Paper CDA Congress Murcia, September 2004


Macro Contexts


Teun A. van Dijk

Universitat Pompeu Fabra, Barcelona


Second draft, May 2005





In linguistics, discourse analysis and most of the humanities and social sciences it has become a truism that all phenomena are understood and hence should be explained as part of their ‘context’. It is not surprising, therefore, that there are thousands of books that feature the concept of ‘context’ in their titles. Despite this vast amount of ‘contextual’ studies, however, there is not a single monograph that provides an explicit theory of context¸ although there edited books on context (e.g., Duranti & Goodwin, 1992). This means that the notion is commonly used in a more or less informal way, namely to refer to the explanatory situation or environment of some phenomenon, that is, its conditions and consequences.

This is also true in linguistics and discourse analysis, where the notion of ‘context’ should be made explicit for many reasons, if only because of its etymological meaning as an environment of ‘text’: con-text. This may mean, on the one hand, the verbal context of words or sentences, that is structures of text or talk, and on the other hand the social situation in which a communicative event takes place. In discourse analysis, verbal context needs no special treatment, since it is precisely one of the major aims of discourse analysis to study the discursive ‘surroundings’ of words and sentences, for instance in the study of coherence, co-reference, anaphora, and so on.

The other notion of context, accounting for the situatedness of talk or text, however, still needs explicit theorizing, for instance in terms of settings, relevant properties of participants and their social identities, roles and relations, as well as the social actions performed in such a situation. In systemic functional linguistics, there has been sustained interest in the notion of ‘context of situation’ for a long time (see, e.g., Ghadessy, 1999; Leckie-Tarry, 1995), articulated in terms of the notions of ‘field’, ‘tenor’ and ‘mode’. However, I have shown that the SFL approach to context is theoretically inadequate (Van Dijk, 2003).

One of the theoretical difficulties of a theory of context formulated in terms of social situations is how to avoid having to introduce all properties of such situation, and not only those that are relevant for the discourse. Also for this reason, there are directions in discourse and conversation analysis that are reluctant to introduce social context in the description of text or talk, especially if such context is construed by the analyst, rather than explicitly oriented to by the participants themselves (Schegloff, 1987, 1992a, 1997). A point of debate in this case is what exactly counts as ‘being oriented to’, in the same way as the more general notion of ‘relevance’ (see the debate between Schegloff, Wetherell and Billig in Discourse & Society)

In recent work, I have repeatedly argued that an explicit theory of context cannot and should not be accounted for only in terms of the properties of the communicative or interactional situation (Van Dijk, 1999, 2001, 2003). Situations do not directly condition discourse structures. Nor do discourse structures directly influence situations, for that matter. If that would be so, all people in the same situation would talk or write in the same way. Also, such a context theory would be deterministic or probabilistic: some social event would in that case (more or less probably) ‘cause’ specific discourse properties.

The situation-discourse relation is necessarily indirect, and established by the participants. More specifically, the interface is cognitive: It is the way participants understand and represent the social situation that influences discourse structures. According to contemporary cognitive psychology, I assume that such representations take the form of mental models, stored in episodic memory, as is the case for all mental models of specific events and situations (Johnson-Laird, 1983; Van Dijk & Kintsch, 1983; Van Oostendorp & Goldman, 1999). Mental models represent what we informally call ‘experiences’, and communicative events in which we participate are just one type of everyday experience. Since mental models have theoretically and experimentally been shown to be crucial in discourse production and understanding, these ‘context models’ are able to explain many of the relations between discourse and social situations: they explain how participants orient to aspects of the current communicative situation.

Context models account, first of all, for the notion of relevance: whatever is construed as part of the context model is by definition relevant. Indeed, context models provide an explicit theory of relevance. Secondly, context models are subjective – they depend on the previous experiences, including previous discourses, of participants. They show how and why some situational properties may be relevant for some participants, but less so, or not at all, for other participants, or for the same participant at other moments. Thirdly, mental models are ongoingly and strategically constructed and modified, and hence account for the dynamic nature of an ever changing context throughout text or talk. In this way, context models flexibly control many aspects of discourse production and understanding.

One of the issues, however, not yet accounted for in such a theory is the ‘scope’ of the context models. We have assumed that models subjectively represent social situations. But this raises the problem of the definition of a ‘social situation’. If we limit such situations to typical everyday conversations, interaction or other face-to-face encounters, we may define them in terms of a setting of place and time, participants and their properties (including identities, roles, relations, aims, knowledge, etc.) and their actions. This is how much contextualization studies thus far have informally conceptualized the situatedness of discourse (Auer & Di Luzio, 1992; Brown & Fraser, 1979; Duranti & Goodwin, 1992; see Van Dijk, 1999; for further references).

