5 Text and cognition I

5.1 Cognitive stylistics

In this chapter and in Chapter 6 we focus on an area of stylistics in which interest has grown considerably over the last decade or so. This is an area which has come to be known generally as cognitive stylistics (‘cognitive poetics’ is another term currently in use, though there is arguably no significant difference between what the two terms signify). Cognitive stylistics focuses primarily on hypothesising about what happens during the reading process and how this influences the interpretations that readers generate about the texts they are reading. It proceeds on the assumption that reading is an active process and that readers consequently play an active role in constructing the meaning of texts. Cognitive stylistics has drawn considerable influence from work in areas such as cognitive science generally, psychology, computing and artificial intelligence. Although Stockwell (2002a: 1) claims that cognitive stylistics (he calls it cognitive poetics) ‘is all about reading literature’, there is in principle no reason why cognitive stylistics should not also deal with non-literary texts (the arguments for why stylistics generally is appropriate for the analysis of both fiction and non-fiction writing can be found in Chapter 1).

In recent years, the theories and analytical frameworks of cognitive stylistics have been outlined in a number of key texts (see Tsur 1992, Semino and Culpeper 2002, Stockwell 2002a and Gavins and Steen 2003), though the roots of such work may be traced back much further. Gavins and Steen (2003: 41) cite Tsur's research in the 1970s as being an important precursor to current work in the area, while West (2007) notes that the work of I. A. Richards in the 1930s may be seen as protocognitive in orientation. In fact, any evaluative stylistic analysis (cf. purely descriptive stylistics; see Leech 2008 on this distinction) will always necessitate some consideration of the reading process, and most stylisticians would claim that they have always tried to take account of this. We might, then, see cognitive stylistics as an effort to systematise the way in which this aspect of stylistic analysis should be taken into account.

We will focus in this chapter on the cognitive processes by which readers respond to particular aspects of texts, and the real-life schematic knowledge that they bring to bear in interpreting them. In so doing we will begin with a discussion of schema theory, a theoretical consideration of how we package world knowledge and use this in the interpretation of texts. We will then consider the cognitive concept of figure and ground and how prior knowledge drives our recognition of spatial foregrounding, before going on to discuss cognitive metaphor theory, a theoretical exposition of how we use schematic knowledge to structure our view of key concepts in our lives. With this groundwork in place, we will go on in Chapter 6 to focus on cognitive theories designed to explain how readers monitor their progression through a text as they read.

5.2 Schema theory

Cognitive stylistics is predicated on the notion that readers are actively involved in the process of meaning-making. Meaning is not located solely in the formal structures of the text but is, in a sense, negotiated as a result of readers utilising aspects of their pre-existing background knowledge of the real world as they read. In a discussion of how readers make sense of the fictional worlds of poems, Semino (1997: 125) makes a useful distinction between what she terms projection and construction. Although her discussion is centred primarily around explaining how readers generate images of fictional worlds, Semino's terms are useful more generally for explaining the way in which readers make sense of what they read. Texts project meaning while readers construct it. That is, texts contain triggers which activate aspects of readers’ background knowledge. This then allows readers to construct mental representations of the world of the text.

A useful and related notion is the psychological distinction between bottom-up processing and top-down processing. Used in relation to text comprehension, the former refers to the practice of inferring meaning from textual cues while the latter term describes the practice of utilising background knowledge to aid understanding. Bottom-up processing maps onto the notion of projection, since both refer to the means by which meaning is drawn from a text by the reader. Top-down processing maps onto the notion of construction, since both of these terms refer to readers utilising pre-existing world knowledge to make sense of what they read. In practice, reading a text involves engaging in both bottom-up and top-down processing simultaneously. What we will concentrate on in this section is what it means to invoke pre-existing knowledge in the interpretation of texts. In this respect, we will explain some of the theoretical notions concerning how readers are able to store prior knowledge and access it as they read. The various suggestions for how this might work may be taken together as constituting what is often referred to as schema theory.

5.2.1 Schemas, scripts and frames

The term schema refers to an element of background knowledge about a particular aspect of the world. We have schemas for people, objects, situations and events. As an example, most people will have a JOB INTERVIEW schema which they will use to help them navigate their way through such a situation. This schema may include such information as how the interview room is likely to be set out, how many people are likely to be on the interview panel, what is expected of the candidate, etc. Schemas (or schemata to use the Latinate plural) also encompass linguistic behaviour. For instance, anyone with a well-developed JOB INTERVIEW schema will be aware that it is common practice in such a situation for interviewees to violate the Gricean maxim of quantity (Grice 1975; see section 4.2.2) as they attempt to provide as much information about themselves as is possible (Culpeper and McIntyre 2010 provide a full discussion of what the linguistic contents of a JOB INTERVIEW schema might be, and how particular stylistic effects might be generated by deviating from this).

