A corpus linguistic perspective on the relationship between metonymy and metaphor.

1. Introduction

In the 2002 special issue of Style on metaphor, several writers discuss the importance of embodiment in metaphor. For instance, Raymond Gibbs and Nicole Wilson discuss a number of idioms involving lexis for parts of the human body, arguing that these are "not simply rhetorical devices but cognitively underlie and help structure our understandings" (528). Their examples include "turn a deaf ear" and "to go in one ear and out the other," two phrases used to talk about someone who does not attend to what others say to him or her. It will be shown in this paper that there is a strong argument for considering these expressions to be metonymic, at least in origin, rather than purely metaphorical, an argument that can be applied to many of the expressions quoted in the literature of Conceptual Metaphor Theory. Analysis of metaphors in naturally occurring language, by Alice Deignan ("Corpus-Based Research") and Jonathan Charteris-Black, among others, has shown that expressions such as those cited by Gibbs and Wilson are frequent, both as types and as tokens. To see them now as metonymous rather than metaphorical raises the question of whether metonymy should be regarded as a more central trope than metaphor. A less strong position is to argue that expressions of this kind have elements of both metonymy and metaphor, the view that will be put forward in this paper.

I begin by describing some central examples of metonymy, then use corpus data to discuss the apparent metonymic basis for many metaphors. I argue that there is a cline from metonymy through to metaphor, following Gunter Radden (94). I also assert that some fixed points can be established on this cline and that these points are reflected in linguistic patterns. (The theoretical model outlined here is described in more detail in Deignan's Metaphor and Corpus Linguistics.) I identify three groups of figurative expressions resulting from mappings located at different points on this metonymy-metaphor cline, and I analyze concordance data for these expressions. I am particularly concerned with the relationships of synonymy, antonymy, and hyponymy between the source and target domain uses of each group, because, as I will show, these indicate the degree of coherence with which source-domain relationships are metaphorically mapped onto the target domain. The degree of metonymic basis of each mapping, I argue, has an effect on the lexical structure of the target domain, the structure having a greater degree of overall coherence where the mapping is metaphorical and being more fragmented where metonymy is partly involved.

Research in Conceptual Metaphor Theory has drawn extensively on language data, complementing psycholinguistic techniques for investigating the processing of literal and nonliteral language. However, much of the language data used has been elicited from informants and is therefore, to a corpus linguist, suspect. Corpus linguists have argued for several decades that there is a mismatch between the patterns found in elicited examples and those found in naturally occurring corpus citations. It seems that people are not good at producing naturalistic language data when asked to do so consciously (Sinclair, Corpus, Concordance, Collocation). This is one reason for using corpus data in the investigation of metaphor and metonymy where possible. Another reason is the richness of the data that are available through a large corpus and the ease with which they can be sorted and examined. The corpus used in the research described here is a 59 million word section of the Bank of English, owned by HarperCollins Publishers and held at the University of Birmingham, England. The texts in the corpus have been produced within the last fifteen years, are both written and spoken, and are mainly British English, with some U.S. and Australian English. All examples are taken from the corpus, unless otherwise stated.

2. Metonymy Possibly the best known examples of metonymy are uses such as ham sandwich, in sentences like "The ham sandwich is sitting at table 20" (Nunberg 149). Here, ham sandwich is used by waiters to refer metonymically to a customer who has ordered a ham sandwich. Corpus searches suggest that this is not an especially frequent type of metonymy but that metonymy in general is frequent in speech and writing. For example, in the following citations, university and auction house stand metonymically for people who work in those institutions.</p> <pre>

We didn't want to get into a horrible shouting match with the whole university. The clash came when the smaller auction house

attempted to put a gloss on poor figures. </pre> <p>Numerous other examples can be found where a word referring to a building or institution is used to talk about the people who work there, and it seems likely that these uses are unmarked for speakers.

