of making strong recommendations, or even veiled threats ("if she cannot make these changes..."). It is this speech act itself that places the WP as an actor in the complex field of international policy, viz., of a long tradition of U.S. intervention in Central America. The final disclaimer of the editorial precisely confirms the strategic way such editorial intervention in international policies and politics takes shape.
 

REFERENCES

Abelson, R. P. (1973). The structure of belief systems. In R. C. Schank, & K. M. Colby (Eds.). Computer models of thought and language, (pp. 287-340). San Francisco: Freeman.

Abelson, R. P. (1976). Script processing in attitude formation and decision making. In J. S. Carroll, & J. W. Payne (Eds.), Cognition and social behavior, (pp. 33-46). Hillsdale, NJ: Erlbaum.

Abercrombie, N., Hill, S., & Turner, B. S. (1980). The dominant ideology thesis. London: Allen & Unwin.

Abercrombie, N., Hill, S., & Turner, B. S. (Eds.). (1990). Dominant ideologies. London: Unwin Hyman.

Altschull, J. H. (1984). Agents of power: the role of the news media in human affairs. New York: Longman.

Alvarado, S. J. (1990). Understanding editorial text. Kluwer Acad. Pubs. Group.

Antaki, C. (1994). Explaining and arguing. The social organization of accounts. London: Sage.

Arkes, H. R., & Hammond, K. R. (Eds.). (1986). Judgment and decision making: an interdisciplinary reader. Cambridge: Cambridge University Press.

Bagdikian, B. H. (1968). Editorial responsibility in times of urban disorder. In C. V. Daly (Ed.), The media and the cities, (pp. 13-24). Chicago, IL: Chicago University Center of Policy Study.

Bagdikian, B. H. (1983). The media monopoly. Boston, MA: Beacon Press.

Bar-Tal, D. (1990). Causes and Consequences of Delegitimization: Models of Conflict and Ethnocentrism. Journal of Social Issues, 46(1), 65-81.

Bar-Tal, D. (1990). Group beliefs: a conception for analyzing group structure, processes, and behavior. New York: Springer-Verlag.

Bar-Tal, D., & Kruglanski, A. W. (Eds.). (1988). The social psychology of knowledge. Cambridge: Cambridge University Press.

Baumgartner, P., & Payr, S. (1995). Speaking minds. Interviews with twenty eminent cognitive scientists. Princeton, NJ: Princeton University Press.

Bennett, W. L. (1980). Public Opinion in American Politics. New York: Harcourt Brace Jovanovich.

Billig, M. (1982). Ideology and Social Psychology. Oxford: Basil Blackwell.

Billig, M. (1987). Arguing and Thinking: a rhetorical approach to social psychology. Cambridge: Cambridge University Press.

Billig, M. (1991). Ideology and Opinions: studies in rhetorical psychology. London: Sage.

Bogart, L. (1985). Polls and the awareness of public opinion. New Brunswick, NJ: Transaction Books.

Bolivar, A. (1994). The Structure of Newspaper Editorials. In Coulthard, Malcolm (Ed.), Advances in Written Text Analysis. (pp 276-294). London, England: Routledge.

Bonnafous, S., & Fiala, P. (1984). L'argumentation dans les editoriaux de droite et d'extreme droite (1973)-(1982). Paris: Institut de langue francaise (CNRS) and Ecole Normale Superieure de Saint Cloud.

Breakwell, G. M., & Canter, D. V. (Eds.). (1993). Empirical approaches to social representations. Oxford: Clarendon.

Cacioppo, J. T., Petty, R. E., & Sidera, J. A. (1982). The effects of a salient self-schema on the evaluation of proattitudinal editorials: Top-down versus bottom-up message processing. Journal of Experimental Social Psychology, 18, 324-338.

Chang, T. K. (1989). The Impact of Presidential Statements on Press Editorials Regarding United-States China Policy, 1950-1984. Communication Research, 16(4), 486-509.

Ckomsky, N. (1985). Turning the tide. U. S. intervention in Central America and the struggle for peace. Boston: South End Press.

Chomsky, N. (1987). On power and ideology: the Managua lectures. Boston: South End Press.

Chomsky, N. (1989). Necessary illusions: thought control in democratic societies. South End Press.

Chomsky, N. (1992). Deterring democracy. London: Vintage.

Chomsky, N. (1993). Year 501. The conquest continues. London: Verso.

Clancey, W. J., Smoliar, S. W., & Stefik, M. J. (Eds.). (1994). Contemplating mind. A forum for artificial intelligence. Cambridge, MA: MIT Press.

Cohen, L. J. (1992). An essay on belief and acceptance. Clarendon Press; Oxford: Oxford University Press.

Crespi, I. (1989). Public opinion, polls, and democracy. Boulder: CO: Westview Press.

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

Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Orlando: Harcourt Brace Jovanovich.

Eisenberg, N., Reykowski, J., & Staub, E. (Eds.). (1989). Social and moral values: individual and societal perspectives. Hillsdale, NJ: Erlbaum.