At this point we should make it clear that while schema is a general term for an element of background knowledge, some writers prefer alternative terms in order to flag up the varied nature of schemas. Minsky (1975), for example, uses the term frame to describe knowledge related to visual perception (e.g. background knowledge about different kinds of buildings). Schank and Abelson (1977) introduce the term script, explaining that scripts are composed of schematic information about complex sequences of events. These have various slots which will be filled by different elements according to the particular script. So, there may be slots for props (desks, chairs, etc), roles (interviewer, interviewee), scenes (entering, greeting the interviewer), entry conditions (having applied for the job) and results (being offered the job, being turned down for the job). Scripts can also have different tracks, each of which may have different slots. For example, a person's JOB INTERVIEW schema may include a ‘Saturday job’ track, an ‘academic job’ track and a ‘corporate management job’ track, and each of these will vary in respect of their constituent slots. For instance, in the props slot for a ‘corporate management job’ track, we are likely to expect the clothing of the interviewee to be very formal (a suit, perhaps). The equivalent slot for the ‘academic job’ track may be filled with a different expectation – perhaps that the interviewee may be dressed much more informally. With regard to the scenes slot, we might expect a restricted number for a ‘Saturday job’ track, while the ‘corporate management job’ track may well include many more – for example, an assessment centre scene, a team-building scene, a social event scene, etc. It is worth being aware of this varied terminology; we will use schema as an overarching term and frame and script where appropriate.

It should be clear from our discussion so far that schemas are distilled from past experiences. Nevertheless, we do not have to have experienced something directly in order to build a schematic conception of it. For example, it is likely that most people have a script for a murder investigation, though only a few people will have constructed this schema from direct experience. For the majority of people, such a script will have been composed from watching television police dramas and reading crime novels. It follows too that schemas are not static. Rather they are dynamic, in the sense that they develop as a result of experience. For example, a football supporter's FOOTBALL TEAM schema may be developed if the team in question acquires a new player. This kind of schema development is referred to by Rumelhart (1980) as accretion; that is, the new information extends the schema but requires no fundamental change to it. Tuning, on the other hand, refers to the modification of an existing schema to take account of new experiences. For example, floppy disks have now been supplanted by hard drives as the primary means of storing electronic data, and this development will have involved a modification to the WORD-PROCESSING schema of anyone familiar with the old way of saving documents. Finally, restructuring refers to the generation of new schemas. New schemas may be based on pre-existing ones (a person new to flying might generate an AIRPORT schema out of a BUS STATION schema) or may, theoretically, be induced solely through experience, though this latter type of schema creation is arguably much rarer.

5.2.2 Triggering schemas

Knowing that we bring our schematic knowledge to bear when we read texts, the next issue to consider is how schemas are triggered. Schank and Abelson (1977: 49–50) propose that scripts are triggered by headers – that is, textual cues that relate to elements of the script in question – and that at least two headers are required to instantiate a schema. Headers come in the following four types:

One issue with schema theory as we have outlined it so far concerns our capacity for confusing elements of our various schemas. For instance, a child who accidentally calls her school teacher ‘mummy’ is clearly mixing up her TEACHER and PARENT schemas. If schemas are discrete, as we have so far suggested, why should such slip-ups occur? Schank (1982) proposes an explanation for such occurrences by postulating the existence of what he calls memory organisation packets or MOPs. He begins with the suggestion that specific schemas have a subordinate relationship to more general schemas. So, for instance, a DOCTOR/PATIENT INTERVIEW script will be embedded within a higher-level INTERVIEW schema. Schank refers to this superordinate schema as a scene. He then proposes that MOPs act as organising structures for particular groups of scenes, all of which are oriented towards the achievement of some particular goal. For example, experiential knowledge concerning going to an academic conference is encapsulated in a CONFERENCE MOP (M-CONFERENCE). Such a MOP might incorporate the following scenes: PLAN + SUBMIT ABSTRACT + BOOK ACCOMMODATION + PAY REGISTRATION FEE + TRAVEL TO CONFERENCE + CHECK IN + REGISTER + PRESENT PAPER, etc. What is significant about MOPs is that some of their constituent scenes may be shared by other MOPs. So, the REGISTER scene that is a part of M-CONFERENCE may also be found in M-MARATHON RACE, M-DANCE COMPETITION, M-CONVENTION, and others. Returning now to the example of the school pupil mistaking her teacher for her mother, one potential reason for this might be that the activity in which the child is engaged when she makes the mistake is a scene that is a constituent part both of a MOP that is appropriate to school and one that is appropriate to home – say, READING A BOOK. In school, this scene may be part of M-LESSON, while at home it may be part of M-GOING TO BED. If, within M-GOING TO BED, the child's schematic knowledge postulates a mother in the role slot for the READING A BOOK scene, it is perfectly likely that this role slot may be activated within the READING A BOOK scene in M-LESSON, thereby causing confusion (and probably embarrassment) for the child.

In addition to MOPs, Schank also proposes the notion of TOPs, or thematic organisation points. TOPs can begin to explain why it is that we see similarities between different events or experiences. For instance, some viewers may discern a similarity between the Francis Ford Coppola film Apocalypse Now and the Joseph Conrad novel Heart of Darkness. The concept of MOPs cannot explain this, whereas the notion of a TOP can. A TOP, according to Schank, is an impermanent memory structure that arises during processing. The reason for seeing similarities between Apocalypse Now and Heart of Darkness is that both share goals (exploration), conditions (brutality, imperial adventure) and features (exploitation of another culture, madness) that taken together constitute a TOP.

