Vous êtes sur la page 1sur 24

Applied Linguistics 28/1: 124 doi:10.

1093/applin/aml046

Oxford University Press 2007

Critical Discourse Analysis and the Corpus-informed Interpretation of Metaphor at the Register Level
KIERAN OHALLORAN
The Open University One aspect of Critical Discourse Analysis (CDA) involves examining how metaphors in texts, particularly hard news texts (reports of very recent conicts, crimes, etc.), imply certain values. The usual theoretical basis for such analysis is Lakoff and Johnson (1980). My article shows problems with transplanting Lakoff and Johnsons discourse-level approach to a CDA register-level one. I use Lees (1992) analysis and interpretation of what he identies as metaphors in a hard news text as a case study to show the following: problems with how CDA prototypically draws on Lakoff and Johnson (1980) to critically analyse metaphor at the level of register. I draw on evidence from a large corpus in order to show collocational and phraseological evidence around what Lee identies as metaphors. I show how this evidence questions not only his interpretation of these expressions, but also his Lakoff and Johnson (1980) inspired analysis. In doing so, I offer the concept of register prosody as well as a corpus-based method for checking over-interpretation of linguistic data as metaphorical, in relation to regular readers of a range of registers.

1. INTRODUCTION
1.1 Background
Linguistic analysis which seeks to systematically detect and articulate how values and ideologies are represented in text is one part of Critical Discourse Analysis. (See for example: Fowler et al. (1979), Lee (1992), Hodge and Kress (1993), Caldas-Coulthard and Coulthard (1996), Chilton and Scha ffner (1997), Fairclough and Wodak (1997), Fairclough (2001).) In CDA, hard news stories are a staple for analysis, given their salience in contemporary culture. My article is embedded within the tradition of CDA and will also analyse hard news material. The following from Bell (1991: 14) provides a denition of hard news. I follow this denition in this article: reports of accidents, conicts, crimes, announcements, discoveries and other events which have occurred or come to light since the previous issue of [the] paper . . .. The opposite to hard news is soft news, which is not time-bound to immediacy. Features are the most obvious case of soft news . . .. Hard news is also the place where a distinctive news style will be found if anywhere.

CORPUS-INFORMED INTERPRETATION OF METAPHOR

One focus of CDA is highlighting how metaphors can be ideologically signicanthow metaphors can help to construct evaluation of the situations being described. For instance, Chilton and Scha ffner (1997: 222) point to the use of the argument is war metaphor in politics, for example the oppositions claims were shot down in ames, a metaphor which constitutes adversarial debate as a quasi-natural state of affairs. In taking this position on metaphorical naturalisation of thinking, CDA has incorporated Lakoff and Johnsons (1980) approach to metaphor, what has become known as conceptual metaphor theory.1 Although Lakoff and Johnson (1980) do not invoke the work of Foucault, its perspective is that of discourse in the Foucauldian senseways of talking and thinking about the world which promote dominant world views, cutting across a variety of situations in a culture. So, for example, Lakoff and Johnson (1980) make much of the argument is war metaphor as a dominant way of talking and thinking. Chilton and Scha ffners (1997) perspective in the quotation above is likewise at the level of discourse since they are referring to a way of talking and thinking in politics generally. CDA has also drawn on Lakoff and Johnson (1980) to examine metaphor at the level of register. By register, I am referring to a concept associated with systemic functional linguistics but which has been drawn on extensively in text linguistics generally. Registers are varieties of language which are typically associated with a particular situational conguration of eld, tenor and mode, (Halliday and Hasan 1985: 389). Registers are thus distinct varieties. On this denition, newspaper journalism would not count as a register since it could consist of hard news, soft news, reviews, recipes, astrology forecasts, sports reports, etc. Being at a higher-level than hard news, newspaper journalism could be seen as a genre; genres being groups of texts which perform a similar function (Wales 2001: 338).2 In the case of newspaper journalism (hard news, soft news, etc.) a central function is the imparting of up-to-date information. So, while newspaper journalism would not count as a register, on Halliday and Hasans denition above hard news would. Lakoff and Johnson (1980) do not take account of register-specic meanings with regard to metaphor since it is concerned with everyday language (see quote in Section 2.1), although it seems to be based on introspective data, invented data or data which seem to have been elicited from informants (Deignan 2005: 27). But as Deignan argues, there is the danger that informants may tend to produce examples that are rare in normal conditions (Deignan 2005: 27).

1.2 Aims 1.2.1 Orientation


When CDA draws on Lakoff and Johnson (1980) to analyse metaphor at the level of discourse, as Chilton and Scha ffner (1997) do, then this level of

KIERAN OHALLORAN

analysis is logically consistent since Lakoff and Johnsons (1980) perspective on metaphor is also at the level of discourse. But when Lakoff and Johnson (1980) is applied in CDA at the level of register this is not logically consistent. I will demonstrate this problem via a case study, an analysis by Lee (1992) of metaphor in a hard news text, an examination which is prototypical of the way CDA draws on Lakoff and Johnson (1980) to analyse metaphor at the register level. In turn, given this prototypicality, the results of my examination will have ramications for CDA more generally in indicating that when it transplants Lakoff and Johnsons approach from discourse level to register level, this is potentially problematic.

1.2.2 Use of corpora


I demonstrate problems in Lees (1992) use of Lakoff and Johnson (1980) by taking evidence from large corpora, something that is becoming more prevalent in work that takes a critical perspective on language use (e.g. Stubbs 1996; Piper 2000; Widdowson 2000; OHalloran and Cofn 2004; Orpin 2005). Corpora have also been used in metaphor analysis more generally to show ways in which conceptual metaphor theory has ignored real language data (e.g. Deignan 2005). My article is a contribution to a corpus perspective on metaphor but mainly with regard to the way conceptual metaphor theory has been used in CDA. Using a large corpus of 260 million words which includes hard news text, I show the following: how regular readers of the hard news register would be exposed to particular collocates and phraseologies around the metaphors that Lee (1992) identies, which conict with Lees analysis and interpretation. On this evidence, I show where Lee produces an over-interpretation of his linguistic data as metaphorical when it is not, in relation to habitual readers of the hard news register. It is on this basis that I regard Lakoff and Johnsons (1980) discourse-level approach as problematic when applied at register level. (I use words such as problematic in this article rather than wrong to characterise Lees analysis and interpretation since my examination of Lee (1992) is relative to only one corpus, albeit a very large one, and I do not use other methods in support, e.g., reader response studies.) Another of my aims is to offer a convenient method for CDA to help reduce over-interpretation of linguistic data as metaphorical, in other registers, in relation to the perspective of regular readers of those registers.

1.2.3 Register prosody


The concept of semantic prosody has had wide currency in corpus-based linguistics (e.g. Sinclair 1991; Louw 1993; Hunston 1995; Stubbs 1996; Channell 2000; Sinclair 2004). Here is a recent denition from Sinclair: A corpus enables us to see words grouping together to make special meanings that relate not so much to their dictionary

CORPUS-INFORMED INTERPRETATION OF METAPHOR

meanings as to the reasons why they were chosen together. This kind of meaning is called a semantic prosody; it has been recognised in part as connotation, pragmatic meaning and attitudinal meaning. (Sinclair 2003: 178) Sinclair (2004: 305) gives the example of the seemingly neutral phrase, the naked eye. Corpus investigation reveals a common phraseology, visibility preposition the naked eye, which in turn reveals a negative semantic prosody such as in too faint to be seen with the naked eye or it is not really visible to the naked eye. To call into question Lees interpretations of the metaphors he identies, I will use a concept analogous to semantic prosody. This concept, which I have termed register prosody, indicates that some prosodies have probabilistic relationships to register. This is in contrast to the non-register specic notion of semantic prosody.

