Dominik Lukeš University of East Anglia

The central argument of this paper is that discourse analysis must be conceptual before it can be critical. The construction of discourse is above all the construction of a conceptual discourse space (cf. Fauconnier, Werth). This space is patterned by conceptual devices such as frames (cognitive models), folk theories, metaphors, metonymies, etc. described in the tradition of cognitive linguistics (Lakoff, Johnson, i. a.). While the recently popular blending theory provides a very alluring hypothesis for the kind of cognitive operations functioning on the conceptual patterns in discourse, a practicing discourse analyst is still mostly left to rely on her intuition limiting her to very small parts of the discourse under investigation rather than having a reliable method of identifying the underlying conceptual patterns. So far, relatively little work has been done in this direction. There have been attempts to outline the types of textual representation or triggers for metaphor (Goatly, Steen, Semino, Heywood) but we have to go to the conversation analysis-inspired frame analysis for a more detailed account (Tannen). This paper first outlines the kinds of conceptual devices and patterns that might be available for a critical treatment. These patterns are then linked to their potential textual representations. It is argued that a successful critical analysis of discourse needs this manner of textual evidence to be able to determine the conceptual patterns underlying the texts it subjects to scrutiny. Only then can we start exposing “hidden” meanings to the possibility of contestation.


Chapter Eight

Really meaning: On criticism and analysis
It is one of the topoi of introductions to linguistics that the study of language is somewhat peculiar in that it uses language to describe language. Thus the term of metalanguage is introduced as a very early concept.1 While the significance of the language vs. metalanguage dichotomy can be easily overestimated, it is perhaps of more relevance to discourse analysis and, in particular, critically-orientated discourse analysis. The source of potential problems is two-fold. First, unlike in syntax or lexicology, discourse analysts engage with political meanings, and sources of potential bias are therefore amplified. Second, through their linguistic pedigree discourse analysts have a long tradition of positioning themselves in expert roles, using the relative obscurity of linguistic analysis as a convenient pedestal from which to launch their critiques. In Critical Linguistics, this goes as far back as Hodge and Kress’s [1978] use of the commuter as an example of a recipient more easily misled by transformations in text than a theory-equipped analyst. This shortcoming was pointed out in one of the early reviews (Sharrock and Anderson, ), as well as recently, in much greater detail, by Widdowson ()2. A further complication a would be expert (or meta) discourse analyst has to contend with is the fact that there is no shortage of willing folk-theoretical discourse analysts. Indeed, it seems that the ability to analyze discourse is just as part of the general language faculty as the ability to make predicates agree with subjects or vowels to come into harmony. Picking apart the various ways in which discourse can be not what it seems or can be moulded into shapes it was not first intended to take is a commonplace linguistic activity which has so far eluded the attention of linguists and by and large conversation analysts, as well. What, then, does “expert” discourse analysis have to bring to the table that the folk-theoretical “picking up apart” does not?3 It certainly, sets its bar high enough. Can it live up to its potential? An illustrative recent example of discourse analysis taking on the mantle of expertise is a definition of Critical Discourse Analysis (CDA) in an introductory text. Locke () defines CDA as a discipline concerned with “social effects of the meanings a reader is being positioned or called upon to subscribe to” and attempting through analysis of discourse seen as “language in use” to expose these ‘hidden’ meanings to the possibility of “contestation”. [my emphasis]. However, the assumption that these meanings were previously “uncontestable” is left unquestioned. The discourse analyst, then, is implicitly positioned as someone who can tell the non-expert “what they really mean”. But what does “really meaning something” entail? As the following three examples demonstrate, telling someone what they “really mean” is a rather common activity.4

Types of Evidence for Conceptual Patterns in Discourse ‘I’m not a racist but…’ really means ‘I am a racist and…’. (Mark Steel, English comedian) [my emphasis]


In this case the “real meaning” is perceived to be the largely latent attitudes of the speaker. However, the second example not only alludes to latent attitudes but also an explicit hidden agenda.
“This is bad news, as America keeps losing the race to other countries to attract the world’s best and the brightest”

I think he really meant:

“This is bad news, as America keeps losing the race to other countries to attract the world’s cheapest” (Anonymous online comment) [my emphasis]

And assuming hidden “real” meanings is certainly not limited to comedians or participants in online discussions. The following is taken out of George Lakoff’s suggested pro-Democratic partisan manifesto.
Smaller government is, in conservative propaganda, supposed to eliminate waste. It is really about eliminating social programs. (Lakoff, : 94)

All of these examples not only take a critical stance but also use conceptual integration to allude to very explicit political positions. For instance, the racism quip could easily be rephrased as:
Many people only use non-racist language out of social convention. However, as their statements clearly imply, their underlying attitude is one of non-tolerance and much closer to racism than is fashionable to admit.

Clearly, the power of the simple image blend, is lost but the propositional content is by and large preserved. What could a discourse analyst add to this incisive, if partial comment? Perhaps a heuristic for identifying elements of text that are good candidates for contestation. However, as all discourse analysts know (albeit some


Chapter Eight

of them only deep down), no such heuristic can exist. A particularly telling example is this recommendation from an introductory text on discourse analysis for social psychologists (Potter and Wetherell, ). The authors identify ten steps in extracting information from text:
1. Research questions


Sample selection


Collection of records













10. Application

(Potter and Wetherell, : 159ff.)