5.2.3 Applying schema theory in stylistic analysis

So far we have considered schema theory as a means of explaining the ways in which readers’ background knowledge is triggered for use in the interpretation of texts. Schema theory has considerable explanatory power beyond this, however. Consider, for instance, the following extract from Umberto Eco's novel The Name of the Rose:

What becomes apparent as we read extract 1 is that Adso has limited frame knowledge about what we would call ‘glasses’ or ‘spectacles’. He does not know a word to describe them and is therefore forced to describe their appearance and purpose. In effect, Adso is what Fowler (1986) describes as ‘underlexicalised’. This leads to a fairly long-winded description, in which the precondition header ‘William preferred to read with these before his eyes’ might be said to fully invoke our frame knowledge of the object Adso describes. What this also suggests is that Adso's implied addressee is not a reader such as ourselves. Adso's supposition is clearly that his addressee will not have frame knowledge of what he describes, hence the necessity of his extended description. One of the effects of this is to distance the reader of the novel from the character of Adso, since there is a substantial difference in the amount and type of schematic knowledge that Adso has.

Schemas may also be manipulated to create foregrounding effects, which in turn can lead to incongruity and absurdity, as in the following example:

The initial stage directions in this extract contain both a locale header (‘He is sitting in an office, with a placard saying “Exchange and Mart, Editor” on his desk. He points to it rather obviously.’) and an instrumental header (‘enter Bee, a young aspirant job hunter’) which are likely to trigger for the reader a JOB INTERVIEW script. What follows, though, deviates from this schema considerably. The content of Glans's turns is more appropriate for a SELLING schema, until turns 11 to 19, where Glans appears to be behaving as we would expect according to our JOB INTERVIEW script. However, Glans then deviates again, generating further absurdity. The incongruity that arises in this extract comes about because we cannot reconcile the linguistic behaviour of Glans with the JOB INTERVIEW script that is triggered at various points throughout the sketch.

Examples 1 and 2 show how schema theory might be used to account for particular stylistic effects in texts, though schema theory has also been used more generally to discuss the notion of literariness. Cook (1994), for example, has argued that literary texts are distinguished from non-literary texts by their capacity to induce schema refreshment. Cook's argument is that literary texts are ‘representative of a type of text which may perform the important function of breaking down existing schemas, recognizing them, and building new ones’ (Cook 1994: 10). In this sense, we may view certain texts as literary because they cause restructuring or tuning of existing schemas, such that we see the world differently as a result. This, clearly, is a cognitive variant of the theories of defamiliarisation which were discussed in Chapter 2 and demonstrate that a strong tradition of believing in the separateness of literature as a text-type has outlasted the move from formal stylistic to more contextualised and in this case cognitively-informed practice.

Semino (1997) draws on Cook's work to produce an ‘approach to text analysis that combines linguistic description with schema theory in order to account for how readers imagine (different types of) text worlds in reading texts’ (Semino 2001: 345). Whilst Semino explicitly distances herself from Cook's use of schema refreshment as the defining feature of literary texts, she accepts the usefulness of the distinction Cook makes between schema refreshment and reinforcement.¹ Jeffries (2001) takes issue with Cook's proposal that literary texts are primarily distinguished from other texts by their schema-refreshing function, and argues that they may also cause schema affirmation; that is, we find literary texts satisfying because they reinforce our world view by reflecting our schematic knowledge. This ‘thrill of recognition’ (Jeffries 2001: 340) is significant in many cases, she argues, because individual readers may have a set of schemas resulting from their particular identity (as a woman, as a black person, as a lesbian, as a victim of crime and so on) and some of these schemas may be limited to the particular groups that s/he is a member of. Such ‘suppressed meanings’ (Jeffries 2001: 340) may be made explicit in literary or other texts and produce a strong reaction in a reader. Like schema refreshment, these may be quite significant for the reader, but they will not necessarily result in cognitive changes as a result. This concept of schema affirmation, Jeffries argues, applies also to non-literary texts such as propaganda and advertising; hence, the notion is unsuitable as a defining characteristic of literature.

Schema theory, then, provides a useful theoretical base both for understanding text-types and for identifying the local effects that are generated in particular texts. A broad appreciation of the notion of schemas is also useful for understanding a number of other cognitive stylistic approaches, as we shall explain in sections 5.3 and 5.4.

5.3 Figure and ground

In Chapter 2 we explained the importance of foregrounding theory for stylistics, outlining how foregrounding effects are achieved through deviation and/or parallelism at one or more linguistic levels. The psychological reality of foregrounding has been demonstrated in empirical tests (see van Peer 1980, 1986, 2007), the results of which suggest that readers do indeed attach more interpretative significance to foregrounded elements of texts. The concept of figure and ground adds a further cognitive dimension to the notion of foregrounding by providing an explanation of why we are attracted to deviant and parallel structures.

Notions of figure and ground have their origins in the work of the Gestalt psychologists of the early 1900s, and particularly in the work of Rubin (1915). Rubin proposed that our visual field is organised in such a way that we make a distinction between figures and backgrounds, and that we are able to distinguish the contours of separate objects when there is a strong contrast between their respective colours and degrees of brightness. For example, a particularly bright object will stand out against a dull background and will consequently be perceived as figural and therefore prominent. Rubin's famous illustration of figure and ground is the face/vase illusion, reproduced in figure 5.1.