2. CDA AND LAKOFF AND JOHNSON (1980)


2.1 Assumptions of metaphorical processing in Lakoff and Johnson (1980)
Since the CDA treatment of metaphor is based on Lakoff and Johnson (1980), let me outline the position of this book in a little more detail. I reproduce the following since it has been regularly quoted and endorsed in CDA as an example of analysis of metaphor: . . . let us start with the concept ARGUMENT and the conceptual metaphor ARGUMENT IS WAR. This metaphor is reected in our everyday language by a wide variety of expressions: ARGUMENT IS WAR Your claims are indefensible. He attacked every weak point in my argument. His criticisms were right on target. I demolished his argument. Ive never won an argument with him. You disagree? Okay, shoot! If you use that strategy, hell wipe you out. He shot down all of my arguments. . . . Many of the things we do in arguing are partially structured by the concept of war. Though there is no physical battle, there is a verbal battle, and the structure of an argumentattack, defense, counterattack, etc.reects this. It is in this sense that the ARGUMENT IS WAR metaphor is one that we live by in this culture; it structures the actions we perform in arguing. (Lakoff and Johnson 1980: 4) Lakoff and Johnson (1980) argue that human conceptual systems, relating to how we both think and act, are at base metaphorical. In the above,

KIERAN OHALLORAN

Lakoff and Johnson invoke a macro-concept, war, to draw together shoot down, target, etc. Their assumption here is that metaphorical processing is structured by this macro-concept. Indeed, this pattern of inferencing from an instance of a metaphor to a macro-concept is made throughout Lakoff and Johnson (1980), and is endorsed in Lakoff and Johnson (2003) and in work in conceptual metaphor theory elsewhere such as Lakoff (1987), Lakoff and 00 Johnson (1999), Ko vecses (2000), and Gibbs and Wilson (2002).

2.2 Metaphor analysis in CDA


With regard to analysis of metaphor in CDA, here are some applications of Lakoff and Johnson (1980) in CDA at the level of register as well as genre and discourse. In other words, for all these authors there is the assumption, implicit or explicit, that metaphorical processing is structured by a macroconcept: Fairclough (1989/2001) (hard news);3 Kress (1989) (school textbooks); Lee (1992) (hard news); van Teeffelen (1994) (popular novels); Patthey-Chavez et al. (1996) (erotic romances); Santa Ana (1999) (news articles); Koller (2004) (types of business English); Charteris-Black (2004) (nancial reporting, political manifestos, American presidential speeches, sports reporting). Some of Lakoffs own work has had more of a political focus, such as his examination of the political use of metaphor in the Gulf Wars (Lakoff 1992, 2003). This work thus chimes more explicitly with work in CDA.

2.3 Metaphor and discourse exclusion


Another part of Lakoff and Johnsons (1980) perspective is that dominant ways of metaphorisation can help to exclude other ways of thinking and talking in a culture, that is, exclude other discourses. Let me give an example. In British politics, the largest party not in government is known as the opposition and practices for the opposition involve sitting, adversarially, opposite the governing party in the elected chamber (the House of Commons) of the British parliament. Indeed, argumentative exchanges between the prime minister and the opposition leader, at prime ministers questions, have often been angry ones, with the parties behind them jeering and hurling abuse. This practice helps to reinforce the argument is war discourseit is not so far-fetched to say that prime ministers questions is, in Lakoff and Johnsons words, partially structured by the concept of war (see quotation above); such argumentative framing is often reinforced in press accounts.4 This discourse also helps to exclude less dramatic but potentially more rationally-based political discourse, for example where a British prime minister listens to the perspectives of different members of the House of Commons before deciding on a course of action rather than trying to defeat opposition to his or her own pre-determined perspective. Indeed, because Lakoff and Johnsons (1980) perspective is a discourse-based one, and discourses can be accompanied by reinforcing practices (e.g. oppositional

CORPUS-INFORMED INTERPRETATION OF METAPHOR

seating arrangements and hurling of abuse in the House of Commons) one might agree that, at the discourse level, dominant ways of metaphorisation can help to exclude other types of discourse. So, there would seem to be legitimacy for CDA to invoke a Lakoff/Johnson perspective for analysis of metaphor at the level of discourse (e.g. Chilton and Scha ffner 1997). But is it legitimate to do this at the level of register?

2.4 Case Study: Lee (1992)


Lee (1992) has proved to be a popular textbook, continuing to be recommended reading for courses on CDA. Lees (1992) analysis of metaphor in a hard news text is then reasonably well known. Indeed, it is a salient example of Lakoff and Johnson (1980) inspired metaphor analysis in CDA since Lee (1992) devotes an entire chapter to Lakoff and Johnsons (1980) view of metaphor. Lee comments upon a hard news report from the British newspaper, The Guardian, on 4 August 1976, concerning events in Soweto in South Africa. Here are the rst few paragraphs of the article which Lee (1992: 912) reproduces: Police open re as Soweto erupts again5 From STANLEY UYS, Cape Town, August 4 The black township of Soweto, which has been simmering with unrest since the riots on June 16 and the shooting of 174 Africans, erupted again today. At least three Africans were shot dead, according to witnesses, although police deny this. The black hospital of Baragwanath nearby was reported to be overcrowded with injured Africans. The Minister of Justice, Mr Jimmy Kruger, announced in Pretoria this evening that he is reimposing the ban on public gatherings which lapsed last Saturday. The ban will continue until the end of the month. The nightmare of many whites in Johannesburg of a black march on their city almost came true today when between 20,000 and 25,000 angry Africans began moving in procession out of Soweto towards John Vorster Square, police headquarters in Johannesburg, where they planned to protest against the detention of black pupils. Police with automatic ries and in camouage uniform headed the marchers off after they had swept through a roadblock. They allegedly red long bursts at the leading marchers and also rained a barrage of tear-gas canisters on them. A reporter said he took a dead African to hospital, and witnesses said at least two other Africans were lying dead in the veld. [My bold] ( Guardian News & Media Ltd 1976) Now Lees (1992: 93) comments: [A] feature of the rst paragraph that could be said to derive from a white perspective is the metaphorical process that treats the

KIERAN OHALLORAN

people of Soweto as some kind of natural force, specically here as a volcano which has been simmering with unrest and then erupted. This is echoed in the later report that the marchers had swept through a roadblock, like a river. Note, too, that the emotions of individuals and the actions that they give rise to are transferred onto the place where they live. It is the township that has been simmering and that now erupts, rather than the Sowetans experiencing feelings of anger and deciding to march. The effect of these processes of metaphor . . . is arguably to distance the reader from the subjects of the report. In speaking of the Sowetans as a natural force and as a place, the emotions of the people involved and the decisions which they make to engage in particular actions are eliminated from the process of interpretation. The situation is seen as resulting from some kind of inevitable set of natural laws rather than from human feelings and decisions. As can be seen above, Lee argues that the natural force metaphors distance the reader from an understanding of Sowetans as human beings who are capable of acting as agents. Lakoff and Johnson (1980), at their discourselevel perspective, infer a macro-concept, war, to encompass shot down, right on target, etc. It should be apparent that Lee too infers a macroconcept, volcano, to draw together erupt and simmer. He also infers a broader macro-concept, natural force. Lees analysis is then in line with Lakoff and Johnson (1980) in inferring macro-concepts. His interpretation relates the metaphors to natural forces specically and generally. Lee (1992: 93) goes on to say that the fact that Sowetans are represented as a volcano is, in part, due to banality of journalistic style. Since, as Lee seems to acknowledge, erupted has conventional usage in hard news, how far can one assume that a routine reader of the hard news register would come to this text and understand Sowetans in terms of the macro-concepts, natural force/volcano?