Analysis, is conspicuously relegated to a single item in a long list of necessary steps, hiding a complex process underneath an innocent-looking label. A reader expecting further elucidation will be treated to the following:
Analysis of discourse is like riding a bicycle compared to conducting experiments or analysing survey data which resemble baking cakes from a recipe. There is no obvious parallel to well-controlled experimental design and test of statistical significance. (: 169) [my emphasis]

Types of Evidence for Conceptual Patterns in Discourse


The simile of “riding a bicycle” is an intimation of not just the analogue (rather than digital) nature of discourse analysis but also of its general accessibility. Other introductory texts are equally silent on the question of the exactness of the analytical method. Let’s take an a look, then, at an actual example of a piece of analysis by a professional and respected analyst. MacLure () scrutinizes the following statement by Chris Woodhead, a former Chief Inspector of Schools in England.
I used to try to read these [academic] journals. Life is too short. There is too much to do in the real world with real teachers in real schools to worry about methodological quarrels or to waste time decoding unintelligible, jargon-ridden prose to reach (if one is lucky) a conclusion that is often so transparently partisan as to be worthless. (Woodhead, writing in the New Statesman in 1998) (Quoted by MacLure, : 12)

MacLure provides the following commentary:
The Chief Inspector’s appeal to the real is a common tactic in the construction of binary arguments about relevance, an issue that has a long pedigree as a boundary that constructs educational heroes and villains. The ‘discourse of derision’ referred to above drew much of its force from the claimed irrelevance of teachers’ outmoded views and values. Thus teachers have been constructed as both the enemies and the defenders of relevance. (MacLure, : 12) [my emphases]

However, she provides no evidence of this “tactic” being “common” nor does she do much to avoid Woodhead’s accusation of partisanship by labelling him as being a part of a “discourse of derision”. In fact, her claim “teachers have been constructed” would have been one of the first to be described as ideological by Hodge and Kress for its use of the unelaborated passive construction. That is not to say that one cannot engage with Woodhead’s assertions or point out that his use of language is reminiscent of other similar uses, as part of that engagement. However, it is difficult to see what discourse analysis as such brought that serves to expose some “hidden” meanings. In fact, Woodhead’s meaning appears anything but hidden. He states his opposition to certain approaches to education and does so in no uncertain terms. To engage with his views, discourse analysis is not necessary. Rather than clarifying it makes his meaning more opaque and uses this selfimposed opacity as an instrument to legitimize the critical analyst’s views. If anything, it is the analyst herself who has a hidden agenda, viz. to argue with Woodhead’s actual stance but attacking it by the proxy of discourse. This concealed argument under the guise of analysis is anything but uncommon in critically oriented discourse analysis. Elsewhere, I pointed out similar problems in


Chapter Eight

Locke’s () account of a newspaper column dealing with Maori education in New Zealand (Lukeš, ). Widdowson ( collects many other examples. MacLure, unlike others, however, sets out an alternative. She proposes a discourse-oriented approach to education.
If I were to define a discourse-based approach to educational research (a problematic task […]), one of the more general things I might say is that you have to suspend your belief in the innocence of words and the transparency of language as a window on an objective graspable reality. (p. 12) [my emphasis]

She suggests that (presumably, unlike the likes of Woodhead):
A discourse-oriented educational research would attend to the multiplicity of meanings that attach to (and divide) the people, spaces, objects and furniture that comprise its focus–the teachers, children, classrooms, textbooks, policy documents– and to the passion and politics that are inevitably woven into those meanings. It would not try to distinguish the ‘real’ teacher from the rhetorical ones. But it would be immensely interested in how appeals to ‘real teachers’ and ‘real worlds’ work as rhetorical power-plays that try to install some version of reality by disqualifying others. (MacLure, 2003: 12) [my emphases]

This, as we saw earlier with Locke, sets the discipline of discourse analysis a very difficult task. The purpose of this chapter is to investigate whether this is a realistic undertaking, and in particular, whether we can amass sufficient and reliable evidence to support such claims. I suggest that to do so, it is necessary to establish how texts actually construct meaning and what this constructed meaning looks like. The original Critical Linguistics assumed that meaning was carried and concealed by units of language subject to transformational rules. CDA has proposed Systemic-Functional linguistics to reveal units that expose and conceal meaning. Meaning, however, is seen as rather unproblematic, if complex. It simply happens and we somehow know what it is or what it could be if it hadn’t been concealed. It is sufficient, CDA seems to propose, to analyze text using whatever linguistic method available, and then find the concealed meanings, variously encoded by linguistic structures. While we have precise methods for analyzing text as such, uncovering meaning is still more like “riding a bicycle”. In my view, this will inevitably lead to the situation where linguistic method and its terminology are used to support an argument that would have been made anyway. Rather than serving a purely analytical purpose, they serve just as much the political purpose of discrediting the opponent. The fact that the method of linguistic analysis is rather inaccessible to the general public should also not be discounted as a factor in the power play between the text, its author, and the analyst.