Conceptual Metaphor Theory sees metonymy and metaphor as both having important roles in structuring thinking and therefore language (Gibbs, The Poetics of Mind; "Speaking and Thinking"). However, metonymy involves a different kind of mental mapping. While metaphor involves mapping between domains, metonymy is an intradomain phenomenon. For instance, the metonym university, which has a "literal" meaning of "institution for higher education," is from the same domain as the entity "people who work at the university," for which it stands in the above citation. In contrast, in one of George Lakoff and Mark Johnson's best known examples of metaphor, the word attack is taken from the source domain of "war" and is mapped onto the different domain of "argument"(4).

In Metaphor and Corpus Linguistics Deignan shows that metonyms can be divided into nonconventional and conventional. Nunberg's ham sandwich is an instance of a nonconventional metonym. Because nonconventional metonyms are not established meanings, they can be understood only with reference to context. Deignan cites a use of toothache as a metonym for "person suffering from toothache and requesting urgent attention" (48), a use that is probably interpretable only in its context, a dentist's waiting room. Gibbs, in The Poetics of Mind, shows how literature makes use of nonconventional metonyms. Among other examples, he cites a humorous passage written by Erma Bombeck, in which potential roommates are referred to by their possessions, for instance, a steam iron, stereos, and an electric typewriter, in sentences such as "Steam irons never have any trouble finding roommates" (334). Nonconventional metonyms are of interest in what they show about mental mapping, and they are often colorful. However, like many studies in corpus linguistics, this paper is concerned with the typical patterns of language and therefore focuses on conventional metonyms from here on.

Conventional metonyms include the uses of words for institutions to stand for the people who work in them, as in the examples of university and auction house, cited above. Unlike nonconventional metonyms, these expressions can be understood readily without reference to a specific context. Any preliminary examination of naturally occurring language suggests that they are extremely frequent. For instance, the following five citations are all taken from a single edition of the popular newspaper The Sunday Mirror:</p> <pre> Hospitals are there to treat the sick. Department of Trade investigations tend to be long-winded, unwieldy and drag on for years. </pre> <p>Here, hospitals and Department of Trade stand for the people who work there.</p> <pre> The label that's big in Fairtrade is Cafe Direct. ... inefficient, over-regulating, overpaid pen-pushers. Employees can expect to pick up no more than 4,000 [pounds sterling]. </pre> <p>Here, label stands for a company, pen-pushers stands for people whose work involves processing written documents, and pick up for "take, receive" in a wider sense than the purely physical act of picking an object up.

Many now conventional expressions are metonyms, such as the verb shelve, meaning to postpone plans indefinitely or cancel them, derived from the action of putting objects onto a shelf, out of the way. There are 776 citations of this verb in the section of the Bank of English corpus searched, including the following:

A new school exam for 16 year olds has been shelved by the government.

Moon's corpus study of fixed expressions and idioms shows numerous instances of words for parts of the body and clothing being used to stand metonymically for human states, behavior and actions. These include be long in the tooth, have one's eye on something, set sail, cap in hand, and hold on to someone's apron strings. While they often go unnoticed, all the examples cited here are nonetheless clear-cut examples of metonymy. Recent work within the Conceptual Metaphor tradition, however, has shown that many linguistic expressions have more complex origins. Some of this discussion is summarized in the following section, using corpus examples.