3. METHOD
3.1 Lexicogrammar
Lexicogrammatical patterns can be sensitive to register, as Halliday and Hasan (1985: 389) argue: [S]ince it is a conguration of meanings, a register must also, of course, include the expressions, the lexicogrammatical and phonological features, that typically accompany or REALISE these meanings. And sometimes we nd that a particular register also has indexical features, indices in the form of particular words, particular grammatical signals . . . that have the function of indicating to the participants that this is the register in question. . . . Once upon a time is an indexical feature that

CORPUS-INFORMED INTERPRETATION OF METAPHOR

serves to signal the fact that we are now embarking on a traditional tale. Given the above, just a prototypical lexicogrammatical fragment from a register could cue that register for a reader. Investigation of large corpora bears out the fact that there is a greater likelihood of some lexicogrammatical patterns in certain registers than others. So corpus investigation can show clearly the distinctive style of hard news (see earlier quotation from Bell (1991)).6 Using large corpora of news texts can create a sense of what regular readers of news text are conventionally exposed to. As Stubbs (2001: 20) says with regard to this issue of convention: our (unconscious) knowledge of what is probable . . . involves expectations of language patterns. Our knowledge of a language involves not only knowing individual words, but knowing very large numbers of phrases . . . and also knowing what words are likely to co-occur in a cohesive text . . . If it can be shown through corpus investigation that: (i) erupted, simmering, and swept through tend to have conventional meanings in hard news text which are bound up with their prototypical lexicogrammatical patterning in this register, (ii) these meanings are different to the meanings Lee makes in his interpretation, then this will raise doubts about his analysis of metaphor as inspired by Lakoff and Johnson (1980). In turn, this would have implications for what linguistic data CDA identies as metaphorical in other registers.

3.2 The corpus used in this investigation


To perform this investigation, I will draw upon The Bank of English, a corpus of 450 million words made up of separate subcorpora. It is particularly skewed towards newspapers, which suits my purposes. I shall use 260 million words from six newspaper subcorpora from the period 19992003: 60 million, UK The Times; 30 million, UK regional newspapers; 45 million, UK The Sun and The News of the World; 51 million, UK national news; 38 million, US News; 36 million, Australian newspapers. They are not pure register corpora since they consist of many different newspaper texts, not just hard news. However, since the Bank of English allows the investigator to expand concordance lines to ve lines of co-text, it is thus possible to establish whether the texts come from hard news or not by inspecting whether the text is in line with Bells (1991) denition given in Section 1.1. Clearly in using contemporary corpora, I am not attempting to reconstruct how a reader in August 1976 would have come to the Soweto text.

KIERAN OHALLORAN

This would be difcult to achieve since to the best of my knowledge there is no corpus of newspaper texts of comparable size (260 million words) from the early to mid-1970s, let alone a sizeable corpus of Guardian hard news texts up to August 1976. Since my purpose is ultimately to show how CDA metaphor analysis, which imports an approach from the discourse level to a register level, can be problematic, I treat the text non-historically (like Lee).

3.3 Collocation: frequency and t-score


To get an initial sense of how the metaphors identied by Lee regularly function in the hard news register, my rst step is to look at their collocates using the 260 million word newspaper corpus. I do this because, as Sinclair (2004) points out, regular collocation patterning indicates delexicalisation of words: The meaning of words chosen together is different from their independent meanings. They are at least partly delexicalized. This is the necessary correlate of co-selection. If you know that selections are not independent, and that one selection depends on another, then there must be a result and effect on the meaning which in each individual choice is a delexicalization of one kind or another. It will not have its independent meaning in full if it is only part of a choice involving one or more words. (Sinclair 2004: 20) So, for Sinclair, lexical items in collocation have a strong tendency to become delexicalised. Delexicalisation in collocation brings delocalisation7 of meaningmeaning is not then located in a single lexical item but across the collocation. By extension, evidence for regular collocational patterning in hard news would tell us whether apparent metaphors are actually delexicalised in this register (for example, is erupted, in the Soweto text, likely to involve the meaning volcano?). In all of my collocation searches in this article, I use a span of 4:4, the default span of The Bank of English; that is, I look for collocates within a span of four places to the left and four places to the right of the search term (simmering etc.), also known as the node word. I produce the raw frequencies for collocates. Comparing raw frequencies is initially useful in seeing which collocates are recurrent. However, it is difcult with raw frequencies to attach a precise level of attraction between a collocate and a node word. The statistical measure, t-score, provides this information.8 More precisely, it measures the certainty of collocation (Hunston 2002: 73) because it takes into account the size of the corpus used. The Bank of English software automatically generates t-scores in collocate searches. Since a t-score of more than 2 is normally taken to be signicant (Hunston 2002: 72), I include collocates with such a value. The advice that one should regard

10

CORPUS-INFORMED INTERPRETATION OF METAPHOR

t-scores of more than 2 as being signicant derives from the experience of corpus linguists that such words are likely to be the most interesting (Barnbrook 1996: 98). Furthermore, the larger the corpus, the more reliable t-scores will be. The 260 million word corpus that I am using is a very large corpus and so t-scores calculated are fairly reliable.

3.4 Phraseology
Recent advances in corpus investigation have thrown up many insights about the nature of phraseological language (Wray 2002; Butler 2005). Since lexicogrammatical patterning is sensitive to register, taking a phraseological approach enables me to use more syntagmatic information to check how simmering, erupted/erupts and swept through are used routinely in hard news. Inspecting phraseologies of the metaphors that Lee (1992) identies forms the second part of my investigation. I examine, in turn, each of the metaphors Lee identies.

4.RESULTS AND ANALYSIS


4.1 Simmering
The black township of Soweto, which has been simmering with unrest since the riots on June 16 and the shooting of 174 Africans, erupted again today.