Types of Evidence for Conceptual Patterns in Discourse


Meaning and text
There now appears to be a general agreement that meaning is constructed. But how and when is it constructed? And, since we’re dealing with text, how is constructed over text? The central claim of this chapter is that we must first look at the conceptual patterns of meaning constructed and then try to see what evidence we can find for them in text. Conversely, we must pay much closer attention to how parts of the text contribute to the construction of meaning. Cognitive linguistics has much to offer on this point thanks to its rejection of boundaries between form and meaning, as well as semantics and pragmatics. The remainder of this chapter will deal with how concepts introduced in cognitive linguistics such as cognitive models and conceptual integration can provide a viable theory of the construction of meaning that can then be subject to the analysis of discourse. This influence, however, is not a one-way street. I will also try to show how the cognitive linguistic enterprise can benefit from paying closer attention to discourse. We must start with acknowledgement, however, that the concept of text and discourse is in itself not a straightforward matter. While a linguist will know a text when she sees one, it is her presumptions about its properties that will influence the approach to analysis she embarks upon. Rather than attempting an exclusive definition of text and discourse let me dwell briefly on the various metaphors for text available to discourse analysts (see Fig. 1). Hoey () identified the prevailing metaphor of text in text linguistics as viewing text as a sentence. This leads researchers to looking for structures in text that they are used to finding in sentence: lexical-like units, syntactic structural relations, finite set of generative rules, grammaticality, completeness, etc. This view of text is typical of early text linguistics. However, even a brief examination of this metaphor reveals that text is unlike a sentence in more ways that it may resemble it. Hoey suggests an alternative metaphor of text as an “oeuvre” of a single author’s works. While individual works progress over time in a linear fashion, they will contain many meanderings and a variety of interconnecting strategies maintaining overall cohesion without structural dependencies equivalent to concepts such as hypotaxis that we might look for if we considered text to be like a sentence. There are other analogies that may help us elucidate how text is dealt with in various analytic contexts. Parallel to text as sentence is the metaphor of text as a dialogue. In this case, a researcher will look for patterns of interactivity with the intended audience, other texts, evidence of turn-taking, etc. This perspective can be found in literary criticism as well as many postmodern approaches to text. Hoey () suggests another metaphor for certain kinds of texts, namely, that of a colony or beehive. Colonies allow for the inclusion of texts such as lists (shopping list, phone book, dictionary) or collections (encyclopaedia, journal). While Hoey’s


Chapter Eight

focus is on some less studied texts, there are ways in which any text can be viewed as either a colony or a part of a colony (see for instance Page, ). A final (for the purposes of this chapter) metaphor of text is that of a mental or discourse space. This metaphor is most closely associated with cognitive approaches to text but it has multiple incarnations and sources. In its one incarnation, it can be traced to psychological concerns with memory and attention scopes. This can be found in literature as diverse as topic-focus articulation in generative functional grammar (Hajičová, ), anaphoric co-reference (Ariel, ) or psychology of text comprehension (Sanford and Garrod, ; see also O’Halloran, present volume). A more pertinent conception of text as a space can be found in cognitive linguistics proper. This idea, building on Fauconnier’s () mental spaces and Lakoff’s () cognitive models, was most clearly elaborated by Werth (). In this context, text as a mental space corresponds nicely with the concept of frames and models which I will investigate in the following section. My aim in outlining the different metaphors of text present in the research literature was neither to provide a definitive classification, nor to declare an allegiance to one or the other approach. Nor are the scholars cited expected to subscribe without reservation to the metaphor as presented here. Rather, my intention is to demonstrate the complexity of text and often overlapping and contradictory textual patterns that can be identified. This will come in good stead when we look at the variety of textual evidence available to us as discourse analysts.
sentence dialogue oeuvre (Hoey) colony (Hoey) mental (discourse) space single, uncomplicated meaning, hierarchical organization, tight structure, codification interaction with audience, turn taking, extra linguistic features progression over time, linking with other texts, development and change of argument thematic unity, random organization, collective meaning spatial relationships, non-codified hierarchies, visual arrangement, on-line processing, relationship to real world

Figure 1. Metaphors of text and their consequences Before moving on, let me briefly address the vexed issue of the definition of discourse. So far, I have used the term virtually interchangeably with “text” and given the variety of metaphorical embodiments of text, this seems to be an acceptable terminological compromise that is adopted by many researchers. However, the indeterminacy of the term makes it both very versatile and potentially misleading. In the literature, we find “discourse” referring either to a

Types of Evidence for Conceptual Patterns in Discourse


single conversation or text, to a collection of texts or conversations, or to a shared way of talking or creating texts (equivalent to code). The last of these frequently leads to plural use of “discourses”. Yet another use is the instantiation or use of text in interaction (similar to the langue/parole distinction) which views discourse essentially as a process rather than product. This chapter uses the term referring to all of these possible uses, only specifying which when confusion could arise. This allows me to benefit from the power and versatility of indeterminacy while avoiding the pitfalls of terminological confusion.5

Discourse and text as mental space
The metaphor of text as a space has received particular attention in the past two decades. As noted above, part of its appeal is its correspondence to a variety of other concepts outside linguistics proper, in particular, model in psychology and frame in social sciences. Perhaps the most detailed elaboration of this idea was provided by Werth () in his concept of “text world”. Having borrowed the term from van Dijk, he describes it as:
a deictic space, defined initially by the discourse itself, and specifically by the deictic and referential elements in it. It falls within the definition of ‘mental space’ (Fauconnier, ). The deictic and referential elements are given by the discourse. The referential elements in their turn, activate relevant areas of memory, including complex conceptual structures known as frames. Frames are whole chunks of experience and situations, codified and stored in memory as single items. These operate to ‘flesh out’ the discourse from the knowledge and imagination of the participants. This accounts for the fact that every individual will build up a slightly different text world from the same discourse input. At the same time, there are strong restrictions on this so that individual differences remain within accepted boundaries.” (Werth, 1999: 20) [my emphasis]

The text world, for Werth, is a result of the process of “discourse” which he defines as:
A deliberate and joint effort on the part of a producer and recipients to build up a ‘world’ within which the propositions advanced are coherent and make complete sense. (p. 51)

Text world, for Werth, is a similar concept to “mental space” (Fauconnier), “semantic frame” (Fillmore), or “idealized cognitive model” (Lakoff). By viewing discourse as the process of constructing text worlds, Werth makes it possible for cognitive linguistics to study discourse outside the confines of text as a sentence


Chapter Eight

metaphor. His approach is similar to that of Langacker () in the emphasis of constructing a cognitive grammar of discourse. However, Werth’s approach also demonstrates a number of weaknesses of a straightforward application of a theory of grammar. First, his reference to memory storage and activation is overly simplistic. He does not have a good theory of online processing or entrenchment that blending theory would bring only later6. His assumption that frames are “built up out of the repetition of ‘similar’ text worlds.” (p. 51) is little more than wishful thinking. But most importantly, Werth did not provide a satisfactory account of how conceptual structures are represented at all levels of text. His examples of analyses focus mostly on the construction of simple propositional text worlds dealing with issues of counterfactuals and deictics (similarly to Fauconnier , ). In the process, some of the key achievements of cognitive and construction grammar, mentioned above and in the Introduction, are left unexplored. Also, Werth only insufficiently elaborates the concept of frame itself, defining frames as “areas of memory which relate to areas of experience and knowledge encoded as complex conceptual structures.” (p. 51 [my emphasis]) What is the nature of the “complexity” of these structures and how is this complexity represented in discourse? That is the subject of the following section.