However, as soon as we take different types of text and talk, and especially specialized, institutional or professional discourse, then we might want to include some relevant aspect of the institution and its properties in the model. That is, we may want to ‘situate’ medical discourse in a hospital, educational discourse or conversations in schools or universities, and political discourses in parliament, for instance (see, among many other references, Drew & Heritage, 1992; Drew & Sorjonen, 1997; Sarangi & Roberts, 1999; Thornborrow, 2002) . And when we include the social identities of the participants, such as gender, race or class, then we have another important category that requires to be included in a context model (Bucholtz, Liang & Sutton, 1999; Davis, 1993; Fischer & Todd, 1988; Kotthoff & Wodak, 1997; Van Dijk, 1984, 1987, 1991; 1993; Wodak, 1997). The same is true for group relations, e.g., of power or competition (Corson, 1995; Fairclough, 1989; Wodak, 1989). Before we know it we are including as possibly ‘relevant’ a host of other social conditions of communicative events – conditions that demonstrably influence what people say and how they say it. In other words, we would need to include much of the ‘variables’ that have been studied in sociolinguistics for years.

However, in the same way as behaviorist and other older theories of language did not want to study ‘meaning’ because this would imply a study of the whole world, also such a theory of context involving all relevant micro and macro properties of social situations, would soon mean a virtual explosion of possibly relevant conditions, and hence an unmanageable context model.

Especially when we think of the broader social, political and cultural conditions of discourse, the notion of context of culture has been used, e.g., in the tradition of British empiricism as it gave rise to systemic linguistics: Malinowski, Firth and Halliday (see, e.g., Eggins, 1994). Such context of culture was usually distinguished from the ‘context of situation’, as described above. However, it is not so easy to establish such a distinction. For instance, the social identity and institution of the participants might be seen as situational or more broadly sociocultural. Here we also touch upon the well-known but problematic distinction between micro and macro structures or levels of description in the social sciences (Alexander, 1987; Huber, 1991; Knorr-Cetina & Cicourel, 1981).

It is my aim in this paper to theoretically account for the concept of this more ‘global’ social, political or cultural context, how it differs from situational context and how such a notion can be accounted for in terms of the theory of context models. It obviously cannot be the aim of this paper to study the myriad of societal influences on discourse, as they have been studied in a wealth of literature. We only want to theoretically account for the fact that such (indirect) influences are possible in the first place, namely through context models.




Context Models



Context models are subjective representations of communicative situations. As all other models, they are assumed to be stored in episodic memory, that is, together with all our other personal experiences. Indeed, communicative events are experiences like other events, and may be remembered or forgotten as such. In other words, if our personal experiences are represented as episodic models, context models are a special type of such models, namely those in which at least one of the ongoing actions is discursive.

Context models strategically control discourse processing, in such a way that a discourse is produced or understood that is appropriate in a given communicative situation. This means that anything that can vary in discourse may thus become controlled by the context model, such as deictic expressions, politeness formulas, style, rhetorical structures, speech acts, and so on.

As suggested above, context models subjectively represent the properties of the communicative situation that are relevant for each participant. This means that if context models are not well coordinated, communication problems and conflict may arise, for instance when the participants represent each other in terms of identities or roles, or with aims and beliefs that are different from those represented by the other. In most informal studies on context, thus, we find various categories, such as Setting (Time, Place), Participants (and their group identities, roles and relations), and Actions, since the contents of these categories are usually relevant for discourse. These categories form a more or less fixed schema that allows participants to analyze and represent a potentially infinite number of social situations. Such an analysis needs to take place in matters of seconds, given the fast changing situational circumstances of language users. This means that context models need to be relatively simple and not too big.

Cognitively speaking, language users cannot possibly handle vast and complex situation models, if only because of the limitations of working memory (or perhaps some form of intermediate control memory; see Kintsch, 1998). During talk or text, they need to keep track of changes in the situation, and when relevant represent them and thus update their context model, as they also do with ‘semantic’ models of the events discourse refers to (Morrow, Bower & Greenspan, 1989).

Whereas the identities and relationships of participants may be relatively stable during interaction (although also negotiation of identities and relations may take place; see, e.g., Firth, 1995), at least one contextual dimension is constantly changing, so that context models need to be updated ongoingly: their knowledge, if only on the basis of the meanings or information (and their inferences) of the previous segments of the current discourse.