3. Combinations of Metaphor and Metonymy

3.1. The Overlap

Many metaphor theorists now believe that some of the most frequent and central metaphors have their basis in our physical experience of the world, especially in bodily sensations. Antonio Barcelona puts forward the arguments for this view. Gunter Radden is one of several researchers to point out that this proposed experiential basis suggests an overlap with metonymy (94). For instance, Lakoff and Gibbs argue that the basis for the conceptual metaphor EMOTIONS ARE TEMPERATURES is the physical sensations we experience together with emotion (Lakoff, Women, Fire and Dangerous Things; Gibbs, The Poetics of Mind). At the linguistic level, this means that a word describing the physical sensation that is associated with an emotion is used to stand for that emotion. In the following corpus citations, cold is used to stand for several emotions.</p> <pre> [He had a] very restrictive life, with his cold mother, his press secretary ... ... capable of taking a human life in cold blood The couple, who have known each other for eleven years, were to have been married last Saturday at church. Friends believe he got cold feet and did a runner. </pre> <p>If the motivation for this use is that a sensation of reduced body temperature is an aspect of being unemotional or frightened, the choice of cold could be interpreted as using "an aspect of something to represent or stand for the thing as a whole," Gibbs's definition of metonymy (The Poetics of Mind 320). This use of cold then becomes an intra-domain mapping, not cross-domain. Zoltan Kovecses uses this argument to suggest that numerous expressions, such as to have cold feet, are metonymically based. He found that the relationship of cause-effect, which produces the expression cold feet, was one of the central metonymies in his data, producing both CAUSE-FOR-EFFECT and EFFECT-FOR-CAUSE metonymies (Metaphor: A Practical Introduction 154). Feyaerts constructs a similar argument to show that the mapping KNOWING IS SEEING, which is usually talked of as a conceptual metaphor, can be regarded as a metonymy if we decide that the concept SENSORY PERCEPTION belongs in the same domain as MENTAL PERCEPTION. The argument for treating sensory and mental perception as separate domains is certainly difficult to sustain, given that there is no generally agreed, watertight understanding of the term "domain," and that visual perceptions are surely linked in a causal relationship with many instances of mental perceptions. These mappings are especially significant given that they are claimed to be universal or near universal in human cultures. Lakoff (Women, Fire, and Dangerous Things) writes of the centrality of emotions as temperatures, while Sweetser cites expressions from a large number of languages to demonstrate the pervasiveness of KNOWING IS SEEING across cultures.

It seems that some expressions that were traditionally regarded as metaphors have at least an element of metonymy. However, calling them all "metonyms" is not the best way forward, because this would put cold meaning "frightened" in the same category as university meaning "people who work at a university." A more detailed classification is needed, one that can do the following:

* distinguish clear-cut cases of metonymy such as ham sandwich, university, pen-pusher, and long in the tooth from unambiguous cases of metaphor

* split clear-cut metonyms such as the above expressions into subcategories--for instance, separating the context-bound ham sandwich from conventionally established metonyms

* account for more complex cases, such as the above uses of cold, that have traditionally been regarded as metaphors but that seem to have an important element of metonymy

3.2. Goossens's Classifications

Goossens uses his database of figurative conventionalized expressions, taken from dictionary entries, to argue that metaphor and metonymy often interact, in several distinguishable ways. He lists four potential types of interaction, two of which seem to occur only very rarely, if at all. The other two types are metaphor from metonymy and metonymy within metaphor. Metonymy within metaphor is found where "a metonymically used entity is embedded within a (complex) metaphorical expression" (336). One of Goossens's examples is the expression bite one's tongue off, where tongue stands metonymically for speech, and the expression as a whole is used metaphorically to mean "prevent yourself from speaking." Metonymy within metaphor can only describe multiword expressions, because with the exception of some compounds, a single word could not usually contain a metonymy within a larger unit of meaning. Deignan and Potter found that this category was rare in corpus data. The other main category in Goossens's classification, metaphor from metonymy, is found where an expression develops a meaning through metonymy and that meaning is then mapped metaphorically onto another domain. Goossens's examples include say something with one's tongue in one's cheek, beat one's breast, and be close-lipped. These expressions can be understood on two levels. First, they can be seen as literally true but as having wider connotations. For instance, someone may literally beat his or her breast in an expression of remorse. At a more figurative level, the literal action may be absent, and the expression may refer to the emotion alone. In this case, the expression is a metaphor rather than a metonymy but has been derived through a process of metonymy. Goossens's categories, used as the basis for a corpus exploration of the interaction between metaphor and metonymy, led to an adapted categorization, described in the following section.