Conceptual patterns in discourse
Investigating the conceptual underpinning of discourse7 has always been, whether explicitly or implicitly, the primary purpose of discourse analysis. It is also the area where the least rigor has been applied and almost no explicit consensus exists. 8 That is one of the reasons while I choose to talk of patterns rather than structures as Werth and others do. The concept of structure carries with it certain assumptions of stability, connectivity, and hierarchy that may prejudice the outcome of an inquiry in a way that the idea of a pattern will not. However, although we cannot identify a theoretical let alone methodological consensus, there seems to be a general convergence around the idea of a frame. Conceptual frames have become the nexus of investigations by an astonishing variety of disciplines: sociology and anthropology, social psychology, cognitive psychology, artificial intelligence, cognitive linguistics, politics, education, and policy studies. While the origins of the concept can been traced much further into the past (e.g. Tannen, : 15-17), interest in it only led to research of current relevance in the 1970s and 1980s. Unfortunately, this research is not only scattered across disciplines frequently unaware of each others’ accomplishments which is made even more intractable by the terminological variety. The same concept has been called a frame, a script, a schema, or a model to name just the most common. There isn’t even a consistency within scholars’ individual oeuvres. Goffman, one

Types of Evidence for Conceptual Patterns in Discourse


of the most influential proponents of the concepts in sociology, calls for frame analysis () but later uses the term “footing” to refer to what is essentially a frame (). Lakoff introduces and elaborates “idealized cognitive models” (, ) but reverts to frame in his popularizing and applied work (). Ensink and Sauer () conducted a literature review in which they identified seven disciplines in which a concept similar to “frame” was referred to by eight different terms. This review does not include domains, mental spaces or Idealized Cognitive Models from cognitive linguistics, stereotypes from social cognition, and fails to mention frame analysis in policy studies (e.g. Schön and Rein, ). Broadly speaking, the concept of a frame appears in two overlapping contexts which are close to forming two separate research traditions: 1. Social (sociology, anthropology, policy analysis) 2. Psychological (social cognition, cognitive development, psychoanalysis) In general, there is only scant acknowledgement from students of one of the contexts of accomplishments achieved in the other. Accordingly, linguists and discourse analysis applying the concept tend to align themselves more strongly with one or the other research tradition. Tannen () provides an excellent overview of discourse linguistics conducted within the social tradition while Lee () gives an overview of linguistics more influenced by the psychological tradition. The psychological context should further be divided into scholars adopting the perspective of social psychology (exemplified by the work of Teun van Dijk) and those working more with cognitive psychology in mind (exemplified by the work of George Lakoff) Although, these three traditions share many assumptions and analytic goals, they remain remarkably resistant to mutual exchange of ideas. The others’ work is occasionally but, by no means always acknowledged9, but rarely used to validate or question one’s assumption. This is even more prominent in the applied disciplines such as policy analysis where many cite influences of Goffman (), van Dijk (e.g. (, ), or Lakoff (typically as Lakoff and Johnson, ) but almost never combine the two10. Fisher () is one of the few overviews that gives equal attention to all of these traditions. This fragmentation and insularity is detrimental to the study of conceptual patterns (for terminological convenience I will refer to them as frames from now on) since each of these contexts reveals something interesting about them. Below I briefly outline what relevance the two different contexts of frames have for a critically as well as conceptually minded discourse analysis.


Chapter Eight

Frames in the social context
When seen in the social context, frames reveal important facts about their importance for social interaction and the variety of real-world contexts in which their study becomes relevant. In fact, Schön and Rein () demonstrate that in the world of policy studies, the analysis of frames should be the primary rather than ancillary approach. The social context also profiles certain aspects of frames that might otherwise remain hidden. In line with Goffman (), Tannen defines frames in terms of expectations and actions:
prior experience or organized knowledge [that] takes the form of expectations about the world, and in the vast majority of cases, the world, being a systematic place, confirms these expectations, saving the individual the trouble of figuring things out a new all the time. (Tannen, 1993: 29) [my emphasis]

However, Schön and Rein add a practical dimension that makes frames real in a specific case of social interaction.
Frames are not free-floating but are grounded in the institutions that sponsor them, and policy controversies are disputes among institutional actors who sponsor conflicting frames. (Schön and Rein, : 29)

The emphasis on the interactive, negotiated, nature of conceptual frames is one of the aspects of frame-like structures often overlooked in linguistic and discoursebased studies. However, in the social context, and in particular in applied areas, such as the study of policy decisions, the interaction (constructionist) aspects are hard to miss. In fact, Benford and Snow (), dealing with collective frames of action in social movements, rebuke scholars who “tend to treat collective action frames in a fashion that is more consistent with psychological concepts such as ‘schema.’ thereby overlooking the interactive, constructionist character of movement framing processes.” (p. 614) They go on to quote Gamson who maintains that “collective action frames are not merely aggregations of individual attitudes and perceptions but also the outcome of negotiating shared meaning” (Gamson, : 111). In cognitive linguistics, there is much talk of theoretical constructs being “psychologically real” and considerable effort to infuse research with such realism (see for instance the contributions by Maalej, O’Halloran, and indirectly de Landtsheer, and Sánchez-García in the present volume). However, the research conducted in the social context brings into mind the need for “sociological realism” (or perhaps ethnographic) to go hand in hand with “psychological realism” when we discuss conceptual patterns in discourse.11