Erikson, R. S., Luttbeg, N. R., & Tedin, K. L. (1988). American public opinion: its origins, content, and impact. New York: Macmillan.

Fairclough, N. L. (1995). Media Discourse. London: Edward Arnold.

Farr, R. M., & Moscovici, S. (Eds.). (1984). Social representations. Cambridge: Cambridge University Press.

Fasold, R., Yamada, H., Robinson, D., & Barish, S. (1990). The Language-Planning Effect of Newspaper Editorial Policy: Gender Differences in The Washington Post. Language in Society, 19(4), 521-539.

Fedler, F., Smith, R. F., & Counts, T. (1985). Voter uses and perceptions of editorial endorsements. Newspaper Research Journal, 6, 19-23.

Fischer, H. D., & Fischer, E. J. (Eds.). (1990). Political editorial, 1916-1988. Muenchen: Saur.

Fiske, S. T., & Taylor, S. E. (1991). Social cognition. (Second edition). New York: McGraw-Hill.

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

Fraser, C., & Gaskell, G. (Eds.). (1990). The social psychological study of widespread beliefs. New York, NY: Clarendon Press/Oxford University Press.

Frijda, N. (1987). The emotions. Cambridge: Cambridge University Press.

Graham, G. (1993). Philosophy of mind. An introduction. Oxford: Blackwell.

Gruner, C. R. (1989). A quasiexperimental study of the effect of humor preference and other variables on understanding/appreciation of editorial satire. Psychological Reports, 65(3,, Pt 1) 967-970.

Hackett, R. A., & Zhao, Y. Z. (1994). Challenging a Master Narrative: Peace Protest and Opinion Editorial Discourse in the United-States Press During the Gulf-War. Discourse & Society, 5(4), 509-541.

Harre, R., & Stearns, P. (Eds.). (1995). Discursive psychology in practice. London: Sage.

Herman, E. S., & Chomsky, N. (1988). Manufacturing consent. The political economy of the mass media. New York: Pantheon Books.

Himmelweit, H. T., & Gaskell, G. (Eds.). (1990). Societal psychology. London: Sage.

Hintikka, J. (1962). Knowledge and belief. An introduction to the Logic of the Two Notions. Ithaca, NY: Cornell University Press.

Holloway, H., & George, J. H. (1986). Public opinion: coalitions, elites, and masses. New York: St. Martin's Press.

Iyengar, S., & McGuire, W. J. (Eds.). (1993). Explorations in political psychology. Duke studies in political psychology. Durham, NC: Duke University Press.

Jaspars, J., & Fraser, C. (1984). Attitudes and social representations. In R. M. Farr, & S. Moscovici (Eds.), Social representations. 101-124. Cambridge: Cambridge University Press.

Johnson-Laird, P. N. (1983). Mental models. Cambridge: Cambridge University Press.

Kinder, D. R., & Sears, D. O. (1985). Public opinion and political action. In G. Lindzey, & E. Aronson (Eds.), The handbook of social psychology, 3rd ed., vol 2, pp, 659-741. New York: Random House.

Kornblith, H. (Ed.). (1994). Naturalizing epistemology. 2nd edition. Cambridge, MA: MIT Press.

Krueger, E., & Fox, J. D. (1991). The Effects of Editorials on Audience Reaction to Television Newscasters. Journalism Quarterly, 68(3), 402-411.

Kruglanski, A. W. (1989). Lay epistemics and human knowledge. New York: Plenum.

Larrain, J. (1979). The concept of ideology. London: Hutchinson.

Lee, M. A., & Solomon, N. (1990). Unreliable sources. A guide to detecting bias in news media. Carol Pub. Group.

Lehrer, K. (1990). Theory of knowledge. London: Routledge.

Lichter, S. R., Rothman, S., & Lichter, L. (1990). The media elite. America's new powerbrokers. New York: Hastings House.

Lippmann, W. (1922). Public opinion. New York: Harcourt Brace and World.

Love, A., & Morrison, A. (1989). Readers' Obligations: An Examination of Some Features of Zimbabwean Newspaper Editorials. ELR Journal, 3, 137-172.

Martin, L. L., & Tesser, A. (Eds.). (1992). The construction of social judgments. Hillsdale, NJ: Erlbaum.

Merron, J., & Gaddy, G. D. (1986). Editorial Endorsements and News Play: Bias in Coverage of Ferraro Finances. Journalism Quarterly, 63, 127-137.

Milburn, M. A. (1991). Persuasion and politics: The social psychology of public opinion. Pacific Grove, CA: Brooks/Cole Publishing Co.issues]

Moravcsik, J. M. E. (1990). Thought and language. London: Routledge.

Myers, D. S. (1982a). Editorials and foreign affairs in recent presidential campaigns. Journalism Quarterly, 59, 541-547.

Myers, D. S. (1982b). Editorials on the economy in the 1980 presidential campaign. Journalism Quarterly, 59, 414-419.

Noelle-Neumann, E. (1984). The spiral of silence: public opinion--our social skin. Chicago: University of Chicago Press.