Figure 5.1 Rubin's face/vase illusion

When we look at the picture, we either see two black faces in profile or a white vase-like object. The point here is that we can only see one of these images at any one time; we cannot see both the vase and the faces simultaneously. The explanation for this is that the object we choose to concentrate on (the figure) seems to us to have special properties. For example, if we choose to concentrate on the faces then they appear to the viewer to have form and structure, whereas the white ground seems formless and unstructured. Alternatively, if we concentrate on the white space then that appears to have contours that allow us to discern the figure as a vase. The figure appears to be in front of the ground and we therefore perceive it as being more prominent. Research also suggests that we attach more importance to figures, that they are memorable and that they are likely ‘to be associated with meaning, feeling and aesthetic values’ (Ungerer and Schmid 1996: 157).

5.3.1 Image schemas

It is relatively easy to see how the concept of figure and ground as employed in the visual arts may be extended by analogy to language, and how, in so doing, the notion of a figure equates to the linguistically foregrounded elements of texts. It is also the case that textual figures may be dynamic; that is, they may be imbued with a sense of movement. This is explained by the cognitive notion of an image schema.

Stockwell (2002a: 15) makes the point that in prose fiction, characters are figures against the ground of the story's setting. He explains that we can view them as being figures ‘because they move across the ground, either spatially or temporally as the novel progresses, or qualitatively as they evolve and collect traits from their apparent psychological development’ (Stockwell 2002b: 16). Within a fictional world, of course, it is not just characters who constitute figures, but other objects too. Stockwell goes on to explain how movement is prototypically represented in the verb phrase, through verbs of motion, and/or by locative expressions of space and time, realised through prepositional phrases – for example, ‘over there’, under the table’, etc. Key to our understanding of movement in texts are image schemas, defined by Ungerer and Schmid as ‘simple and basic cognitive structures which are derived from our everyday interaction with the world’ (1996: 160). Essentially what this means is that as a result of repeated experiences of certain concepts, we form a schema for these in the same way that we have schemas for people, places, objects and situations (see section 5.2). So, with regard to movement, one of the image schemas we have is of the locative expression OVER/UNDER. This arises out of our repeated encounters with objects moving over other objects. (There are other image schemas of movement, of course. Stockwell 2002b: 16 lists some of these as JOURNEY, CONDUIT, UP/DOWN, FRONT/BACK and INTO/OUT OF.) In this image schema the figure is referred to as a trajector, and the ground that it is moving over is called the landmark. And as a trajector moves over a landmark, it follows a path. Consider the following extract from Milner Place's poem ‘Favela’:

Here, the clouds are the trajector following a path above the sea, which constitutes the landmark. All OVER/UNDER image schemas make use of this basic conceptual structure.

5.3.2 Figure and ground in stylistic analysis

Having established an understanding of the concepts of figure and ground and image schemas, we might now examine how these concepts can be employed to help us understand and interpret longer texts. To do this, we will consider how the concept of figure and ground is useful in explaining the mechanism by which a reader might make sense of the following poem by Edward Thomas:

In this poem the speaker describes his² favourite spot in a farmyard, evoking a nostalgic image of a place neglected and overgrown with nettles. As we read, we are likely to picture the scene being described and we are likely to see it in a particular way. For example, it is probable that we will picture the nettles as being in the foreground of the scene. We might picture the harrow, the plough and the roller to be behind the nettles, only barely visible. Why is it that we are likely to ‘see’ the scene in this way? Obviously, this is in part to do with the propositional content of the phrasal verb ‘cover up’, but the concept of figure and ground can also help us to explain why it is that we imagine the scene in the way that we do.

In terms of figure and ground, the nettles can be seen as the figure in the poem for a variety of reasons. The noun phrase ‘tall nettles’ is foregrounded as it is the title of the poem. It is also the grammatical subject of the first sentence. We therefore read this noun phrase twice as we begin reading the poem, and because it is unusual to have the same noun phrase repeated in immediate succession, it is likely to attract our attention, causing the object to which it refers to be perceived as figural. The items that the nettles cover up, the rusty harrow, the worn out plough and the stone roller, are in the object position in the sentence, and are preceded by an adverbial that further distances them from the subject of the sentence. This grammatical structure corresponds with the propositional content of the phrasal verb ‘cover up’ and helps to evoke an image of the nettles being in front of the farm implements. As a result, we are likely to picture the nettles as being in the foreground of the scene. In effect, the ‘tall nettles’ constitute a trajector moving over several landmarks – the rusty harrow, the worn out plough and the stone roller. The way that we ‘see’ the scene is governed by the OVER image schema that we have.