Types of Evidence for Conceptual Patterns in Discourse


Frames in the psychological context
Of course, although we encounter them at their most interesting in social and other applied contexts, frames are primarily creatures of the mind. It is therefore no surprise that the concept emerged almost simultaneously in theory of social interaction (Goffman, ) and in theory of mind represented in the mid-1970s by artificial intelligence12. In the 1980s, the concept of cognitive models gained traction in cognitive psychology (e.g. Johnson-Laird, , or Gentner and Gentner, eds, ). Of direct relevance to discourse study, however, is the work of George Lakoff (particularly and ). Lakoff introduced the term Idealized Cognitive Model (ICM) to account for whole range of mental phenomena that are only hinted at in the approaches hinted at so far. More specifically, ICMs were conceived to account not only for propositional encyclopedic knowledge structures and implicit “expectations about the world” but also for metaphorical and metonymic conceptualization, image schematic structures, prototype effects in categorization, and online processing. For Lakoff, ICMs are not simply lexical-like items that can be looked up, they provide structure to mental spaces (see above). Although issues related to online processing have not received sufficient attention until Fauconnier and Turner’s () theory of conceptual integration or blending theory (see below), Lakoff’s chief contribution was to demonstrate that frames (ICMs)13 are not simply amorphous mental structures but that they are richly structured by a variety of cognitive phenomena. This, in combination with blending theory, is of paramount importance for discourse analysis. Lakoff () suggests four structuring principles for frames (ICMs) (p. 68):
1. Propositional structures

2. Image-schematic structures

3. Metaphorical mappings

4. Metonymic mappings

This means that to fully account for frames, we need to take into consideration things like: cultural scripts and scenarios of varying levels of schematicity narratives (stories, anecdotes, archetypes) images of varying levels of schematicity (image schemas, rich images)


Chapter Eight

primary and secondary metaphors typical and ad hoc metaphorical and metonymic understanding14 folk theories and expert theories generative principles, cognitive strategies and algorithms typical examples, paragons, cultural stereotypes categorial structures (including basic-level and prototypical phenomena) etc. This is purposely not intended to be an exhaustive list of discrete items. These are merely examples of the consequences of Lakoff’s four frame-structuring principles. Lakoff () lists further examples. What is of particular interest to our present enterprise which is to identify textual evidence for frames is that Lakoff demonstrated that the phenomena involved in framing are not limited to discourse level but occur at all levels of language from phonetics to text. Frames are then likely not to be textually represented by any one kind of text pattern, they can be evidenced by words, phrases, phonemes, morphemes, genre, register, and other stylistic devices, codeswitching, intonation and other suprasegmental devices, orthography, paralinguistic phenomena, corpus-level phenomena such as semantic prosody (Louw, , Stubbs, ), text-level phenomena such as cohesive harmony. Framing, conceived this way, also explains most of pragmatics’ and semantics’ most intractable problems. How is this relevant to a critically-oriented study of discourse. Lakoff presented several illustrations of how frames (ICMs) are relevant in this context. One was his illustration of the complex metaphorical patterns of anger including investigation of how discursive representation of domestic violence can be explained by metaphorical frames. Unfortunately, the study of metaphorical frames in criticallyoriented discourse analysis has largely confined itself to the study of metaphor in the limited sense of Lakoff and Johnson () without reference to the complex issues associated with framing—see Maalej and O’Halloran in this volume for a more detailed critique. For an example of a more complex illustration, we have to go to Lakoff’s () Moral Politics, which has been inexplicably almost completely ignored in the discourse literature. In Moral Politics, Lakoff applies the ICM (frame) concept to the study of partisan political debates and shows how frames not only help participants make sense of their discourse but also provide an internal logic and coherence to seemingly incongruous positions. This last point is what ultimately makes discourse possible, at all. However, Lakoff’s inquiry is almost completely outside the text linguistic and discourse analytical tradition going back to the earlier work of Halliday (e.g. Halliday and Hasan, ) and van Dijk (e.g. van Dijk and Kintsch, ). As such Lakoff is agnostic as to issues regarding the nature of text and discourse outlined above. As a consequence, he offers no theory of the

Types of Evidence for Conceptual Patterns in Discourse


textual representation of the conceptual structures he posits. And, more importantly, he has no theory of the relative value of the evidence he presents for his conclusions. Investigating the possibility of such a theory and its implications is the task of the rest of this chapter.

Frames in the context of text and discourse: Types of textual evidence for conceptual patterns
The investigation of the representation of conceptual frames in the context of text and discourse has been mostly neglected by scholars as an issue to which significant research resources should be directed. I propose, that in addition to the psychological and sociological realism of frame theory, a third type is added, namely that of textual realism. For a complete picture, frames must be considered with reference to their mental representation, their function in social interaction, and their textual/discourse representation. Here, it behooves us, again, to recall the various metaphorical castings of the “text” and “discourse” outlined above. Textual realism must take into account all of them.15 Below, I investigate two complementary approaches to identifying conceptual structures in text. Van Dijk and Kintsch () represent a similar attempt at a systematic model of text-based discourse modeling, which has become the basis of a strand of critical research into political and media discourse (e.g. van Dijk, , ). However, this approach does not identify specific structures of textual evidence.