4.1.1 Collocates for simmering


Collocates for simmering in the newspaper corpus of 260 million words are as follows. For all my inspection of collocates in this article, the rst number in brackets is the frequency, the second is the t-score: water (103 10.0), row (39 6.2), dispute (30 5.4), tensions (29 5.4), resentment (25 5.0), feud (21 4.6), tension (21 4.6), anger (17 4.1), conict (17 4.0), discontent (17 4.1), debate (8 2.6), rivalry (8 2.8), saucepan (8 2.8), unrest (7 2.6), violence (7 2.5), boiled (5 2.2), frustration (5 2.2), scandal (5 2.1), [volcano (2 1.4)]. There would seem to be two different semantic elds represented in the collocates: boiling water (perhaps in relation to cooking?) and dispute/ anger. There are only two instances of volcano; the t-score for volcano is non-signicant as it is less than two. So, initially, the newspaper corpus evidence would seem to call into question reading simmering in the Soweto text as connected with volcanoes. I shall look at this issue specically with regard to hard news when I examine phraseologies for simmering in 4.1.2. Water occurs as a collocate of simmering 103 times and has the highest t-score (10.0) of any of the collocates. A t-score in double gures is very

KIERAN OHALLORAN

11

signicant (Hunston 2001: 16). (In personal communication with Susan Hunston, I was told that the comment about t-scores in double gures is based purely on experienceyou dont often come across gures that high in a large general corpus). So does the collocational evidence suggest that, actually, people could read The black township of Soweto, which had been simmering, . . . in terms of water being heated? If this is the case, Lee might have been amiss in inferring volcano, but it could still be argued that Lees (1992) original interpretation is basically in accord with a water being heated reading. In other words, Soweto would still be viewed as nonhuman and thus readers would still be distanced from the subjects of the report. Indeed, it might also be argued that the large number of collocations which include water bolsters a Lakofan container metaphor view. Lakoff (1987: 383) argues that the ANGER IS HEAT metaphor (e.g. dont get hot under the collar) when applied to uids combines with the metaphor THE BODY IS A CONTAINER FOR THE EMOTIONS to yield the central metaphor of the system. He then goes on to offer as evidence a number of examples which includes simmer down!. So it could be said, from a Lakofan perspective, that the township of Soweto is being conceptually metaphorised as a container of angry uid rather than as a collection of human beings. I will return to this point shortly.

4.1.2 Auxiliary been simmering


Simmering in the Soweto text occurs in the following: The black township of Soweto, which has been simmering with unrest . . .. Following up the collocation search with a phraseological one, I searched for the pattern auxiliary been simmering. Figure 1 offers a random sample of 20 lines; the node words are right-centred to allow more left-span co-text and thus to make it easier to see the nature of the phenomenon which has been simmering. Interestingly, of the 63 instances of has/have/had been simmering, around 70 per cent are associated with the hard news register and overwhelmingly show a semantic preference for human phenomena such as violence and conict as can be seen in the concordance lines of Figure 1. Indeed, this is reected in the Soweto text by the prepositional phrase, with unrest, following has been simmering. Only two of the instances are related to water boiling and these are not from the hard news register. The reason for the very high t-score for the collocate water is because water and simmering (rather than has/have/had been simmering) commonly show up as collocates in newspaper recipes. One can say, then, that phraseological evidence for simmering, when used in the perfect progressive in the hard news register, suggests it has hardly any associations of water boiling nor indeed of volcanoes. In turn, this negates a Lakofan reading that has been simmering would trigger a conceptual metaphor in which the township of Soweto is likened to a

12

CORPUS-INFORMED INTERPRETATION OF METAPHOR

morning to demand a fairer wage. This discontent gone on since fighting last May. The dispute are agitating for action. <p> The union dispute thority to withhold payment? The seating dispute the excesses of the war against drugs A row between the Israeli authorities and the EU most notorious fans in the game and ill-feeling been received. A low-level separatist insurgency but a row between the two Cabinet ministers chairwoman. <p> The row between Hynes and Power moves by Railtrack to dilute their safety role ause of foot-andmouth disease, the resultant row is making her sick and ruining her life. The row separatist movement against Jakarta's rule The broad Franco-Italian squabble hard feelings for Mr Horan. Leadership tension current account deficit. Why does a problem that precedented service highlights the conflict that whites didn't budge. Fueling a controversy that ut the case has also highlighted a problem which

has has has has has has has has has has has has has has has has has has has has

been been been been been been been been been been been been been been been been been been been been

simmering simmering simmering simmering simmering simmering simmering simmering simmering simmering simmering simmering simmering simmering simmering simmering simmering simmering simmering simmering

all week since Eritrea for the since June at the state for more since the teams in the for months since the over since October on. Earlier for more than for years. since the in the for years for years for years, in the

Figure 1: Sample concordance lines for auxiliary been simmering from the 260 million word newspaper corpus
container of angry uid and thus dehumanised in the eyes of the reader. More signicantly, it creates a difculty for generating macro-concept inferences from simmering in the Soweto hard news text, that is transplanting Lakoff and Johnsons (1980) approach from discourse level to register level.

4.1.3 Was simmering/simmered


I continued my investigation using the 260 million word newspaper corpus and looked at was simmering. I found only four instances, one of which was from the hard news register (anger was simmering in Afghanistan over continued US bombing . . .). Expanding my search of was simmering to the whole of the Bank of English, apart from recipes, there are a number of instances from ction (a smaller part of the Bank of English includes ction) such as in the following: He lifted her and settled her, stomach to ridged abdomen, then lled her mouth with caresses that stirred her to her toes. She was simmering when he drew back. Brushing his lips across her forehead, he murmured, I fantasized for weeks about having the countess reveal herself to me. His palms skimmed down her naked back to cup her . . .. She had ridden in sports cars with Bobby, who was truly a giant, and had never felt a bit of the wariness, eagerness, and sense of sensual risk that was simmering in her blood now. As Chase watched Nicole hesitate about moving up to front seat, he wondered what was going on in her calculating little mind.

KIERAN OHALLORAN

13

Rather than anger, simmering in the above examples relates to sensuality and sexual feelings, hardly negative in themselves. In the following example, again from ction, the simple past form, simmered, is used in a positive co-text: Ellel wishes me to announce that she is only days away from having in her custody the Gaddir childno, the Gaddir young woman. Ander simmered delightedly under their incredulous stares. Youre bbing, whispered Berkli. At the very least, youre exaggerating! No, hes not, said Mitty, gravely. Although there are only a few examples from ction, nevertheless the positive associations for the past tense (with or without progressive aspect) clearly contrast with the negative associations of the perfect progressive use, auxiliary been simmering, in hard news and neutral associations in recipes. They thus provide an indication of how the lexicogrammar of simmering can potentially realise meaning in a genre-sensitive way (in ction)9 as well as in a register-sensitive way (in recipes, hard news), for which there is much more empirical evidence in the Bank of English. Thus, given the different context-dependent values for the lemma, simmer, instead of thinking in terms of a semantic prosody, I judge instead that it is better, for hard news reporting, to think of simmer in terms of a register prosody. So, has been simmering has a negative register prosody for hard news.

4.2 Erupted
The black township of Soweto, which has been simmering with unrest since the riots on June 16 and the shooting of 174 Africans, erupted again today.

4.2.1 Collocates for erupted


As he did with simmer, Lakoff (1987: 385) refers to erupted when discussing HEAT OF FLUID IN A CONTAINER as the source domain and to ANGER as the target domain for the metaphor, ANGER IS HEAT. He gives examples such as: When I told him, he just exploded; She blew up at me; We wont tolerate any more of your outbursts. He goes on to say that this can be elaborated upon and gives the example, volcanos: She erupted. Once again, we see the characteristic Lakofan macro-inference, volcano, from an instance, She erupted, which is mirrored in Lees analysis. Lakoffs perspective on erupted would seem to bolster Lees interpretation that the people of Soweto are being metaphorised as a volcano in the hard news text. The corpus evidence, however, would suggest otherwise. Let me begin once again with a collocate search (frequency and t-score) in the 260 million word newspaper corpus. Erupted is in the past tense in the Soweto text. Since the collocate search can only look for the form erupted, this will include collocates for the past participle erupted as well.