Since context models need to be relatively simple in order to be able to fit memory and in order to feed the information of the control processes of discourse production and understanding, there is no obvious way that they can also deal with vast amounts of social or cultural characteristics of the current ‘global’ situation.

However, if on the other hand it is also true that speakers and their discourses are traditionally said to be controlled by social structure or culture, then we need to find out how such information may fit the context model anyway.

In order to solve that theoretical dilemma, should we perhaps add or integrate some kind of macro model representing the relevant social or cultural structures to the ‘micro’ context model postulated above? And if so, how can we do so within the space and process limitations of controlling context models? Secondly, whereas we may have relatively simple schematic representations of situations, no obvious representation format seems to be available for the social and cultural (macro) structures that form the environment of interaction and communication: indeed, how to represent institutions, groups, power relations, and so on, in a few categories? And thirdly, if we need to select the relevant aspects of such social and cultural environments, how are language users able to do such a ‘social’ analysis of situations, and do this in real time, that is, immediately previous to, or during text or talk?  Thus, if professors and students interact, how and how much knowledge do they need to activate about their mutual identities, roles and relations; how much about universities, teaching and learning; how much about their mutual age groups, and so on? The same is true for all other text and talk, e.g., in families, on the job, or in service encounters, among many other settings and genres. Moreover, each culture has different norms and rules of interaction, such as norms of politeness, which would need special cultural conditions of speech acts, text or talk. Again, the number and variety of such potentially relevant social and cultural conditions is vast – and they add to the properties of situations (participants, etc.) mentioned above. Obviously, in order to accommodate even the potentially relevant properties of ‘global’ situations, we need a device that is able to do so within the constraints of realistic mental models of human participants of communication.




Macro Context Models



In order to answer the various questions above let us speculate a moment on ‘macro’ context models. At the moment it does not matter whether these are separate macro models, or whether ‘macro information’ is simply part of the context model as defined above. However, whatever our ideas on the ways participants bring to bear their ‘understanding’ of society and culture in their interaction and discourse, the format must be compatible with the usual memory and other constraints of discourse processing. That is, whatever sociocultural information is discursively relevant, it must be organized in a relatively simple way, or must be easily retrievable. So let us construct macro models theoretically and see where they lead us as feasible control structures for discourse.



Macro categories


We have assumed that any kind of macro model is manageable only if it has a relatively simple structure -- just like situational context (micro) models -- and if such a structure consists of a schematically organized categories. What are these categories? One theoretical heuristic to find these categories is by comparison to the structure of situations, but instead of the micro categories (setting, participants) we now use the corresponding macro-categories, e.g., as follows:


Micro                                               Macro

Setting                                              Macro Setting                        

interaction time                                 period (days, months years)

location                                   space (city, country)               

Participants: persons                         Participants: groups, institutions, organizations

Identities: professor                          Identities, e.g., ethnic group, school

Roles: teach                             Roles, e.g., education

Relations: personal power                  Relations, e.g., institutional power

         Aims                                       Group goals

         Personal knowledge                           Group – social knowledge  

Action, e.g., explain                           Macro act of group, institution: educate, etc.


Let us examine these potential ‘macro’ categories, in some more detail. Thus, by way of example, let us imagine first the (macro) context model of a journalist writing a news report for the press. She first of all represents the current Setting, as here and now, the place of writing; with herself in her roles of, e.g., writer, journalist, correspondent of newspaper X, with the aim to write a news report, based on her professional knowledge about an assassination. Now, what kind of macro-information about the social environment does she need in order to be able to write an appropriate news report on such a topic?



Macro Setting


Obviously, first of all, the Setting information available to her is not limited to the here and now of her writing, at home or in the office. She most likely also ‘knows’ what city and country she is writing in, as well as the date and the year. She may not have such knowledge activated and present in working memory at all moments, but it is ‘situational’ knowledge that is immediately accessible when needed, for instance in some kind of control memory (Kintsch, 1998). Indeed, she may need to signal such information in the dateline of the news article, as well as when using deictic expressions such as ‘this year’, ‘last month’, or ‘in this city’, and so on, for instance when comparing to other assassinations that have taken place in this city or country and during this year. News discourse, as well as many other discourse genres, make extensive use of reference to the places and periods of events, and these also require contextualization with respect to the place and time frame of the speaker or writer. Such ‘settings’ are not merely ‘local’ (here and now), but also more ‘global’ (this city or country; this week or this year).