Textual evidence for metaphor
Given that metaphors are one of the most prominent conceptual devices in discourse research, it is not surprising that the most attention has been given to their identification in text and the implications of their presence for the meaning of text (e.g. Steen, , , ; Wallington et al., ; Charteris-Black, ; Musolff, ; Deignan, ). One of the more comprehensive surveys of how metaphor is represented in text that can be of use in the current context was compiled by Goatly (). He lists a number of ways in which metaphors are signaled in text including the following:
Markers: “metaphorically speaking”, “literally”

Intensifiers: “actually”, “almost”, “in fact”, “regular”

16 Hedges: “in a way”, “technically”

Chapter Eight

Metalanguage: “in one sense”

Mimetic terms: “image”, “likeness”, “picture”

Causal similies: “as if”, “though”

Perceptual, cognitive and other processes: “seemed”, “sounded”, “looked”

Others: orthography, modals, conditionals, etc.

(adapted from Goatly, 1997: 174-175)

Goatly also describes the ways in which all these markers signal different types of metaphor and also may have an intensifying or reductive effect on the metaphor. It is also important to remember that not all, and perhaps not even most, metaphors are signaled by one of these markers and conversely not all of the occurrences of these markers are predictive of the presence of a metaphor (see Wallington et al., ). A key marker of metaphor, of course, is a lexical representation of the domain mapping, such as “Students invested into their learning.” However, the identification of non-lexical markers for metaphor is an important step toward identifying a broader textual evidentiary base.

Textual evidence for frames
The only work dealing explicitly with the textual representation of frames is Tannen’s () analysis of transcripts of a group describing a film to a person who had not seen it16. Tannen identified sixteen types of evidence of “structures of expectation” that can be “seen to reveal the existence of these expectations (or scripts or frames or schemata)” (p. 41).
1. Omission; 2. Repetition; 3. False starts; 4. Backtrack; 5. Hedges and Hedgelike Words or Phrases; 6. Negatives; 7. Contrastive Connectives; 8. Modals; 9. Inexact Statements; 10. Generalization; 11. Inferences; 12. Evaluative Language; 13.

Types of Evidence for Conceptual Patterns in Discourse Interpretation; 14. Moral Judgement; 15. Incorrect Statement; 16. Addition (Mentioning a character or episode not present in the film)


(Tannen, 1993: 41)

Her list exhibits some similarity with Goatly’s, and in fact, it could be said it subsumes his categorization. For instance, the definition of hedges, connectives and modals could be extended to account for a significant proportion of Goatly’s markers (metalanguage, intensifiers, markers, causal similies, etc.) while others derived from the nature of spoken language (repetition, false starts, backtracking) can find written language equivalents (orthography, elegant variation, etc.). Most notably, however, a large proportion of Tannen’s types of evidence has to do with pragmatic issues of context (evaluative language), and reference (incorrect statements) that cannot be identified simply by reference to the text without extraneous concerns. In particular, she can only identify “incorrect statement” and “addition” by virtue of having privileged access to the original film which will frequently be hidden from a discourse analyst. Furthermore, not all of these types of evidence point to the same “level of frame” since “[a]ny speech event represents the overlapping and intertwining of many relations concerning the context as well as the content of communication.” (p. 22) Her speakers then provide frames of themselves in relation to the text, positioning themselves as subjects of an experiment, somebody telling a story, describing a film, or watching a film. Tannen also looked at what kinds of expectations are revealed: events (personal encounters, confrontation, accident, etc.), objects (roads, trees, etc.) or locations. Tannen concludes to have shown that
the notions of script, frame, and schema can be understood as structures of expectation based on past experience, and that these structures can be seen in the surface linguistic form of the sentence of a narrative. Furthermore, the structures of expectation which help us process and comprehend stories serve to fill and shape perception. That is why close analysis of the kinds of linguistic evidence I have suggested can reveal the expectations or frames which create them. (p. 53)

Textual evidence for conceptual patterns
Both Goatly and Tannen operate within the confines of their subject matter. Goatly is limited by focus on metaphor as an isolated conceptual structure and Tannen by focusing on spoken language recounting a film-mediated event. When we contrast their types of evidence with the richness of conceptual frames as structured by Lakoff’s four principles, some gaps emerge. There is no mention of the types of


Chapter Eight

descriptions, such as images, descriptions of social relationships, categorization, espousal of folk theories, etc. Here, in no particular order, are a few examples of additional features of text which might be necessary to include in a conceptual discourse analysis:
Evaluations; Definitions; Hypostasis; Analogies (and other tropes); Lists; Categorical hierarchies; Hypercorrection; Linguistic purism; Cohesive patterns; Elegant variation; Quotations; Prefacing; Freezes; Terminology; Disclaimers; Sourcing (paragons); (Pop culture) references; Stories; Images; Exegesis; Expression of expectation; Semantic prosody; Cohesive harmony; Pronoun reference; Participant role description; Summaries; Spelling; Graphical organization of text; Genre context; Social context

So far, I have not explored the linguistic nature of these types of textual evidence. When we look at linguistic theories of the last hundred years or so, we find, that all of these textual devices would not ever be classed together and many of them would not even be considered to be of interest to linguistics proper at all. It is only now with the advent of usage-based models of language (Kemmer and Barlow, ) and most notably cognitive and construction grammar (e.g. Langacker, , Goldberg, , and Croft, ) that it is possible to consider things like spelling and genre to contribute directly to the construction of conceptual patterns (cf. also Lukeš, a). Whereas, traditionally, every linguistic device belongs to one in a hierarchy of levels (phonological, morphological, lexical, syntactic, textual, stylistic) the concept of construction obviates the need for these artificial boundaries forcing classification where it is not only needed but is often detrimental to understanding. What’s more, Fauconnier and Turner’s (e.g. ) blending theory (see Hart, present volume, for detailed exposition), provides a mechanism for conceptual integration that works equally well on all kinds of constructions.