Figures and grounds, though, are not necessarily fixed across the course of a whole text. Instead, as we read, other elements of a text are likely to catch our interest and shift our attention. In the case of this poem, other trajectors appear and follow paths across other landmarks. In the fourth line of the first stanza, we are likely to shift our attention on to the image of the elm butt. This is a new grammatical subject, and likely to attract our attention because it is a new element of the scene. ‘Newness’ plays a major part in making something figural in a text. Additionally, phonological parallelism help to foreground the image of the elm butt. The adverb ‘only’ and the noun ‘elm’ both contain lateral and nasal sounds. Parallel structures like this set up a relationship of either opposition or equivalence. In this case it would appear to be a relationship of equivalence, and this foregrounds the relationship between the two words and emphasises the fact that the elm butt is the only object not entirely covered by the nettles. This is contrasted with the alliterative relationship between the object noun ‘nettles’ and the adverb ‘now’, both of which begin with the nasal consonant /n/. Again we can see a relationship of equivalence between ‘nettles’ and ‘now’. However, there is a wider parallel structure at work within this clause, at both the level of phonology and grammar. Both the subject and the object of the clause are phonologically parallel with their related adverbials. Also, we can note that before the predicator there is an adverbial (‘only’) followed by a noun phrase (‘elm butt’), and after the predicate this structure is reversed as we encounter a noun phrase (‘the nettles’) followed by an adverbial (‘now’). This parallelism sets up a relationship of opposition between the subject and the object of the clause, which we might see as a figure/ground relationship. This throws into relief the image of the elm butt and focuses our attention away from the nettles and on to this object, thus changing the figure in the poem.

In the second stanza, the syntactic deviation that arises as a result of the grammatical object being placed first in the clause works to foreground the image of ‘this corner of the farmyard’. One way of imagining the scene that is conjured up by the poem is to think of it in filmic terms. Up until this point it seems that we have had a shot of the tall nettles, followed by a shot of the elm butt. We now pull out from a close-up of the elm butt to a wider view of the corner of the farmyard as a whole, almost as if a camera were pulling back on the scene being shot. ‘This corner of the farmyard’ thus replaces the ‘elm butt’ as the figure, and the image of the elm butt is backgrounded, along with the image of the tall nettles.

As we continue to read, the figure changes again. This time our attention is focused on ‘the dust on the nettles’. The fronting of the adverbial ‘As well as any bloom upon a flower’ causes us to pay particular attention to the clause ‘I like the dust on the nettles’ (notice that this is because we would expect this clause, prototypically, to come before the adverbial). Within this clause, we can also notice that a figure/ground relationship emerges between the image of the dust and the image of the nettle. This can be explained by the locative prepositional phrase which suggests that the nettle forms the ground on which the dust settles. Consequently we see the dust as the figure. The dust, then, is the trajector here and the nettle becomes the landmark onto which the trajector moves. If we return to the film camera metaphor, it would appear that what we experience at this point in the poem is another close-up shot, this time of the nettle leaves.

The figure/ground distinction is a useful addition to basic foregrounding theory as it allows us to explain more clearly how we as readers shift our attention between various parts of a text as we read. It also helps to address the common criticism that stylistics is too concerned with formal features of language and does not take into account what real readers do when they interpret texts. We will continue in this vein in the next section, where we shall examine cognitive approaches to the analysis of metaphor. This move from examining figure and ground to looking at metaphor is especially pertinent, given Stockwell's (2002a: 105) point that much of the language we use when discussing figure and ground is itself metaphorical. The underlying patterns of thought behind such terminology should become clearer once we have discussed the way in which we as readers process figurative language.

5.4 Cognitive metaphor theory

One of the most influential aspects of cognitive stylistics has been cognitive metaphor theory, developed initially by Lakoff, Johnson and Turner (see Lakoff and Johnson 1980, Turner 1987, and Lakoff and Turner 1989) and subsequently developed by, amongst others, Semino et al. (2004), Crisp (2002), Steen (2007) and Semino (2008). Cognitive metaphor theory proceeds on the basis that metaphor is not limited to literary texts but is a pervasive phenomenon in all text-types. Furthermore, it asserts that metaphor is not merely a feature of language but a matter of thought which is central to our conceptual system and the way in which we make sense of ourselves and the world we live in.

A basic definition of metaphor is that it is the practice of talking about one thing as if it were another, on the grounds that there are some notional similarities between the two entities. The following simple example from a Tony Harrison poem illustrates clearly a prototypical metaphor:

Using the terms of traditional rhetoric, ‘death’ is the tenor of the metaphor (i.e. that which is being discussed), whereas ‘a debt’ is the vehicle (what Leech (1969: 151) calls the ‘purported definition’). Clearly the meaning that is being conveyed here is that death has some of the qualities we associate with debts – perhaps in the sense of being unwished for, being something inescapable or something to be feared. These are what traditional rhetoric would refer to as the grounds for comparison. Leech (1969: 156) describes metaphor as ‘covert comparison’, contrasting this with the overt comparison of simile, where the comparison between the two entities is made explicit, as in the example below:

However, the traditional distinction between metaphor and simile is one that only exists in formal grammatical terms. Both figures of speech (to use the generic term from rhetoric) involve the comparison of two separate entities; fundamentally, they achieve their effects via the same conceptual process. From this it becomes clear that a traditional rhetorical approach to metaphor is limited to providing descriptive categories for different types of metaphor. Such an approach offers no insight into the cognitive basis of metaphors, nor the means by which we process and make sense of them. For this we must turn to cognitive metaphor theory.