Ortony, A., Clore, C. L., & Collins, A. M. (1988). The cognitive structure of emotions. New York: Cambridge University Press.

Oskamp, S. (1991). Attidudes and opinions. (2nd edition). Englewood Cliffs, NJ: Prentice-Hall.

Paletz, D. L., & Entman, R. M. (1981). Media, power, politics. New York: Free Press.

Pomerantz, A. (1984). Agreeing and disagreeing with assessments; Some features of preferred/dispreferred turn shapes. In J. M. Atkinson, & J. Heritage, (Eds.). Structures of social action, 57-101. Cambridge: Cambridge University Press.

Potter, J., & Wetherell, M. (1987). Discourse and social psychology: Beyond attitudes and behaviour. Beverly Hills, CA: Sage Publications, Inc.

Potter, J., & Wetherell, M. (1988). Accomplishing Attitudes: Fact and Evaluation in Racist Discourse. Text, 8, 1-2, 51-68.

Price, V. (1992). Public opinion. Newbury Park, CA: Sage.

Rokeach, M. (1973). The nature of human values. New York: Free Press.

Rokeach, M. (1979). Understanding Human Values: Individual and Societal. New York: Free Press.

Rumelhart, D. E., & McClelland, J. L. (1986). Parallel distributed processing: Explorations in the microstructure of cognition. 2 vols. Cambridge, MA: MIT Press.

Rystrom, K. (1994). The why, who, and how of the editorial page. Strata.

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

Searle, J. R. (1995). The construction of social reality. London: Penguin Books.

Sinclair, K. (1982). British press editorials on leadership during 1979 campaign. Journalism Quarterly, 59, 230-234.

Taylor-Gooby, P. (1985). Public opinion, ideology, and state welfare. London: Routledge & Kegan Paul.

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

Thompson, J. B. (1990). Ideology and modern culture: critical social theory in the era of mass communication. Stanford University Press.

Tirkkonen-Condit, S. (1987). Argumentation in English and Finnish editorials. In F. H. van Eemeren, R. Grootendorst, J. A. Blair, & C. A. Willard (Eds.), Argumentation across the lines of discipline. (pp. 373-387). Dordrecht: Foris.

Tulving, E. (1983). Elements of episodic memory. Oxford: Oxford University Press.

Van Dijk, T. A. (1987). Episodic models in discourse processing. In Rosalind Horowitz, & S. Jay Samuels (Eds.), Comprehending oral and written language. (pp. 161-196). San Diego, CA: Academic Press.

Van Dijk, T. A. (1988a). News Analysis. Case studies of international and national news in the press. Hillsdale, NJ: Erlbaum, 1988.

Van Dijk, T. A. (1988b). News as discourse. Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.

Van Dijk, T. A. (1989). Race, riots and the press. An analysis of editorials in the British press about the 1985 disorders. Gazette, 43, 229-253.

Van Dijk, T. A. (1991). Racism and the press. London: Routledge.

Van Dijk, T. A. (1992). Racism and argumentation. In F. H. van Eemeren, et al. (Eds.) Proceedings of the Second Int. Congress of Argumentation. Dordrecht: Foris.

Van Dijk, T. A. (1995a). Discourse semantics and ideology. Discourse & Society 6(2), 243-289.

Van Dijk, T. A. (1995b). Power and the news media. In D. Paletz (Ed.), Political Communication and Action. (pp. 9-36). Cresskill, NJ: Hampton Press.

Van Dijk, T. A. (1996a). Context models and text processing. In M. Stamenow (Ed.). Cognition and consciousness. (In press.)

Van Dijk, T. A. (1996b). Discourse, power and access. In C. R. Caldas-Coulthard and M. Coulthard (Eds.), Texts and Practices. Readings in Critical Discourse Analysis. (pp. 84-104). London: Routledge.

Van Dijk, T. A. (1996c). Discourse, racism and ideology. (in press).

Van Dijk, T. A. (Ed.). (1985). Discourse and communication. Berlin/New York: de Gruyter.

Van Dijk, T. A., & Kintsch, W. (1983). Strategies of discourse comprehension. New York: Academic Press.

Van Eemeren, F. H., & Grootendorst, R. (1992). Argumentation, communication, and fallacies. Hillsdale, NJ: Erlbaum.

Van Oostendorp, H., & Zwaan, R. A. (Eds.). (1994). Naturalistic Text Comprehension. Norwood, NJ: Ablex.

Wegman, C. (1981). Conceptual representations of belief systems. Journal for the theory of social behavior 11, 279-305.

Wegman, C. (1994). Factual Argumentation in Private Opinions: Effects of Rhetorical Context and Involvement. Text, 14(2), 287-312

Wyer, R. S., & Srull, T. K. (1989). Memory and cognition in its social context. Hillsdale, NJ: Erlbaum.

Wyer, R. S., & Srull, T. K. (Eds.). (1984). Handbook of social cognition. 3 vols. Hillsdale, NJ: Erlbaum.

Zaller, J. R. (1992). The nature and origins of mass opinion. New York, NY: Cambridge University Press.

-----------------