14

CORPUS-INFORMED INTERPRETATION OF METAPHOR

The following collocates were found, shown here with their frequency and t-score: violence (214 14.5), row (190 13.7), ghting (87 9.2), fury (82 9.02), scandal (79 8.8), war (78 7.8), crisis (53 7.0), controversy (53 7.2), trouble (50 6.8), volcano (41 6.4), rioting (34 5.8), gunre (32 5.6), battle (30 5.0), riots (28 5.2), dispute (28 5.2), clashes (21 3.8), furore (20 4.4), protests (19 4.27), conict, (18 4.0), feud (14 3.7), protest (13 3.39), revolt (12 3.4), tensions (12 3.4), chaos (10 3.03), killing (8 2.4), struggle (8 2.5). This time, there are more instances of volcano (41). There are also eight instances of Vesuvius and ve of Nyiragongo. But still, the number of instances of volcano and names of volcanoes as collocates actually amounts to only around 4 per cent of the total. The largest t-score, 14.5, is for violence whereas the t-score for volcano is much lower at 6.4. Since the t-score for violence is over 10, it is very signicant. These instances of violence refer to human phenomena. So, overwhelmingly, erupted has a semantic preference for human phenomena. As with simmering, the newspaper corpus evidence initially raises doubt about reading erupted in the Soweto text as connected with volcanic meaning.

4.2.2 Erupted in the past tense


There are 2,509 instances of erupted in the 260 million news corpus. Around 90 per cent of the instances of erupted are in the past tense. A random sample of 20 lines of erupted in the past tense from the 260 million word news corpus can be seen in Figure 2. Around 75 per cent of the instances are from the hard news register, and again overwhelmingly show a semantic preference for human phenomena (e.g. ghting erupted; row erupted). Given that in hard news there is repeated evidence for a phraseology of abstract noun for human phenomenon erupted in the past tense, erupted in the past tense in hard news could well be understood prototypically, by regular readers of this register, in terms of violence, conict, etc. rather than volcanoes. Since these meanings are overwhelmingly negative, one can say there is a negative prosody for erupted in the past tense. But is it a register prosody or does erupted in the past tense carry a negative prosody across registers, that is to say a semantic prosody? One example of erupted in the past tense which has positive associations, and which is not from the hard news register, is the pub erupted. This example is from the sports report register. Here is the expanded co-text: Just as another undeserved German victory loomed, up popped Robbie Keane to score a dramatic last-minute equaliser. The pub erupted. Another heroic draw for the Irish to celebrate. I suspect Roy Keane would have been furious that they had failed to win.

KIERAN OHALLORAN

15

issue since the scandal in America in the West Bank and gun battles called by passers-by after a dispute tourists. <p> Last week's fighting at Kabul airport. Fresh fighting nto the city since renewed fighting one said. On Friday, when firing Officers slam 'cosy deal' <p> FURY <dt> 18 May 2002 </dt> <p> FURY Terminal when the gunfire new system." <p> Other protests have <p> The tension of the occasion Then seven years later the problem <p> Rumbles of controversy recently the two clubs continues. <p> The row their silence. But the scandal of the council that horrible scenes the Marines. When a political storm village of Montrado when violence were sent to Bosnia when civil war

erupted. "Behavior which might give erupted in the Gaza Strip yesterday, a day erupted between the driver of a Mercedes erupted around the old Spanish colonial erupted yesterday in northern Afghanistan erupted between Palestinians and Israelis erupted in this corner of southern Kosovo, erupted last night after a police chief erupted yesterday as an arsonist aged 15 erupted. "It echoed all over the airport," erupted in the ghetto satellite district erupted at odd moments. Speaking too long, erupted again when the couple tried to erupted over the world of famed US jock erupted last week when Andrew accused erupted this year with new allegations of erupted. They held out no olive branch, erupted in April over an American spy erupted after the murder of a Dayak boy. erupted there in 1992. They quickly

Figure 2: Sample concordance lines for erupted in the past tense from the 260 million newspaper corpus
What is interesting about this football report example is that there is no modication of erupted, for example, with a postmodier such as with joy. However, we would understand erupted here in a positive sense since football supporters are celebrating a goal. Other metonymic10 collocates, in the sports report register, such as press box, ground, room, and stadium all relate explicitly to eruptions of applause, joy, etc. in relation to the watching of a sports game. The non-metonym, crowd, is particularly marked in this usage; there are 51 collocates with a signicant t-score of 7.0 in the 260 million word news corpus. Indeed, the fact that erupted in the past tense has largely positive associations in the sports report register, but largely negative ones in the hard news register, provides evidence for seeing erupted in register prosody terms rather than semantic prosody terms. I should stress that the concept of register prosody is a probabilistic one. While the meanings around erupted in the past tense in hard news are overwhelmingly negative, there are a small number of instances of erupted in the past tense in hard news which carry positive meanings (e.g. reworks erupted and champagne corks were popping in a story about the rst day of the new millennium). Biber et al. (1999: passim) comment that the need for economy affects lexicogrammatical choices in news given the need to save space in hard news and maximise what is novel. This is why Biber et al. (1999: 477) argue the short passive is common in news, (e.g. Doherty was arrested in New York in June). Extrapolating to the use of erupted in the Soweto text, one might say that because erupted in the past tense carries a negative register prosody in hard news, its use allows the compressed meaning-making that there has been a dramatic initiation of violence without violence actually having to be mentioned. In the Soweto text, it could thus be argued that erupted

16

CORPUS-INFORMED INTERPRETATION OF METAPHOR

serves a textual function in systemic functional terms (Halliday and Matthiessen 2004). Similar things could be said with regard to erupted in the sports report register. Here erupted would seem to have a positive register prosody and so communicate joy without it having to be inscribed in the text. In tune with Lakoff and Johnson (1980), Lakoffs (1987: 377415) case study of the metaphorisation of anger is register non-specic. His examples too relate to everyday language and seem to be based on introspective data, invented data or data elicited from informants. Moreover, most of the sentence examples with metaphors in Lakoff (1987) (and Lakoff and Johnson 1980) have pronouns as subjects just as in She erupted (and like the argument is war examples in 2.1). But in the whole of the Bank of English, the majority of the instances of erupted do not have subject pronouns. (There are only eight instances of Lakoffs example of She erupted). As with the 260 million word newspaper corpus, the overwhelming majority of instances of erupted in the whole of the Bank of English relate not to volcanoes but to negative human phenomena, and human phenomena represented lexically rather than human beings represented through subject pronouns. The phraseological approach as afforded through corpus techniques of investigation shows up the problems with concocting examples, as well as those arising from failing to consider register specicity.

4.3 Erupt(s)
(headline) Police open re as Soweto erupts again I move on to looking at the present tense form, erupts, as this is the form in the headline. Since the results are similar to those for erupted, my coverage here will be briefer than in Section 4.2. Erupts has collocates in common with erupted, with the highest co-occurrence for erupts being violence (38 instances; t-score 6.1). As with erupted, there is a strong semantic preference for human phenomena. In contrast, there are 10 instances of volcano collocating with erupts but the t-score is at borderline signicance at 1.8. There are 1,468 instances of erupt(s) in the present tense. 36 of these instances occur in headlines in hard news. In contrast, erupted occurs in only two headlines and neither of these is from hard news texts (one is from soft news and the other is from a letters page). Volcano makes up 5 per cent of the total lexical collocates of erupt in hard news text bodies and 17 per cent of the total lexical collocates of erupt in headlines. This gure is still only 17 per cent; collocates of erupt(s) in headlines overwhelmingly have a semantic preference for human phenomena such as disputes. The prosody is usually a negative one, as with erupted. In sum, and similar to the evidence for erupted, corpus evidence suggests erupt(s)

KIERAN OHALLORAN

17

in the Soweto headline is not likely to be associated with volcanoes for regular readers of the hard news register.