Some examples
Here are selected passages that illustrate some of the frame building textual devices. They come from a text analyzed for the presence of different metaphor use in more detail in Lukeš (b). The examples below illustrate some of the other features of the text that can trigger a frame in the context of building a mental space associated with the text.
Our report sets out a clear vision for a unified framework of 14-19 curriculum and qualifications. (Prefacing, Referencing; my emphasis)

What is remarkable about this sentence that it contains two potential frame builders yet their contribution to the overall mental space built up by the text is minimal. It

Types of Evidence for Conceptual Patterns in Discourse


first instantiates the context but this is done for purely cohesive rather than space building reasons, since the reader already knows they are reading a report the purpose of which is to “set out” something. “Vision” and “unified framework” are much more interesting since their potential contribution of conceptual frames to the space is great. However, because of the genre context (this is basically a topos, i.e. a ritualized text chunk) it does little more than place the text in the right social context.17
our system is not providing the stretch and challenge needed, particularly for high attainers (Metaphorical semantic prosody, Freeze)

Here, again we see several potentially significant frame triggers blended into the background by integration with transactional elements of the text. The integration of “system” and “providing” with freezes “high attainers” and “stretch and challenge” along with the register and genre context suppresses the conceptual patterns associated with the idea of provision (education as a good for sale) and system (collection of interconnected parts functioning according to an algorithm) and translates into little more than “the smart students are bored in class”. However, the presence of freezes indicates something more important, namely, the fact that patterns of expectation exist and are associated with them. Also the fact, that both of them come with a high degree of semantic prosody.
Despite its weaknesses, the current system has its strengths. Many elements of the reforms we propose can already be found in schools and colleges around the country and we want to build on their good practice. (Disclaimer, Evaluation)

The way we alerted for the presence of frames here is the integration of its transactional function of a disclaimer with its ideational function of evaluation. This suggests that more conceptual richness is available.

Conceptual integration in discourse
How is the above relevant to the job of a discourse analyst? We can see that the presence of frames can be both intensified and suppressed by formal features of the text and that the richness or schematicity of “meaning” of any of the constructions above varies greatly depending on the context. As I pointed out at the outset, identifying the textual representation of frames in text is not the primary purpose of the conceptual analysis of discourse. Our first and foremost aim is the discovery of conceptual patterns in discourse but to achieve this with the most possible effort at


Chapter Eight

empirical realism on all levels (psychological, sociological, textual). What we are really interested in is how these textual framing markers are used to 1. build a mental space (collection of activated and blended frames) of a given text; and 2. contribute to the social repository of culturally available frames?

Building and blending the discourse space
Let us illustrate the first question on another example:
status quo [preface, space builder] is not an option … many young people leave education lacking basic and personal skills [evaluation] … The results [elaboration] are a low staying-on rate [terminology, reference] post-16; employers having to spend large sums of money [reference, scenario description] to teach the ‘basics’ [quotation, evaluation]; HE struggling to differentiate between top performers [terminology]; and young people’s motivation and engagement with education [freeze] reducing [evaluation] as they move through the system [scenario].

Here, we see a number of text segments that can potentially trigger frames (mental spaces) to be blended. If we view them as constructions (in the sense of Langacker and Croft), we need to take into account their schematic meaning (i.e. their role in the transactional business of constructing a text), as well as the schematic genre meaning of the text as a whole. This allows for a much more complete view of the final blend, that could look something like Figure 2.
Generic input spaces: Reader perspective: analyst, teacher, government official Context: debate over quality of exam results, book layout Genre: preface to a government sponsored report Conceptual: selectively activated frames / cognitive models / scenarios Status quo input space: scenario: employers spending money to teach the basics to students who didn’t learn because of poor schooling Change input space: negating the status quo input space

Blended space: Partial and individual scenario of negated frames from the Status Quo input space moderated by knowledge of genre and context

Types of Evidence for Conceptual Patterns in Discourse Figure 2. Example of blending


I do not use Fauconnier and Turner’s graphical notation (see Hart in the present volume) on purpose to preserve the richness of the conceptual patterns structuring the input spaces. Also, the final blend is not one that presents neatly linked mappings between discrete elements of one space with those of another. Rather, the blending of the elements is highly schematic and partial. Schematicity and partiality of the blending process cannot be overemphasized. Furthermore, the level of elaboration will differ depending on the reader’s perspective (see O’Halloran, , and present volume). One of the rather forgotten aspects of blending that Fauconnier and Turner () include in their descriptions, is their opportunism (p. 367)18. The blending process does not produce universally valid and uniform results and does frequently lead the process of building a space in unpredictable ways. Take a recent comment by the US presidential candidate Hillary Clinton when answering a question from a constituent. She merely rephrased the question and paused:
What in my background equips me to deal with evil and bad men? [pause][laughter by audience, speaker joins in] (see for details)

However, the audience, including herself, used (opportunistically) the space afforded by the pause to integrate the other mental images of Clinton’s relationship with her husband and his detractors as well as the public perception of gender discourse to map the image of “evil men” onto the images of variety of other prominent males rather than the one intended by the questioner, i.e. Osama bin Laden. This innocent episode gave rise to quite a volume of commentary trying to analyze what Clinton “really mean” by her pause and joining in with her audience’s laughter. However, a close observation of the speech, as well, as analysis of the text makes it rather clear that her “meaning” was schematic and partial and a highly integrated blend. So far, we have only addressed the issue of the mental space and frames evoked by rather small chunks of text. However, what about a larger text such as that dealt with by Hoey () who can produce an elaborate system of patterns of connections between surface lexical items to show a cohesion of book-length text. However, the complexity of the task increases dramatically, if we try to add conceptual patterns into the mix.