The pervasiveness of metaphor in all discourse types has been one of the major findings of cognitive metaphor theory. In addition, we have noted that cognitive metaphor theory makes the point that metaphor is not just a matter of language but a matter of thought. To see more clearly what is meant by this, consider the following conventional metaphors:

Cognitive metaphor theorists would note that all these sentences are different linguistic instantiations of the same underlying metaphor. Each sentence describes life as if it were a journey that we make. In cognitive metaphor theory this underlying metaphor is called a conceptual metaphor. Conventionally, conceptual metaphors are written in small capitals, as in LIFE IS A JOURNEY. All conceptual metaphors consist of a target domain (that which is being discussed; equivalent to the tenor in traditional approaches) and a source domain (the ‘source’ of the metaphor; similar to the vehicle). In the above example, LIFE is the target domain and JOURNEY is the source domain. When we interpret conceptual metaphors we map concepts from the source domain onto the target domain. For example, when we use the conceptual metaphor LIFE IS A JOURNEY, one of the mappings that we make between the source domain and the target domain is to think of the person living the life as a traveller. For instance, we might say ‘I feel as if I’m going nowhere!’, or ‘He's getting ahead of himself.’ Other mappings (along with corresponding examples) are:

What should be apparent from these examples is that conceptual metaphors are cognitive structures that underpin our metaphorical use of language. Since cognitive metaphor theory makes clear the connection between language and thought, it has the capacity to be a useful analytical tool for cognitive stylisticians, not least because of its explanatory as well as descriptive capability. For example, applying cognitive metaphor theory can show up ideologies that underlie the surface form of the text. This will be apparent in the following example:

Clearly, the notion of Mr Blair being attacked is not meant to be interpreted literally. Instead, ‘attacking’ is a conventional metaphorical way of describing the action of criticising someone. Consider a couple of similar examples:

There is a similarity to examples 12 to 14 because they all share the same underlying (and common) conceptual metaphor: ARGUMENT IS WAR. In this conceptual metaphor, concepts from the source domain of WAR are being mapped on to the target domain of ARGUMENT. Being criticised is likened to being shot at (‘coming under fire’) and supporting someone's argument is likened to physically defending them (‘leaping to Mr Blair's defence’). When we talk about arguments and arguing we use the same kind of language as when we talk about war and fighting. The conceptual metaphor ARGUMENT IS WAR is deeply entrenched in (at least) Western society and the implications of using this metaphor are significant. Cognitive metaphor theorists have begun to see these patterns as evidence that, far from being inactive, these conventional metaphorical ways of talking about particular activities and practices suggest that we conventionally think about these things in particular ways. With regard to ARGUMENT IS WAR, Lakoff and Johnson (1980: 4) explain that ‘Many of the things we do in arguing are partially structured by the concept of war.’ We can therefore begin to see how the way we use language reveals a lot about how we conceptualise the world around us. When we use the metaphor of war to refer to arguments what we’re really doing is showing how much we think about arguing as being like war. Fairclough (1992: 195) makes the same point when he claims that ‘the militarization of discourse is also a militarization of thought and practice’. This is a disturbing concept, and perhaps the very fact that we have these underlying conceptual ideas about how the world works begins to explain why concepts like diplomacy and international relations are so very difficult to manage (see Goatly 2007 for an extended discussion of the capacity of metaphor to shape ideology).

5.4.1 Novel conceptual metaphors

Cognitive metaphor theory, then, provides a means of describing and explaining the prevalence of conventional linguistic metaphors in all discourse types. However, it does not ignore the more obvious kinds of metaphor typically found in literary language. The new and unusual metaphors that we often find in poetry may be explained using the same principle as cognitive metaphor theory uses to explain conventional linguistic metaphors. Underlying all unusual linguistic metaphors will be a novel conceptual metaphor. Again, though, novel metaphors are not restricted to literary language, and as with the examples of conventional metaphors that we have examined, using a novel conceptual metaphor reveals something of the way in which the user conceptualises the world. Consider the following example from a newspaper article reporting Air Marshal Brian Burridge's experience of fighting in the Gulf War of 2003/4:

In example 15 concepts are mapped from a novel source domain onto the target domain of war. The novel conceptual metaphor that underlies the linguistic metaphors that Burridge uses is WARFARE IS MUSIC. Modern warfare is described as jazz, meaning perhaps that it is improvisational and that there are no hard and fast rules. This, for many people, would be quite a disturbing metaphor, since music is more commonly perceived as an art form and participation in music as a pleasure. Here, though, the domain of music is being used to talk about war and arguably the underlying conceptual metaphor gives warfare a frivolity and a licence that it shouldn't have. This could easily be seen to detract from the gravity of war and consequently Air Marshal Burridge lays himself open to criticisms of being gung-ho and inconsiderate of the suffering that war causes.

5.4.2 Types of conceptual metaphor

So far we have shown in general terms how conceptual metaphors underlie particular linguistic metaphors. Lakoff and Johnson (1980) distinguish a number of different types of conceptual metaphor. While we do not have the space to discuss all of these, below is a summary of some of the most prominent.