4.4 Eruption(s)
To obtain a better sense of phraseological behaviour of the (broadly dened) lemma erupt, let me now go a little further and compare the results for erupt(s) and erupted with those for the noun form, eruption(s), in the 260 million word newspaper corpus. Interestingly, collocates for eruption(s) are overwhelmingly connected with volcanoes or related geological phenomena: Eruption: volcanic (68 8.2), Nyiragongo (18 4.2), violence (16 3.9), volcano (15 3.8), Vesuvius (9 3.0), lava (5 2.2), scientists (5 2.2), tremors (5 2.2), violent (5 2.2). Eruptions: volcanic (59 7.6), earthquakes (9 3.0), ash (5 2.2), violence (5 2.2), volcano (5 2.2), volcanoes (5 2.2). The relatively high frequencies and t-scores for volcanic provide evidence that eruption(s) is much more likely to have meanings associated with volcanoes in news than erupted and erupt(s). (The phrase violent eruption(s) refers in the main to volcanic disturbances.) To further explore the phraseological behaviour of eruption(s), I looked more generally across the 450 million word Bank of English, which also contains academic texts amongst two book subcorpora. With this wider exploration, I found that eruption(s), either in hard news or academic texts, is predominantly used to refer to volcanoes. Here is one example from an academic source: Both these methods act as geochemical stopwatches, which are reset to zero in the rocks that are formed from volcanic eruptions. Potassium has been given the symbol K, from the Arabic kali (alkali). It is one of the commoner elements in the earths crust and indeed in our own bodies. Overall, in contrast to erupt(s) and erupted, there is a much greater tendency for eruption(s) to have meanings associated with volcanoes and in a way which would appear to be not so register-specic. It would seem also that in hard news the semantic extension of eruption(s) is much more restricted than for erupted and erupt(s). This is also reected in only two instances of a positive meaning for eruption, that of applause. In other words, delexicalisation of eruption(s) in collocation is less likely to happen in hard news than with erupt and erupted. This has an interesting corollary: if the writer of the Soweto text had chosen eruption in describing Soweto, then the corpus evidence suggests that volcanic meanings would be more likely to be associated with Soweto

18

CORPUS-INFORMED INTERPRETATION OF METAPHOR

than if he had used erupted. This would then have chimed with Lees (1992) interpretation.

4.5 Swept through


Police with automatic ries and in camouage uniform headed the marchers off after they had swept through a roadblock.

4.5.1 Had swept through


Lee (1992) interprets swept through as metaphorising the Sowetan marchers as a natural force like a river with the agentive element downplayed as a result. This is a very revealing analysis and interpretation. It is a macro-inference and so is in line with Lakoff and Johnson (1980). But, in another sense, it is not in line with this book. Since Lee takes a Lakofan/ Johnsonian approach, I would expect him to generate a macro-inference to a source domain, perhaps brooms, or cleaning equipment or something else which is connected to sweeping. There are 38 instances of had swept through in the whole of the Bank of English. There is only one instance of the broom meaning of swept through in the street cleaners had swept through. All the other instances of swept are delexicalised, that is there are no associations of brooms, cleaning equipment, etc., including eight instances that occur in the hard news register (e.g. . . . the u epidemic which had swept through his squad . . .). (Had) swept through is something akin to a phrasal verb with a phrasal meaning of rapid movement. In earlier sections, it was found that (has been) simmering, erupted, and erupt(s) collocate with negative human phenomena actualised by an abstract noun in grammatical subject position. What was not found in signicant numbers were human agents as common collocates. However, in ve of the instances for had swept through in hard news, human agents collocate with had swept through. Here are two such co-texts: That handed the whole of the key Takhar provinceincluding the Talibans main garrison town of Taloqanto the Alliance. But the Alliance army had swept through other provinces. As each fell, trapped Taliban soldiers raced into two mountain ranges near the Tajikistan border for a last stand. On Sunday morning, Serbian special forces wearing white jumpsuits and black masks that showed only their eyes had swept through this Kosovo village, breaking down doors and demanding to see the young men. Eight of the men they found were marched, hands on heads, into a narrow gully in the piney woods . . .. In addition to the meaning of rapid movement, meanings of determination and decisiveness are taken on when the grammatical subject is

KIERAN OHALLORAN

19

a human agent. It must be said, however, that for had swept through, there are only a small number of instances in the Bank of English and so there is a danger of generalising too far beyond this data (which is why I have not generated t-scores). It might, then, be better to have a broader perspective by looking only at swept through, rather than just had swept through, across news corpora in The Bank of English.

4.5.2 Swept through (with or without auxiliary)


In the 260 million word newspaper corpus, I found 437 instances of swept through. Around 10 per cent of these have human agents as in the following: We were here for the sole purpose of expelling terrorists from the country and establishing a government that would not harbour terrorism. Before arriving in Kabul, Mr Rumsfeld swept through the capitals of all three Caucasus republics and predicted an end to US sanctions, and military help for Azerbaijan and Armenia. However, about 30 per cent of collocates relate to natural forces. Around 50 per cent of these collocates (i.e. about 15 per cent of the total 437 instances of swept through) semantically prefer the natural force of re; there are 42 instances of re, 10 instances of blaze, and 8 instances of ames with signicant t-scores of 6.5, 3.2, and 2.8 respectively. Floodwater, in contrast, just about receives a signicant t-score of 2.0. But there are only two instances of oods, its t-score being non-signicant at 1.4. There are no instances of river(s). Since re sweeping through is substantially more represented than river(s)/ ood(s)/etc. sweeping through, and given the t-scores for re, blaze, and ames, the corpus evidence is in conict with Lees intuition of river. But does this mean while Lees choice of specic natural force, river, is problematic, that if he had chosen re his interpretation would have had more validity? In other words, does the corpus evidence suggest that the (Sowetan) marchers referred to in after they had swept through a roadblock could be understood by readers of hard news in terms of the particular natural force of re? On the basis of the hard news corpus evidence, there is indeed the possibility that swept through is partially imbued with the meaning of re. But this still would not debar the human agency of the (Sowetan) marchers. This is because readers of the hard news register will also have been exposed to human agents such as armies, troops, etc. sweeping through villages and the rest, that is in cases where the extra meaning of determination and decisiveness can be activated; there are seven instances of Israeli as agents with a t-score of 2.6; six instances of forces as agents with a t-score of 2.4. Readers will also be aware of swept through collocating with human agents in sports reports; as in this example from the Bank of English: Hewitt swept

20

CORPUS-INFORMED INTERPRETATION OF METAPHOR

through without a hitch and put on a ruthless performance to win 61, 63, 62. In sum: on the basis of the corpus evidence, the (Sowetan) marchers sweeping through a roadblock could, for regular readers of hard news text, have some particular natural force meaning.11 This is in the sense that the marchers are unstoppable as are some res. Nonetheless, given the evidence mentioned above, it is unlikely that readers will not understand that the marchers are human agents.