4. Metaphor and Metonymy in the Corpus

4.1. Differences in Findings from Dictionary and Corpus Data Goossens's categories are extremely useful, but were developed through the study of entries from a corpus-based dictionary, which can give a different picture from direct corpus study. This is because lexicographers are obliged to filter corpus evidence of language in use, in order to select information that matches the dictionary users' priorities. Because virtually all currently available corpus-based dictionaries have been designed for learners of English, they often reflect attempts to clarify and simplify definitions and examples. Dictionary texts are thus accessible for learners, but possibly less than ideal as research data. In contrast, the unfiltered use of corpus data will often show additional patterns to those found in a dictionary and will present the researcher with a large number of citations of each word and expression, rather than one or two examples that have been pre-selected, simplified or, in some cases, invented. In this and the following sections, I discuss interactions between metonymy and metaphor from a corpus perspective. This leads to the description of a cline from "pure" metonymy, through various types of interaction, to "pure" metaphor.

In a study of metaphors and metonyms in English and Italian, Deignan and Potter investigated linguistic expressions from the source domain of the human body, using corpus data ("A Corpus Study"). They studied one thousand concordance citations for each of four items--nose, mouth, eye, and heart--and their inflections in the Bank of English, and they conducted a similar study of their Italian translation equivalents in a modern corpus of Italian. The study was admittedly very limited in the number of words studied, but it was comprehensive and systematic in its use of corpus analysis techniques. Deignan analyzed a much larger number of body-part expressions in English using similar techniques (Metaphor). Both studies showed up many examples of Goossens's metaphor from metonymy but no clear examples of metonymy within metaphor.

A typical corpus example of metaphor from metonymy is turn one's nose up at something, meaning "reject." The expression is a metonym if the action that it describes is literally true, that is, if the person referred to has actually turned his or her nose upwards. This is possible, though we cannot be sure, in the following citation, which describes a young child rejecting food. Here, her turning her nose up may be an instinctive reaction of distaste.</p> <pre> She will feed my daughter healthy tidbits from the centre's fridge if she turns her nose up at her own food. </pre> <p>If the expression is not literally true, it is an instance of metaphor from metonymy. With this meaning, the subject of the phrase does not literally turn her nose upwards, but her rejection is expressed in a different, or more abstract, way. This is probably the case in the following corpus citation:</p> <pre>

Two trees in Horace's garden demonstrate that you should never turn your nose up at inexpensive plants. </pre> <p>The use in the following dialogue from the corpus is similar:

Speaker 1: .. clothes do they ... are they important to get you friends

Speaker 2: It is to some people I reckon

Speaker 3: Mm. It shouldn't be

Speaker 2. But it is. People like react

Speaker 3: Yeah. Yeah

Speaker 2: like they turn their nose up at you or something. It's out of order but they do do it.

The researchers in both the corpus studies described above found dozens of examples of metaphor from metonymy but only dubious examples of metonymy within metaphor. Metonymy within metaphor is therefore not included in the classification developed here.

4.2. Metaphor, Metonymy, and Ambiguity

It was seen above that it was impossible to be completely certain whether turn one's nose up had a literal meaning in each citation. Ambiguity is not generally found in examples cited in dictionaries; it is likely that ambiguous examples are deliberately avoided. However, ambiguity between literal and nonliteral senses of expressions is a prominent aspect of many instances of metaphor from metonymy in naturally occurring data. The following citation illustrates this:

She had been holding her breath and hoping that the agreement would be signed.