The CONDUIT metaphor

Much of our metalinguistic capability uses what has been referred to as the CONDUIT metaphor. As Lakoff and Johnson put it, ‘The speaker puts ideas (objects) into words (containers) and sends them (along a conduit) to a hearer who takes the idea/objects out of the word/containers’ (Lakoff and Johnson 1980: 10). The CONDUIT metaphor is thus an overarching metaphor for the following conceptual metaphors:

Some linguistic instantiations of the CONDUIT metaphor are:

Orientational metaphors

The metaphors that we have examined so far are, Lakoff and Johnson note, ‘structural metaphors’ (1980: 14). That is, they structure one concept in terms of another. Orientational metaphors are somewhat different in that rather than structuring a concept in terms of another, they provide a concept with a spatial orientation. The category of orientational metaphors begins to explain many otherwise odd expressions that we commonly use in language. Why, for example, do we talk about being in high spirits when we’re happy but feeling down when we’re depressed? And why do we talk about climbing a career ladder or being at the bottom of the class? The answers to these questions can perhaps be found in the following conceptual metaphors:

• HAPPY IS UP
• SAD IS DOWN
• CONSCIOUS IS UP
• UNCONSCIOUS IS DOWN
• GOOD IS UP
• BAD IS DOWN
• HEALTH AND LIFE ARE UP
• SICKNESS AND DEATH ARE DOWN
• MORE IS UP
• LESS IS DOWN
• HIGH STATUS IS UP
• LOW STATUS IS DOWN
• THE FUTURE IS AHEAD
• THE PAST IS BEHIND

The above conceptual metaphors certainly help to explain why we use phrases such as ‘he's at the top of his profession’, but we might well ask just why it is that such conceptual metaphors are ingrained in our consciousness. The answer to this seems to be that many of these orientational metaphors have their origins in a physical basis. For example, we talk about waking up and falling asleep because humans sleep lying down and stand up when they are awake.

Ontological metaphors

Lakoff and Johnson (1980) point out that, typically, we tend to conceive of events, activities and emotions as entities and substances. Such ontological metaphors are used for a variety of purposes, among which are quantifying (‘America has a lot of political capital invested in Iraq’), setting goals (‘Dick Whittington went to London to seek his fortune’) and referring (‘her fear of flying is a huge problem’). In each of these cases, the italicised part of the sentence is being treated metaphorically as a substance. A further ontological metaphor noted by Lakoff and Johnson is the CONTAINER metaphor. Noting that humans are territorial by nature, Lakoff and Johnson suggest that the notions of bounded objects and bounded physical spaces are prime candidates for metaphor. A boundary between one space and another may be denoted by what a person's visual field can encompass, leading to the conceptual metaphor VISUAL FIELDS ARE CONTAINERS. Typical linguistic instantiations of this metaphor include the following:

One of the points that become clear when we examine conceptual metaphors is that many of the source domains in these metaphors derive from bodily experience. In this respect we can make a connection between cognitive metaphor theory and the notion of image schemas, namely that image schemas often work as source domains in conceptual metaphors. For example, a CONTAINER image schema is working as the source domain in examples 20 to 22. However, although image schemas are important to conceptual metaphors, there appears to be some principle at work that restricts which image schemas (indeed, which schemas in general) can work as source domains for which target domains. Lakoff (1993) has described this as the invariance principle. The invariance principle hypothesises that the mappings between domains of a conceptual metaphor preserve the schematic structure of the source domain in a way that is consistent with the structure of the target domain.

To illustrate this, Lakoff and Turner (1989: 82) use the example of the conceptual metaphor DEATH IS A DESTROYER, realised in such linguistic instantiations as ‘Death cut him down’ and in the classic image of the Grim Reaper. Lakoff and Turner argue that while DEATH IS A DESTROYER is a commonly understood metaphor, a metaphor such as DEATH IS A MUSICIAN would be less easily understood. This is because DEATH IS A DESTROYER is a specific instantiation of a generic-level metaphor EVENTS ARE ACTIONS. According to the invariance principle, the generic-level schemas of the generic-level metaphor must be preserved in the specific instantiation of it. So, in DEATH IS A DESTROYER, DEATH is a specific instantiation of the generic EVENT. Our schematic conception of death is that it brings about a sudden physical change in an entity. Hence, this schematic concept has to be preserved in the source domain of the metaphor. The action of destroying preserves the schematic concept of bringing about a sudden change, so DESTROYER is an appropriate source domain. The action of playing music, however, does not bring about a sudden physical change, meaning that using MUSICIAN as a source domain seems inappropriate. Of course, while the invariance principle appears to be at work in everyday discourse, literary discourse often disregards the invariance principle in the pursuit of novel metaphor. Stockwell (2002b), for example, examines the breakdown of the invariance principle in surrealist literature.

5.4.3 Blending theory

We will end this discussion of cognitive metaphor theory with a brief consideration of a related area that emerged as a result of efforts to remedy some of the theoretical issues with the notion of conceptual metaphors. The oft-quoted exemplar of a basic problem with cognitive metaphor theory is the metaphor THE SURGEON IS A BUTCHER. Evans and Green (2006) note that this example is not easily explained by cognitive metaphor theory, since the metaphor suggests incompetence on the part of the surgeon, despite the fact that butchery is a highly-skilled profession.³ Cognitive metaphor theory cannot explain where the negative assessment of the surgeon originates from. A potential answer to the conundrum lies in the blending theory developed by Fauconnier and Turner (see, for example, Fauconnier and Turner 2002). Blending theory draws on Fauconnier's mental spaces theory (Fauconnier 1985) by proposing a model for emergent meaning that is based on the idea that meaning is generated as a result of blending elements from different input spaces. While cognitive metaphor theory proposes the notion of mapping concepts from one domain of knowledge onto another, blending theory suggests that meaning construction is dependent not necessarily on such pre-existing knowledge domains, but on temporary knowledge structures created during online processing (the mental spaces of Fauconnier's 1985 work). An integration network is a model of how these temporary structures give rise to emergent meaning. The example, THE SURGEON IS A BUTCHER is explained by the integration network in figure 5.2.