Context model information is not limited to discursive situations, but more generally derives from the ongoing ‘experience models’ people construct and update each moment during the day (Van Dijk, 1999). Thus, often the knowledge about day, month and year, as well as about city or country, among other ‘global’ Setting properties, people are already aware of before engaging in verbal interaction. As is the case for other macro context information, such information may be cognitively backgrounded, and only be (re)activated when it becomes relevant, for instance when setting information is referred to or presupposed in the discourse.

Some of this global information may be more accessible than other information: We usually have direct access to the information about the city or country we are in (except in the humoristic situation of U.S. citizens traveling in Europe and deciding they are in Amsterdam because today is Tuesday). However, the day of the week and the date are usually harder to activate, and sometimes need to be ‘calculated’ on line or even by consulting and agenda or calendar.

From professionals writing ‘dated’ texts such as news reports, however, it may be expected they do have such information readily available during their work. The point is only to show that in many genres and types of interaction, participants use a model in which at least some global setting information is represented. Such information may even extend to such large periods as for instance the current period of the ‘present’ or ‘postmodernity’, when writing about such a period, or when writing about the ‘past’.

The same is true for location information – for instance when speaking or writing from another continent, and referring to ‘here in Europe’ or ‘here in South America’. Although place-and-time coordinates may be conceptualized differently in different cultures (Levinson, 1996), it seems plausible that all human beings represent at least some basic coordinates of time and place. Indeed, it might even be assumed that the representation of such a setting is a crucial component of consciousness, and would be a more or less serious disorder if people have no idea about the time and place where they are now. I shall assume that in many cultures this is not only true for basic coordinates, but also for such environmental knowledge as villages, cities or countries, and weeks, months or years, or similar ‘meso’ or ‘macro’ spatiotemporal’ categories. 



Macro participants


The same journalist from the previous example represents herself as the current writer of the current news report, as a journalist and maybe as a correspondent of a specific newspaper. Similarly, she represents the editors and readers of the newspaper as probable recipients. In both cases, her news report will probably feature expressions that signal such (micro) contextual information, such as a byline with her name, possibly with a description of her function (Correspondent in X, etc.). Reference to the readers in news is less common (but more common in editorials directly addressing the readers), but is standard in many other discourse genres, such as didactic texts addressing students, advertisements addressing clients or buyers, and of course in most forms of conversation.

The question is whether context models of participants also may or must feature information about macro ‘participants’, such as groups, institutions, organizations or nation states, or whether these are typical analyst’s constructs that are (often) irrelevant for the description of text or talk, as long as participants do not explicitly orient towards them (Schegloff, 1992b). In our example of the journalist, this is obviously the case: She knows for what newspaper she is writing, and will write her news report accordingly – she’ll tell a story in a very different way as when she tells the ‘same’ story to her friends or as a witness in court. Also, apart from the current communication medium’s self-description, such as the name of the newspaper on the cover page and possibly on other pages, the news report itself may self-refer to the newspaper as a participant or a writer, e.g., when a journalist writes something like “The witness told The Newspaper….”, where ‘The Newspaper’, is a metonymy referring to the journalist of The Newspaper.

Similarly, readers in their context models usually do not represent such and such a reporter as author of the current news reports – even when a news report is signed – but represent the author or source as the ‘newspaper’ or as ‘The Newspaper’. Reactivation of the context model in later accounts will typically lead to accounts such as “I read in The Newspaper…”.

Similarly, a journalist on the other hand does not have individual readers in mind, but probably a collectivity of readers of the newspaper. She will adapt the discourse to that collective, such as probable average knowledge about previous events (old models) and probable general, sociocultural knowledge (for these strategies of knowledge management, see Van Dijk, 2005). That is, the journalist writes primarily as a member of an organization (the newspaper) and a member of a profession (the journalists), and is able to do so only when representations of such collectivities are present as macro information in her context model.

The same is true for much organizational or institutional discourse, and this may be variously displayed in explicit collective authorship and corresponding deictic expressions (e.g., the editorials of the British tabloid The Sun are labeled “The Sun Says…”).

Such macro representations of agency in context models not only feature identity, but also information about roles and relationships for collective agents. Thus, the newspaper may discursively self-represent itself as opponent of the government or as critic of industrial environmental practices, as would typically be the case in editorials. This means that the writer of such an editorial also must construct a mental model in which such roles and identities for macro agents are represented as such.





Besides the social properties of collective agents, they also may be assigned cognitive properties, such as aims, beliefs, knowledge, attitudes and ideologies. Such shared beliefs will be multiply signaled in organizational or institutional discourse, as is the case in newspaper editorials, reports of NGOs, parliamentary decisions, or government discourse, among many other collective discourses.