5. DISCUSSION
I have shown how, in the hard news register, has been simmering, as well as erupted in the past tense, have a semantic preference for human phenomena, rather than for volcanoes, and carry a negative register prosody. The same is true for erupt(s) although not to the same degree. However, there is evidence that eruption in collocation is much more likely to carry meanings associated with volcanoes inside and outside the hard news register. So, across different forms of the lemma, erupt, there would seem to be a cline of delexicalisation from eruption(s) to erupt(s) to erupted. This extends Louws point (cited in Sinclair 2004: 198) that literal and gurative are points close to the extremities of a continuum of delexicalization since the corpus evidence suggests: (i) a cline of delexicalisation can be related to lexicogrammar; (ii) delexicalisation (e.g. of erupted and erupt(s)) can have a strong afnity with register. Further corpus exploration may identify that context-dependent prosodies can operate at the level of genre as well (i.e. a genre prosody). Looking overall at Lees interpretation of simmering, erupted, and swept through, there is inconsistency in the way he identies the linguistic data as metaphorical. The corpus evidence shows that Lee has a lexicalised interpretation of both simmering and erupted but has a delexicalised reading of swept through (in not relating the data to a broom, etc.). So it is not only Lees analysis and interpretation of individual bits of data that is problematic but also how he makes overall interpretation of the data. Widdowson (2004) uses the term pretextual to describe how certain scholars, particularly some critical discourse analysts, interpret texts in a way which corroborates their values while implicitly assuming that target readers will understand the text in the same way. I have also shown how corpus evidence is useful in avoiding: (i) pretextual metaphorical lexicalisation of textual data (ii) producing a misleading overall interpretation of textual data identied as metaphorical from the perspective of readers who have been routinely exposed to the texts register. I have shown how difculties can arise in transplanting Lakoff and Johnsons (1980) approach to metaphor from discourse level to a register level. This raises the prospect that importing Lakoff and Johnsons discourse level

KIERAN OHALLORAN

21

analysis into other register-based critical analyses of metaphors is potentially problematic. I offer the method used above as a convenient way for CDA to help reduce the following: over-interpretation of linguistic data as metaphorical, in relation to regular readers of a range of registers. Indeed, the method is straightforward enough to be used by students in seminars (with an appropriate and large corpus resource), either on texts of their own choosing or on texts previously analysed by critical discourse analysts together with their interpretations. Finally, I would argue that critical discourse analysts cannot automatically claim that metaphors are lexicalised without reference to empirical evidence, something which, in invoking Lakoff and Johnsons discourse-based perspective, they often take as a matter of course. To be as conclusive as possible, it must be said that a combination of corpus inspection and reader response exploration is needed. This article also offers a method for developing constrained hypotheses about how readers are likely to interpret what might initially appear to be metaphors. In being constrained, such hypotheses are more likely to be worth empirically testing in reader response studies. Final version received March 2006

ACKNOWLEDGEMENTS
Many thanks to the anonymous reviewers for perceptive and useful comments on this article. Special thanks as well to Sarah North for the same. I am grateful to Guardian News and Media Limited for permission to use the extract from the Guardian.

NOTES
1 In its analysis of metaphor, CDA has mostly referred to this pioneering work in cognitive linguistics rather than any subsequent work on metaphor. When in 2003 Lakoff and Johnson (1980) was reissued, it carried a new (second) afterword which, in part, takes stock of the ideas in the book and the impact the book has had in its over twenty years of circulation. There was, however, no revision of the main body of the text. 2 On the distinction between register and genre, I follow McCarthy and Carter (1994) and Wales (2001), as well as systemic functional linguistic work on genre (see, Martin and Rose 2003: 2545), where genre is treated at a higher level than register. As Wales (2001: 338) comments: It is probably easiest to see registers as particular situational congurations of linguistic resources, quite specically contextually determined; genres are larger or higher-level structures, groups of texts which are recognised as performing broadly similar functions in society. So [the genre of] advertising comprises

22

CORPUS-INFORMED INTERPRETATION OF METAPHOR

specic types which vary in choices of linguistic features, for example according to medium (TV, radio, magazine, etc.), eld (beauty products, mobile phones), tenor (target audience, etc). Register is thus a usefully exible concept: we can appreciate genres for their shared elements; but no two registers can be identical. Both registers and genres can cue a variety of discourses (in the Foucauldian sense). So the above registers and genres could potentially cue discourses of lifestyle, health, and nutrition. A text from the hard news register, on the other hand, covering the 11 September 2001 atrocities in the USA, could cue very different discourses, e.g. US nationalism, terrorism, religion, war. Moreover, different registers and genres could cue the same discourse. Hard news texts relating to the recipe register (e.g. on eating in schools) or the food preparation instructions genre all could cue discourses of nutrition. Discourses cut across registers and genres. The analysis of metaphor in a hard news text by Fairclough (1989, 2001) is endorsed by Weber (1996: 7). On how UK parliamentary discourse on one day was framed in the press using metaphors of war, violence etc, see: http://politics.guardian.co.uk/ conservatives/story/0,,1662005,00. html (accessed 21 December 05). Lee (1992) does not provide the headline of this extract. Here is one example of a lexicogrammatical pattern in hard news from Biber et al. (1999) which is marked in contrast to evidence from academic prose, ction, and conversation in English. Hard news is for Biber et al. (1999: 844): particularly marked in its use of after, where it often provides background

information about prior events, following presentation of the main story line: In a related case, four Trinity College Dublin student leaders were cleared of contempt after the society sought to have them jailed for alleged breaches of an earlier injunction restricting distribution of literature on abortion services. (NEWS). 7 Such a delocalised collocational approach ties in with connectionist modelling of language processing. The meaning of lexical items in a sentence is not compositional but distributed in connectionist representations of sentences (see McClelland et al. 1986, Rumelhart et al. 1986). 8 T-score depends on a number of calculations. The rst is the number of instances of the co-occuring word in the specied span. This value is known as the Observed. The second calculation is based on the null hypothesis: the co-occurring word has no effect at all on its lexical environment. In other words, its relative frequency of co-occurrence with the node word in the specied span is the same as its relative frequency in the whole of the corpus under investigation. This value is known as the Expected. The nal calculation that t-score depends on is standard deviation. This calculation involves the probability of co-occurrence of the node and the collocate and the number of words in the specied span in all concordance lines. T-score is calculated by rstly subtracting the Expected from the Observed and dividing this number by the standard deviation value. 9 Given that the Bank of English only allows ve lines of co-text to be inspected, it is difcult to say whether

5 6

KIERAN OHALLORAN

23

the was simmering extracts come from highly specic types of romantic ction which might be categorisable in register terms rather than genre terms. Given also the much smaller amount of ction data available in the Bank of English in contrast with news data, it would be difcult in any case to say denitively whether was simmering has a register prosody in specic romantic ctional texts. But, it may be that was simmering is strongly associated with positive connotations in romantic ction more generally speaking, i.e. it has a genre prosody. With a large corpus of romantic ction textsbroadly denedit would be

possible to investigate the possibility that was simmering instead carries a genre prosody. 10 See Widdowson (2000) for a critique of Lees (1992) perspective on metonymy in the Soweto text. 11 Although I acknowledge the prospect that swept through may carry some associations of a particular natural force, this is not an endorsement of a Lakoff and Johnson (1980) macroinference generation from a metaphor at the register level. This is because a Lakoff and Johnson macro-inference from swept through would be based on a compositional reading and thus a focus on swept.