Figure 5.2 Conceptual integration network for SURGEON IS A BUTCHER (based on Evans and Green 2006: 406)

In the diagram in figure 5.2 there are two input spaces in which we can identify elements associated with SURGEON and elements associated with BUTCHER. The unbroken lines between input spaces 1 and 2 represent mappings between these two domains. The two input spaces are linked by means of a generic space, in which we find higher-order elements that are shared by the input spaces. The blended space includes some information from the two input spaces, and some that is to be found in neither input space. In the case of the SURGEON IS A BUTCHER example, the blended space contains a concept of the SURGEON and the BUTCHER as being one and the same. Only in the blended space can this be the case. The blended space also blends together the goal of healing with the means of butchery. Again, only in the blended space can this be the case. These blends generate an emergent meaning of a surgeon attempting to heal a patient using the techniques of butchery, a clearly inappropriate endeavour which leads to our conception of the surgeon as incompetent.

The value of conceptual blending for stylistics is that it can begin to account for particular stylistic effects. An example of this can be seen in the following extract from James Meek's novel, The People's Act of Love (for reasons of space we have abridged the extract):

This deeply unsettling scene culminates in Samarin's horrified realisation of what his fate is likely to be. This reaction is likely to be shared by the reader since it is at this point in the novel that we are for the first time able to recognise that the actions of the Mohican constitute metaphorical mappings. Conceptual blending theory helps to explain in explicit terms how Samarin's realisation of his fate comes about, whilst also spelling out the dilemma that Samarin faces. This is detailed in figure 5.3.

Figure 5.3 Conceptual integration network for example 23

Input spaces 1 and 2 are akin to the target and source domains of a conceptual metaphor. The mappings between these two domains are indicated by the unbroken lines between them. In the generic space are the shared concepts from the two input spaces. The blended space is where we find the emergent meaning being generated. In this space, the entities of Samarin and the pig are blended together. However, input space 1 contains one structure that has no equivalent in input space 2. This is the offer of escape from the Mohican to Samarin. Since this has no direct equivalent in input space 2 (indeed, the source domain predicates the slaughter of the pig), this causes a mismatch in the blend which gives rise to deep uncertainty for Samarin. The Mohican is offering him a way out of the labour camp but on the (so far unmentioned) proviso that if their supplies run out during their escape march, the Mohican will kill Samarin and eat him. The blended space is where Samarin's horrific dilemma – whether to stay in the camp and die or take his chances with the Mohican – is generated.

5.5 Summary and conclusions

In this chapter we have outlined a selection of key cognitive theories that may be used to explain the active role that readers take in interpreting texts. The concept that links the three theories we have discussed is background knowledge. Schema theory suggests that readers make extensive use of pre-existing world knowledge to make the processes of communication and interpretation efficient. The notion of figure and ground in language works on the basis of our applying schematic knowledge about spatial relations in our analysis of texts. And cognitive metaphor theory takes the notion of image schemas and other types of schematic knowledge and demonstrates how such knowledge structures can function as source domains in conceptual metaphors. In the case of each theory, substantial interpretative benefits can be gained from its application in stylistic analysis. What will be apparent from our discussion, of course, is that some of these theories are more fully developed than others. While there is empirical support for some of these approaches (evidence for the existence of schemas, for instance, can be found in the work of Steffensen and Joag-Dev 1984, while Gibbs 1994 has found empirical evidence for conceptual metaphors), there is still considerable work to be done in this area. Evans and Green (2006: 781), for instance, note the problems of the apparent unfalsifiability of blending theory, postulating as it does the existence of ‘the conceptual structures that it attempts to demonstrate evidence for’. Similarly, though some evidence is produced for the importance of schema theory in reading, Semino (2001: 349) claims that ‘schema theory is rather unprincipled in what can count as a schema, so that it is hard to disprove empirically’. Nonetheless, these cognitive theories are promising new approaches for stylistics and will gain in currency as they develop through empirical and analytical testing.

Further reading

Semino and Culpeper (2002) is an excellent collection of chapters dealing with recent applications of cognitive theories in stylistic analysis, as is Gavins and Steen (2003). Stockwell (2002a) provides an introductory summary of a number of cognitive stylistic approaches, including those discussed in this chapter, while Evans and Green (2006) is a comprehensive (and challenging) introduction to cognitive linguistics in general. Semino (1997) provides a detailed summary of schema theory, taking into account a number of different conceptions of it. A particular strength of Semino's summary is her concentration on the linguistic mechanisms by which schemas are triggered, as well as demonstrations of its application in stylistic analysis. Figure and ground is covered by Ungerer and Schmid (1996) and Stockwell (2002a, 2003). The classic works on cognitive metaphor theory are Lakoff and Johnson (1980), Turner (1987), and Lakoff and Turner (1989). For recent additions to this area, Steen (2007), Goatly (2007) and Semino (2008) are particularly recommended.