As suggested before, prominent and important in context models is the role of knowledge, organized by a special K-device that strategically projects what recipients already know or what is still unknown to them, thus regulating the presuppositional structure of the discourse (Van Dijk, 2003, 2005). Knowledge is by definition certified shared belief, and hence as such characteristic of groups and communities. If journalists express or leave implicit some information that is presupposed to be known to the readers, they do so because they assume that ‘the readers’ as a collective already have this information.

Even if such knowledge presuppositions are regulated by the context models of actual writers-reporters, they are at the same time the shared knowledge of the newspaper as an institution or organization. In legal conflicts, it may be the newspaper, and not the individual journalist who may be prosecuted for publishing secrets or for slander. Journalists are explicitly trained to take into account these and other shared, collective or institutional responsibilities, as is the case for civil servants, business managers and many other professionals working for companies or institutions.

As required by the theory of contextual relevance, such macro model information must be signaled in discourse. This is indeed the case, as we may see in the epistemic structure of assertions-presuppositions in news reports, in which case ‘old information’ that might have been forgotten by ‘the readers’ may be reminded by such expressions as ‘as we reported last week’, or more explicitly as ‘as The Newspaper reported last week’, a previous report that might have been written by another journalist.

In everyday conversation, presupposition of previously communicated knowledge is part of personal context models: ‘I remember what I told you before, and now I remind you of that earlier conversation.’ In collective discourse, such a personal model need to feature also a macro level in which it is not merely the (current or old) context models of the individual journalists are at stake, but also the context models of the organization: journalists need to know and remember what the organization has ‘said’ before.


Macro Action


Although context models are undoubtedly more complex than summarized here, let us finally assume that one of their macro categories is about global actions. This category applies both to personal as well as to collective discourses. Thus, a professor may now self-represent her current act as ‘explaining a problem to the students’, or as ‘helping students’, but such an act will be part of the meso-level act of teaching a class or having a tutorial, which in turn may be part of the macro-level act of teaching, educating, and so on.

My contention is not that at each moment during the performance of base-level acts in interaction speakers are aware of these higher level acts, but that such may be the case as soon as aims, problems, conflicts or complications need to be formulated, expressed or discussed. That is, lower level acts will usually be represented as functional at other levels, and even during lower level acts agents may have more or less cognitively backgrounded consciousness of at least some higher level acts providing functional ‘reasons’ for what they are now doing. That is, the way I explain a problem to students is probably different from the way I explain a problem to friends, to the police or to my kids. That is, local acts are functionally variable within larger activities or social practices.

Such differences should also show up in text and talk. This is not only because of the representation of different participants as part of the context model, but also because of the higher level action representation in such models. Thus a current explanation to students will have a didactic style because of the overall representation of teaching. And a speech of a politician is necessarily adapted to the overall category of a ‘parliamentary debate’, and even more globally to the macro act of ‘legislation’ – and locally multiply signaled as such (Bayley, 2004).

The same is true not only for the macro acts of individual actors, but again also for the macro acts of collectivities, institutions or organizations, as we have seen for the example of newspaper news reports and editorials above. It is the newspaper that is ‘reporting’, ‘publishing’ or ‘attacking the government’. In each news report and editorial journalists knows that through this text they are accomplishing one of the macro acts of the organization, and such contextual knowledge is also discursively displayed by the action descriptions of media discourse (‘Yesterday we reported…’; ‘The Newspaper has always supported the ideas of party X, but…’, etc.).



Micro and macro


For a number of reasons I have modeled participant knowledge about social macro situations after the usual model of micro situations. This is not merely a useful heuristic device, but also might function as a feature of cognitive economy: the same conceptual, categorical structure is being used to represent local as well as global information. This is not a farfetched assumption, because such is also the case for other cognitive areas: agents may be personal or collectives or institutional, as is the case for actions, times and places, which are also discursively coded in the same way whether they are local or whether they are global.

This suggests that people probably do not have two different models, a micro and a macro model, of the same structure, but in fact one model with the categories mentioned above, but with information at various levels of specificity or generality as contents of the categories. Depending on the current situation, language users may foreground or background the global level information. They will typically foreground global information when locally it is necessary to explain, account for, motivate or resolve problems of local activities, e.g., in the following simulated account: ‘I report this event because it shows that this minister is incompetent, and it is the task of this newspaper to act as a watchdog of elected politicians.’