The expression holding her breath seems to have metonymic motivation in bodily experience, because when we are worried about something, we often literally hold our breath. In this citation, the reader has no way of knowing whether the expression was intended as literally true, in which case it is an instance of metonymy because the physiological effects of an action are being described to stand for that action. If the expression was intended as figurative, that is, if the writer does not mean that the subject was literally holding her breath but is trying to tell us something of her feelings, then it is an example of metaphor from metonymy. Other expressions that relate breathing to emotions in this way include take one's breath away and breathless. Similar ambiguity can be seen in the following citations:</p> <pre>

The view from the top across two continents--Europe and Africa--takes your breath away. The players turned in some great performances in three wonderful games that left me breathless. </pre> <p>Other aspects of physical behavior generate ambiguous linguistic expressions, such as raise eyebrows and look over one's shoulder, which have several figurative meanings. The following citations show potentially ambiguous uses:</p> <pre> An insider at OK! says: "This interview has caused

quite a few raised eyebrows in the office. We reckon it is the most controversial thing our magazine has ever done." The envelope was his passport to a new life where he wouldn't have to look over his shoulder continually to see whether his former intelligence colleagues had finally managed to track him down. It sounds like the ... I mean it's the priorities of scientific work are mainly decided internally by you know the scientist here rather than somebody looking over your shoulder and saying we want to know about this. </pre> <p>Corpus searches such as those conducted by Deignan and Potter and by Rosamund Moon have shown that there are numerous other linguistic expressions from the source domain of bodily sensations, processes, and actions that can be interpreted as either literal or as figurative in some contexts. This ambiguity constitutes a problem for dictionaries, which attempt to give decontextualised lists of the meanings of words and expressions, often categorized as "literal" or "figurative."

4.3. Metonymy-Based Metaphor

It has been argued that many expressions generated through the interaction termed metaphor from metonymy can show both ambiguity of interpretation and a cline through the various possible interpretations. However, this argument does not apply to all metaphorical expressions that have metonymic grounding. I now describe a group of expressions that are not usually ambiguous, and I argue that they should be seen as occupying a different part of the metaphor-metonymy continuum: in other words, that the category metaphor from metonymy should be split more finely, using ambiguity as a criterion.

An analysis of the senses of heated exemplifies the nonambiguous group. Heated apparently realizes the conceptual metaphor EMOTIONS ARE TEMPERATURES. Of 500 Bank of English citations of the adjectival form of heated, the vast majority can be accounted for by two senses: a literal sense, describing entities which are physically warm, and a figurative sense, denoting anger or emotional tension. Almost all instances of the figurative sense modify nouns denoting speech events, such as debates or arguments. The following four citations show that it is straightforward to decide which meaning is intended, using only a very small amount of context.</p>

<pre> For instance, I found myself in heated argument the other day ... I rose at 5 a.m. to battle with heated rollers ... Morrison was behind schedule. A heated telephone exchange between ... [There are] special offers every month on heated towel rails and selected ... </pre> <p>In the first citation, it seems unlikely that any reader would interpret heated argument as meaning, literally, "hot." If prompted, a reader might infer that the participants felt warm because of the strong emotions that they experienced, but this inference does not seem to be the primary meaning, although the metonymic motivation is traceable.

The citations can also be analyzed using an approach taken by many corpus lexicographers. This approach distinguishes separate senses of a word by looking for discrete groups of collocates. In this case, if we examine the lexical words occurring most frequently to the right of heated, we can see that they fall clearly into one of two lexical sets: physical entities or speech events. In other words, unlike the expressions described in the previous section, the two meanings are not ends of a cline with an ambiguous area between; they are sharply divided. For these reasons, this figurative sense of heated is regarded as different from the expressions discussed above, where metonymic meaning shades almost indiscernibly into metaphorical meaning. Deignan found the same feature, distinctive collocates for literal and nonliteral senses, in an analysis of the most frequent collocates of warm (Metaphor and Corpus Linguistics). The term metonymy-based metaphor can be used to separate this type from the previous, metaphor from metonymy.

4.4. Summary of the Cline from Metonymy to Metaphor

I have shown that a number of linguistic expressions have features of both metaphor and metonymy to varying degrees, and I have isolated and described the major types of metonymic and metonymically motivated expressions found in my corpus studies. These are summarized in Table 1. Although boxes are used to separate off different terms and examples, they are not intended to suggest discrete categorization. There are cases that fall on the boundaries of the types described. Importantly, many linguistic expressions could fall into more than one of these boxes, depending on their use on context, so this is a classification of uses, not of linguistic expressions in the abstract.