REFERENCES
Barnbrook, G. 1996. Language and Computers: A Practical Introduction to the Computer Analysis of Language. Edinburgh: Edinburgh University Press. Bell, A. 1991. The Language of News Media. Oxford: Blackwell. Biber, D., S. Johansson, G. Leech, S. Conrad, and E. Finegan. 1999. Longman Grammar of Spoken and Written English. Harlow: Longman. Butler, C. 2005. Formulaic language: an overview with particular reference to the cross-linguistic mez-Gonza lez, perspective in C. Butler, M. Go rez (eds): The Dynamics of and S. Doval-Sua Language Use. Amsterdam: John Benjamins. Caldas-Coulthard, C. R. and M. Coulthard. (eds) 1996. Texts and Practices: Readings in Critical Discourse Analysis. London: Routledge. Channell, J. 2000. Corpus-based analysis of evaluative lexis, in S. Hunston and G. Thompson (eds): Evaluation in Text: Authorial Stance and the Construction of Discourse. Oxford: Oxford University Press. Charteris-Black, J. 2004. Corpus Approaches to Critical Metaphor Analysis. London: Palgrave Macmillan. ffner. 1997. Discourse Chilton, P. and C. Scha and politics in T. van Dijk (ed.): Discourse as Social Interaction. London: Sage. Deignan, A. 2005. Metaphor and Corpus Linguistics. Amsterdam: John Benjamins. Fairclough, N. 1989. Language and Power. London: Longman. Fairclough, N. 2001. Language and Power 2nd edn. London: Longman. Fairclough, N. and R. Wodak. 1997. Critical Discourse Analysis in T. van Dijk (ed.): Discourse as Social Interaction. London: Sage. Fowler, R., R. Hodge, G. Kress, and T. Trew. 1979. Language and Control. London: Routledge and Kegan Paul. Gibbs, R. and N. Wilson. 2002. Bodily action and metaphorical meaning, Style 36, 52441. Goatly, A. 2000. Critical Reading and Writing: An Introductory Coursebook. London: Routledge. Halliday, M. A. K. and R. Hasan. 1985. Language, Context and Text: Aspects of Language in a SocialSemiotic Perspective. Oxford: Oxford University Press. Halliday, M. A. K. and C. Matthiessen. 2004. Introduction to Functional Grammar 3rd edn. London: Hodder Arnold. Hodge, R. and G. Kress. 1993. Language as Ideology 2nd edn. London: Routledge. Hunston, S. 1995. A corpus study of some English verbs of attribution, Functions of Language 2: 13358. Hunston, S. 2001. Colligation, lexis, pattern, and text in M. Scott and G. Thompson: Patterns

24

CORPUS-INFORMED INTERPRETATION OF METAPHOR

of Text: In Honour of Michael Hoey. Amsterdam: John Benjamins. Hunston, S. 2002. Corpora in Applied Linguistics. Cambridge: Cambridge University Press. Koller, V. 2004. Metaphor and Gender in Business Media Discourse: A Critical Cognitive Study. Basingstoke: Palgrave Macmillan. 00 Ko vecses, Z. 2000. Metaphor and Emotion: Language, Culture and Body in Human Emotion. Cambridge: Cambridge University Press. Kress, G. 1989. Linguistic Processes in Sociocultural Practice. Victoria: Deakin University Press. Lakoff, G. 1987. Women, Fire and Dangerous Things. University of Chicago Press. Lakoff, G. 1992. Metaphors and war: The metaphor system used to justify the Gulf War in M. Pu tz (ed.): Thirty Years of Linguistic Evolution. Dirven on the Studies in Honour of Rene Occasion of his Sixtieth Birthday. Amsterdam: John Benjamins. Lakoff, G. 2003. Metaphor and war again posted 18 March 2003. http://www.alternet.org/story. html?StoryID15414 (accessed 15 December 2005). Lakoff, G. and M. Johnson. 1980. Metaphors We Live By. Chicago University of Chicago Press. Lakoff, G. and M. Johnson. 1999. Philosophy in the Flesh. New York, NY: Basic Books. Lakoff, G. and M. Johnson. 2003. Afterword in Metaphors We Live By. Chicago: University of Chicago Press. Lee, D. 1992. Competing Discourses: Perspective and Ideology in Language. London: Longman. Louw, W. E. 1993. Irony in the text or insincerity in the writer? in: M. Baker, G. Francis, and E. Tognini-Bonelli (eds): Text and Technology. In Honour of John M. Sinclair. Philadelphia and Amsterdam: John Benjamins. Martin, J. R. and D. Rose. 2003. Working with Discourse: Meaning Beyond the Clause, London: Continuum. McCarthy, M. and R. Carter. 1994. Language as Discourse: Perspectives for Language Teaching, Harlow: Longman. McClelland, J., D. Rumelhart, and The PDP Research Group (eds) 1986. Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Vol.2: Psychological and Biological Models. Cambridge, MA, MIT Press. OHalloran, K. A. and C. Coffin. 2004. Checking overinterpretation and underinterpretation:

Help from corpora in critical linguistics in C. Coffin, A. Hewings, and K. A. OHalloran (eds): Applying English Grammar: Functional and Corpus Approaches, London: Hodder Arnold. Orpin, D. 2005. Corpus linguistics and critical discourse analysis: Examining the ideology of sleaze, International Journal of Corpus Linguistics 10: 1. Patthey-Chavez, G., L. Clare, and M. Youmans. 1996. Watery passion: The struggle between hegemony and sexual liberation in erotic fiction for women Discourse and Society 7, 77106. Piper, A. 2000. Lifelong learning, human capital, and the soundbite, Text 20/1: 10946. Rumelhart, D., J. McClelland, and The PDP Research Group (eds) 1986. Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Vol. 1: Foundations, Cambridge. MA, MIT Press. Santa Ana, O. 1999 Like an animal I was treated: Anti-immigrant metaphor in US public discourse, Discourse and Society 10: 191224. Sinclair, J. 1991. Corpus, Concordance, Collocation, Oxford: Oxford University Press. Sinclair, J. 2003. Reading Concordances. Harlow: Longman. Sinclair, J. 2004. Trust the Text: Language, Corpus and Discourse. London: Routledge. Stubbs, M. 1996. Text and Corpus Analysis: Computer Assisted Studies of Language and Culture, Oxford: Blackwell. Stubbs, M. 2001. Words and Phrases: Corpus Studies of Lexical Semantics, Oxford: Blackwell. van Teeffelen, T. 1994. Racism and metaphor: The PalestinianIsraeli conflict in popular literature, Discourse and Society, 5/3: 381405. Wales, K. 2001. A Dictionary of Stylistics. Harlow: Pearson Education Ltd. Weber, J.-J. 1996. The Stylistics Reader: From Roman Jakobson to the Present. London: Arnold. Widdowson, H. G. 2000. Critical practices: on representation and the interpretation of text in S. Sarangi and M. Coulthard (eds): Discourse and Social Life. London: Pearson Educational. Widdowson, H. G. 2004. Text, Context, Pretext: Critical Issues in Discourse Analysis. Oxford: Blackwell. Wray, A. 2002. Formulaic Language and the Lexicon. Cambridge: Cambridge University Press.

Vous aimerez peut-être aussi