This level-dependent processing of situational information is not limited to contexts, but also well-known in discourse processing itself: language users are able to express and understand meaning at various levels between micro and macro structures (Van Dijk, 1980). This allows them especially to assign global coherence (and hence reduce complexity) to complex sequences of propositions.

The same function may be assigned to global context representations – they allow many ‘local’ settings, participants, actions, etc. to be subsumed under higher level, more abstract ones. Thus, whatever we do in the classroom, it may usually be subsumed as ‘teaching’. Whatever journalists do, their actions may be subsumed as, for instance, ‘reporting’. That is, instead of complicating contexts, global level categories may actually reduce the complexity of contexts, especially for complex sequences of action and interaction, as is the case for meetings, parliamentary debates, and so on.

The implications of this theory of macro context control are quite interesting. In the first place, we do not need to introduce new theoretical units, levels or schemas: Micro-macro processing takes place at all levels of discourse, interaction and information processing. People not only understand and plan complex discourse at global levels of meaning and form, and thus produce topics and themes, or schemas, that control local production and understanding. We now find that they do the same for the environment in which they act and speak: They need to reduce the vast complexity of such environments, and do so in the same way as when reading a news report or a novel: they use general categories of a schema and then represent local information at higher, macro-levels. In a way, thus, they ‘summarize’ the social environment at the macro levels of their context models. It is in this way that they are able to control the local situation variables and their information, and control discourse at the same time at the micro and macro levels, as explained above. Thus reduced, situational information may well fit the constraints of context models, as is also the case for the semantic macrostructures of discourse that are needed to produce or understand globally coherent discourse. Macro level information in context models in the same way control both the local and the global appropriateness of discourse.



Other social information?


The modeling of macro contexts on the basis of micro contexts and their de facto inclusion as higher level information in the categories of the context schema, might have blinded us for the fact that we might be overlooking a large amount of other relevant ‘social’ information that should also be represented in context models. Thus, journalists may self-represent themselves as such, as well as the newspaper as an institutional agent, and show such categorization in discourse. But what about all they know about journalists, newspapers, readers, and reporting – a vast knowledge base that they share with other professionals. Thus, if a journalists writes ‘In our last edition of yesterday, we reported that..’, does that mean that the knowledge about newspapers having different editions must be fed to the context model so as to be able to describe and explain the specific deictic expression ‘in our last edition’? The same is true for all journalists know about sources, press releases, interviewing, newsbeats, deadlines, editorial supervision and constraints, and so on (Gans, 1979; Tuchman, 1978).

These are all possible elements of news production, part of the professional knowledge journalists. It might be shown that much of such knowledge is relevant in news writing, for instance when citing people, using quotation marks and in general in the well-known news strategies of reporting discourse. Does this mean that all such knowledge is to be included in context models? If so, this would blow up the notion of context model to cognitively unmanageable proportions.

At the moment, we have no elegant solution to this problem. We need to recognize the cognitive constraint of quite limited, relatively simple context models as control mechanisms of discourse production and understanding. However, as soon as we reflect on the many conditions of discourse production, especially in professional or institutional situations, then we might not have enough when representing Setting, Participants, Cognition and Action as context schema categories. We have already seen that in each category we may have different levels of more or less local (micro) or global (macro) concepts. Now, we see that for each of these concepts (e.g., journalist, professor; newspaper, university, etc) we might need a vast amount of further information in order to account for the specificities of discourse production, e.g., about types of location and time periods (including all of history), types of agents and institutions and their relevant properties (e.g., that newspapers as organizations publish newspapers as medium) and actions.

In order to avoid cognitive overload of the context model, I shall assume that the schematic categories as proposed are filled with ‘summarizing’ information about the social situation (journalist, newspaper, etc), but that such contents are of course linked with the professional knowledge structures of participants. That is, when needed, journalists may activate and apply their professional knowledge about interviewing or editorial meetings. They do so, of course, during interviewing and during such meetings, as part of the overall institutional action of ‘reporting the news’. But when writing the news report, journalists need not have activated all knowledge about interviewing, and may simply represent the now most relevant information, namely what some source has said.

In other words, context models are also acting as the interface between discourse and knowledge. They not only regulate the knowledge needed to write about what the article is about (say Iraq), but the context model may also activate knowledge about the own social identity, profession, institution, professional acts, and so on. At each point in discourse processing, such information may be activated where relevant, such as the information that this newspaper has several editions when referring to such editions in current article. This means that the permanently relevant information (micro and macro) in context models may act as so many pointers to the vast professional knowledge base of participants. Permanently relevant is that a writer knows she is journalist when she writes for the newspapers. So such summarizing information (after all, she has lot of other knowledge about journalists) about current professional identity must part of the context model because it controls all activities. Also, that one is now writing for the newspaper. But not that the newspaper has several editions, because such knowledge does not control all aspects of the production of the current news report, not even as backgrounded context knowledge. In such cases, then, knowledge about production situations is simply activated and de-activated locally where needed.