Chapter Eight

Building a discourse universe
The second question posed above with regard to the discourse analytic utility of textual framing markers was how they are used to “contribute to the social repository of culturally available frames”? The answer here is even less clear-cut than when we ask about their contribution to the understanding of an individual text. Blending theory provides the concept of “entrenchment” which is similar to the concept of long-term memory that can be found as far back as van Dijk and Kintsch’s () seminal work on discourse and also in van Dijk () and elsewhere. However, this model is predicated on a very simplistic model of memory as the cumulative result of individual impressions in the course of exposure to a particular item being instantiated in a chunk of discourse. Here is a typical example of scenario activated in this context:
the time over which anything resides in explicit focus is a joint function of foregrounding and usage ... Thus, a particular explicitly mentioned entity gains in foregrounding every time it is mentioned, and drops in strength when other things are mentioned instead. Oft-mentioned entities reside in focus longer, and therefore their representations will give rise to stronger representations in long-term memory. (Sanford and Garrod, : 183)

This has a clear parallel in more popular accounts of the long-term large-scale impact of frames (e.g. Lakoff, ). However, as we saw above, a simple use of a frame trigger does not necessarily imply its complete activation nor does it result in predictable mappings. Neither does it take fully into account the process of frame negotiation (as described for instance by Gamson, ). How the frequency of activation (blending) of a frame contributes to its availability in a notional repository of entrenched conceptual blends is simply not a question to which satisfactory answer has been given. I suggest that the warning against making simplistic assumptions about the neural representations of linguistic and conceptual patterns presented by Lamb () be extended to this issue, as well.

Conclusions: The limits of critical analysis
The answer to the question posed in the title of this chapter will be less than satisfactory to the critically-minded discourse analyst. To achieve a satisfactory level of psychological, sociological and textual realism may be too large a task given the hermeneutic and heuristic tools available.

Types of Evidence for Conceptual Patterns in Discourse


The current mainstream approaches to critical discourse analysis make unwarranted assumptions about the activation of conceptual patterns by elements of text as well as the impact of use on both current mental representation and longterm entrenchment. Before we can make decisions about meanings that are being concealed, we should ensure that we have a clear picture about the meaning that is being constructed. If we accept the basic premise that the discourse space is being structured by richly structured conceptual frames, then blending theory does provide us with a very promising, if as yet incomplete, analytical framework. Its utility is magnified when combined with a usage-based view of language as exemplified by construction grammar. Much work is need to apply these insights successfully to the analysis of constructions that go beyond the level of sentence in scope. However, if we accept, as appears inevitable, the fact that conceptual analysis is always partial and has to deal with constant fine-tuning between partially blended frames, no other analytical frame-work is available. As we saw, textual evidence of conceptual structures is multilayered and lacks one-to-one correspondence with the psychological and sociological evidence. This places serious obstacles in the way of an analyst who wants make conclusions about a particular text or a collection of texts and their contribution to the world of a particular discourse. A sustained and detailed analysis of extended text examining the multilayered patterns of integration in minute detail is simply impossible. However, at the very least, an honest critical analysis must contain a careful examination of the proposed evidence19 to ensure that the political concerns of the analyst were not the primary reason for its selection. It is my hope that this chapter provides an illustration of the hermeneutic concerns as well as analytical tools cognitive linguistics might bring to just such an examination.


Through an interesting process of blending, the neologistic attributive ‘meta’ has become a part of popular discourse. I have pursued similar lines of criticism in a recent review of an introduction to CDA (Lukeš, ). 3 Is there even such a distinction as discourse analysis vs. discourse meta-analysis? 4 As inexact a measure it may be, the fact that a Google search for “really mean*” returned 1,410,000 hits, is indicative of something. 5 I leave aside the issue of the use of both text and discourse referring to non-linguistic modalities. An object can be thought of as text and it is certainly part of a discourse in one of its conceptualizations. However, for the purposes of this chapter, this is of limited relevance. 6 Werth’s monograph was largely completed in 1994 before blending theory gained traction and he was prevented from further elaboration by his untimely death. 7 In this case, discourse in the most inclusive possible sense. 8 This is one of the contexts in which the indeterminacy of discourse has had detrimental effect. 9 For example, Lakoff (1987) inexplicably makes no mention of van Dijk’s or Goffman’s ideas and equally Tannen () citing the importance of Goffman’s work, ignores both van Dijk or Lakoff. Van Dijk, for his part, only cites Lakoff and Goffman parenthetically (e.g. van Dijk, ) 10 E.g. Schön and Rein () for policy analysis and Morgan () for organizational studies. 11 There are also parallels here to Lakoff’s ({Lakoff, 1990 #619}) proposed empirical commitments of cognitive linguistics which in practice have looked to almost no textual and little sociological evidence. 12 This was mostly through the work of Minsky on frames, Abelson and Schank on scripts, and Rumelhart on schemas (see Lakoff, : 116-7, for discussion). 13 Since in his popularizing and political activism work Lakoff () adopted the term frame to refer to ICM-like structures and the term is easily recognizable to researchers in a variety of disciplines, I will continue to refer to ICMs as frames for the sake of terminological unity. 14 This is a slightly confusing point. Metaphors can at the same time be one of the elements in a frame 15 I am not unaware of the parallelism between these three types of empirical realism of conceptual frame research and Halliday’s meta-functions of linguistic structures (ideational, interpersonal, textual) which are common currency in CDA as advocated by Fairclough (see Locke, for an overview). 16 The film is Chafe’s famous pear story (Chafe, ) 17 It is not insignificant that this is well reflected in the popular understanding (folk theory) of text, as witnessed by such phrases as “buzzwords”, “buzzword compliance”, etc. 18 Fauconnier and Turner point to the differences in blending across languages based on contingent features such as possible rhyming which may give rise to more highly integrated blend in one language as compared to another. They compare the process of blending to that of evolution, during which new structures emerge off random mutations. 19 In this I am in complete agreement with Widdowson’s () and Stubbs’ () critique of CDA.


Sign up to vote on this title
UsefulNot useful