I now use corpus data to examine some linguistic patterns of groups of expressions located at different points on this cline. First I say something about the assumptions underlying the investigations.

5. Metaphor, Metonymy, and the Lexical Structure of the Target Domain

5.1. Mapping and Lexical Relations

Although its goal is to investigate thought, not develop linguistic theory, Conceptual Metaphor Theory cannot be divorced from its linguistic implications, in part because conceptual metaphor theorists have relied so heavily on linguistic evidence in support of their work. For instance, Lakoff cites "systematicity in the linguistic correspondences" of the realizations of conceptual mappings such as LIFE IS A JOURNEY as the first of his three sources of evidence for conceptual metaphors ("The Invariance Hypothesis" 50). Studying the linguistic metaphors though which life is talked about as a journey, Lakoff found that the lexical relations holding between them echo their nonmetaphorical meanings. For instance, arrivals are mapped onto birth, and departures onto death, with the opposition between the literal meanings of arrival and departure preserved in their metaphorical meanings in the target domain. This is logical; if source and target domains are structured in the same way, then the same relations will hold between different elements of each domain, and these will be realized in the lexical relations between linguistic metaphors.

The idea that lexical relationships in one domain can motivate the development of meanings in another has also been put forward by Eva Kittay, who describes ways in which this potential for metaphorical extension of meaning is exploited in literature. She shows how Wordsworth developed the metaphor VENICE IS A WOMAN to create new meanings in his poem "On the Extinction of the Venetian Republic." Here, the source domain, a woman's life, is mapped onto the city of Venice. Early in the poem, birth is used as a conventional metaphor to describe the creation of the city. Lexical items from the same source domain, including child, maiden, and seduced, are then used in the same target domain, and their new meanings are understood through their semantic relationship to birth in the source domain.

None of the writers cited claims that all the potential metaphorical senses in a semantic field are actually found in language in use, but they do strongly imply that where several words from a source domain are used in the same target domain, semantic relations between the target domain uses will replicate those found in the source domain. The language data on which such claims are based appear to have been generated intuitively by the researchers themselves or are provided by informants, but are not generally extracted from naturally occurring language use. It was noted in the introduction that corpus linguists are skeptical of such data, and I therefore set out to examine some of the central lexicalizations of some widely discussed metaphors using corpus data. The goal was to see whether the lexical relations between metaphors in naturally occurring, nonliterary language are consistent with the source domain, as suggested by the metaphor scholarship. This was the first of two questions tackled in this set of corpus studies.

It has been argued that, unlike metaphor, metonymy does not map one domain onto another; rather it maps within a domain. This means that there is no distinction between source and target domain, so no structural mapping can happen. It thus would not make sense to look for, for example, a source domain structure involving literal university (institution) and to try to find out whether this is mapped onto a target domain structure involving metonymic university (people who work in the institution): both the literal and metonymic meanings of the word signify components of the same domain. What is of interest, however, is what happens where the mapping is not purely metonymic. This leads to the second question to be investigated here: to what extent is the structure of the source domain mapped coherently onto the target domain when the mapping is an interaction between metaphor and metonymy?

In the next three sections, I report corpus studies that investigated the lexicalizations of the metaphor ARGUMENT IS WAR and the metonymy-based metaphors HAPPINESS IS LIGHT and EMOTIONS ARE TEMPERATURES.