In sum, we should leave context models as simple as possible, with a handful (7 plus or minus 2) of categories which themselves may each feature a handful (7 plus or minus 2) summarizing information units describing current Setting (time, location), participants, cognitions and actions, both at the local as well as the global level. Local level information is always active because it needs to control all current discourse structures, and global level information may be more or less backgrounded and activated when needed at certain (explanatory, meta, etc) moments in discourse production.

The same is obviously true for the context models of recipients, which also may represent the communicative event at various levels of specificity and generality.  

So far my argument was limited to the levels of various societal structures. However, macro contexts may be more far-reaching, and include general cultural information. The vast literature on intercultural communication and its possible conflicts seems to suggest that when context models are not culturally matched, communication problems may arise. This is a priori true for the specific cultural knowledge needed to produce and understand discourse, and which is a basis of all processing and interaction.

However, it may not only be true for our ‘knowledge of the world’, but also for our knowledge about communication and interaction and their norms and rules. That is, we may ill understand meanings of discourse in other cultures, but at the same time ignore many rules of interaction, such as those of politeness, turn taking or interruption, conversational postulates, genres or taboo topics, among a host of other communicative knowledge of other cultures (among hundred of other books on cross-cultural communication, see, e.g., Di Luzio, Günthner, & Orletti, 2001; Gudykunst, 2003).

Many of these properties can be represented in the categories postulated above. For instance in the category of Participants we may represent various identities, roles and relations between participants. Thus, we may and often should represent relationships of power and hierarchy, and with such general information contextual rules may constrain discourse in various ways: who speaks first, in what style, who controls topics, and so on.

The point is rather whether in different cultures we might have different basic categories in context models. Thus, whereas ‘we’ (in western and other cultures) might assume that communicative situations are ‘peopled’ by human participants, one could easily imagine that in some other cultures special categories might be needed to represent gods or other metaphysical participants or magic objects – and locations may need to be divided between everyday and sacred, and so on.

We may assume however that the basic principles are the same – namely that the context is schematic and relatively simple, and that its categories and their local and global contents control the variable structures discourse, both locally and globally.

It is an empirical matter to find out for each culture which are these basic categories, whether there are ‘universal’ categories, and how such contextual categories control which structures of text and talk – as when a power relation between participants controls the choice of personal pronouns, or when the knowledge of the speaker about the knowledge of the recipient controls the presupposition structure of the discourse. 



Concluding remark


From our theoretical discussion and from the brief and tentative list of some macro categories of context models, we may provisionally conclude that context models may also feature higher level situational information, for instance in terms of location, period, collective agents and their properties and actions. Such an account is not merely a familiar but problematic social distinction between micro and macro level description. Rather, we thus theoretically account for the bridging problem that has haunted social scientists since decades. That is, the relation between micro and macro is not just a question of levels of sociological description, and not just a question of conceptual inclusion (A is member of group B; action C is an instantiation of global action D; etc.), but rather how micro and macro can be explicitly related, namely through the representations of social members, for instance as context models.

This also implies that the level distinctions and the higher level categorizations are not merely the products of analysts, but genuine members’ categories and devices. Thus, journalists writing a news report are aware that they are producing the text of an organization, that they express shared professional knowledge of the organization, that they are writing for a collectivity of readers, and that they are at the same time realizing, locally, the global personal acts of working for a newspaper, and the global institutional acts of informing the public or criticizing the government. These are not merely analytical macro categories; these aspects of context models also locally control discourse production and understanding: the global categories are relevant for a large number of discourse properties. Indeed, people may explicitly self-describe such higher level categories, for instance in meta-commentary, e.g., when professors tell the students why they explain X or why students need to know X, for instance in terms of educational and learning goals.

Cognitively, the global categories may be less available than the local categories, but they need to be activated in order to be used at any moment as soon as representations of context categories of setting, participants, cognitions and actions are relevant at various levels. Such a foreground-background distinction is a quite general property of processing and memory.

Future work on macro context models needs to further develop the schematic categories needed to account for global situational understanding in different cultures. Comparative discourse studies will be needed to see whether not only the contents but also the very categories are different cultures. Another task will be to spell out the actual cognitive strategies involved in the production and management of context models and the activation and deactivation of global context information. Finally, as discourse analysts our main task is to show how exactly macro contexts influence the structures of text and talk.






Alexander, J. C. (Ed.). (1987). The Micro-macro link. Berkeley: University of California Press.


Auer, P., & Di Luzio, A. (Eds.). (1992). The contextualisation of language. Benjamins.


Bayley, P. (2004). Cross-cultural Perspectives on Parliamentary Discourse. Amsterdam: Benjamins.


Brown, P., & Fraser, C. (1979). Speech as a marker of situation. In K. R. Scherer, & H. Giles (Eds.). Social markers in speech, (pp. 33-62). Cambridge: Cambridge University Press.


Bucholtz, M., Liang, A. C., & Sutton, L. A. (Eds.). (1999). Reinventing identities. The gendered self in discourse. New York: Oxford University Press.


Corson, D. (Ed.). (1995). Discourse and power in educational organizations. Cresskill, N.J.: Hampton Press.


Davis, K. (Ed.). (1993). Negotiating at the margins: The gendered discourses of power and resistance. New Brunswick, N.J.: Rutgers University Press.


Di Luzio, A., Günthner, S., & Orletti, F. (2001). Culture in communication. Analyses of intercultural situations. Amsterdam Philadelphia: J. Benjamins.


Drew, P., & Heritage, J. (Eds.). (1992). Talk at work: Interaction in institutional settings. Cambridge (England New York: Cambridge University Press.


Drew, P., & Sorjonen, M. L. (1997). Institutional dialogue. In: Van Dijk, Teun A. (Ed.), Discourse as social interaction:  Discourse studies:  A multidisciplinary introduction, Vol. 2. (pp. 92-118). .


Duranti, A., & Goodwin, C., (Eds.). (1992). Rethinking Context: Language as an Interactive Phenomenon Cambridge: Cambridge University Press.


Fairclough, N. (1989). Language and power. London New York: Longman.


Fisher, S., & Todd, A. (Eds.). (1988). Gender and discourse: the power of talk. Norwood, NJ: Ablex Publishing Company.


Ghadessy, M. (Ed.). (1999). Text and context in functional linguistics. Amsterdam Philadelphia: J. Benjamins.


Huber, J. (1991). Macro-micro linkages in sociology. Newbury Park, Calif.: Sage Publications.


Knorr-Cetina, K., & Cicourel, A. V. (Eds.). (1981). Advances in social theory and methodology. Towards an integration of micro- and macrosociologies. London: Routledge & Kegan Paul.


Kotthoff, H., & Wodak, R. (Eds.). (1997). Communicating gender in context. Amsterdam (The Netherlands Philadelphia: J. Benjamin.


Leckie-Tarry, H. (1995). Language & Context. A functional linguistic theory of register. Edited by David Birch. London: Pinter.


Levinson, S. C. (1996). Language and space. Annual Review of Anthropology, 25, 353-382.


Morrow, D. G., Bower, G. H., & Greenspan, S. L. (1989). Updating situation models during narrative comprehension. Journal of Memory and Language, 28(3), 292-312.


Schegloff, E. A. (1987). Between Macro and Micro: Contexts and Other Connections. In J. Alexander, R. M. B. Giesen and N. Smelser (Eds.), The Micro-Macro Link (pp. 207-234). Berkeley: University of California Press.


Schegloff, E. A. (1992a). In Another Context. In Duranti, Alessandro, & Goodwin, Charles (Eds.), Rethinking Context: Language as an Interactive Phenomenon. (pp 191-227). Cambridge, England: Cambridge University Press.


Schegloff, E. A. (1992b). On talk and its institutional occasions. In: Drew, P., J. Heritage (Eds.), Talk at work: interaction in institutional settings. (pp. 101-34). Cambridge: Cambridge University Press.


Schegloff, E. A. (1997). Whose text? Whose context? Discourse & Society 8, 165-87.


Thornborrow, J. (2002). Power talk. Language ad interaction in institutional discourse. London: Longman.


Tuchman, G. (1978). Making news: A study in the construction of reality. New York: Free Press.


Van Dijk, T. A. (1980). Macrostructures: An interdisciplinary study of global structures in discourse, interaction, and cognition. Hillsdale, N.J.: L. Erlbaum Associates.


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


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


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


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


Wodak, R. (Ed.). (1989). Language, power, and ideology. Studies in political discourse. Amsterdam Philadelphia: J. Benjamins Co..


Wodak, R. (Ed.). (1997). Gender and discourse. London Thousand Oaks, Calif.: Sage Publications.