5.2. ARGUMENT IS WAR

One of Lakoff and Johnson's most famous examples of conceptual metaphor is ARGUMENT IS WAR. This seems to be close to the "metaphor" extreme of the scale from metaphor to metonymy, described earlier, in that the mapping involves two distinct and easily distinguishable domains: Lakoff and Johnson claim that "argument" is not a type of "war" but is essentially different. According to their data, the linguistic correspondences of the mapping are rich, consistent, and detailed. For my corpus investigation of ARGUMENT IS WAR, I took the following key words from the description in Metaphors We Live By: attack (noun, verb), defence(se), defend, shoot down. I also examined citations for the words (to)fire, battle, and guns, which emerged as significant collocates of the key terms in the corpus, as well as all inflections of these words. I examined corpus citations for literal meanings of the words, to confirm that the following semantic relations hold in the source domain: attack-defend: antonymy; attack-defence(se): antonymy; attack-shoot down/(to)fire: hyponymy. Second, I examined metaphorical uses of these words in the corpus to determine whether the same semantic relations hold in the target domain. If, as Lakoff and Johnson argue, the language of argument is metaphorically structured, the semantic relations of the target domain should reflect those of the source domain.

In all the studies described here, I examined between 200 and 500 citations of each inflection of each word, depending on how frequent the word form was as a whole in the corpus. For example, attack occurs 7,927 times, from which I selected a random sample of 500, while defended occurs 750 times, from which I randomly selected 200 citations. For highly polysemous words, a sample of 1000 citations is advisable, but for this investigation an initial scan of all citations suggested that none of the words studied has more than four or five main senses, so the smaller samples were considered adequate.

The following is an extract of the corpus concordance for nominal attack (NN at the beginning of each line means that attack has been tagged as a noun). All literal citations have been taken out:</p> <pre>

NN hey are always interpreted as 'an attack on John' when it's only

NN of disaster to launch his counter attack in the leadership crisis

NN price. When Robinson unveiled his attack, Granada's quote stood

NN politician Tony Benn, 69, for his attack on Commons privileges.

NN o monetary union. Mr Lamont's attack came as Labour leader

NN 1> and and <ZF0> and Tony Blair's attack on him. And you seemed

NN wn, with Trevor Smith leading the attack, should take another

NN Id a massive campaign against the attack on abortion. And to say NN ico Vinci, a Gedda candidate. The attack on Rossi was a sighting

NN <FOX> The Tories kept up their attack on which they've NN hen, and that idea is still under attack at every moment today. NN me buyers. In a thinly-veiled attack on the Federal Reserve </pre> <p>Five hundred citations of defence (the British English spelling) were examined in detail. Citations for defense (the US English spelling) were then checked to see if any distinct patterns appeared. None did. Many citations are accounted for by expressions such as Ministry of Defence, or are from legal contexts. Metaphorical citations from other contexts include this one:

Voices are being raised in defence of the system.

I concluded from the corpus data that nominal attack and defence are used frequently to talk about verbal conflict and that the relationship of antonymy between them echoes that found in the source domain of WAR.

Verbal attack and defend also appear to share the relationship of antonymy in the source and target domains; the following citations are typical:</p>

<pre> Ministers were attacked for withholding the truth. The teenager was defended by his mum Sue on the doorstep of their family home yesterday. </pre> <p>Two hyponyms of literal attack were studied: shoot down and fire. Both are fairly infrequent in the target domain of argument, though not so much so as to be described as nonconventional. The citations that are found are consistent with the mapping ARGUMENT IS WAR, and seem to be hyponymous to metaphorical attack, as predicted by the mapping, as in these examples:</p> <pre> Fugard has been surprised, he says, by the almost complete lack of adverse criticism, especially in America, "I expected to be shot down in flames."

People always try to shoot us down; it doesn't matter whether it's

a friendly or a big World Cup game, it seems as if we have to win

every match 5-0. The Yorkshire businessman ... fired an early warning shot. </pre> <p>It seems then that the meanings of these linguistic metaphors are consistent with a mapping of semantic relations from the source domain. Some interesting linguistic patterns emerged from this data. Shoot down is rarely used metaphorically, and when it is, the grammatical pattern tends to be different from that in literal citations: literal citations tend to take the form shoot down somebody/something, while metaphorical citations have a tendency to appear as shoot somebody/something down. Further, while the concordance of shoot down shows few metaphorical citations, those citations found to contain the expression shoot down inflames were always metaphorical. Similarly, the verb fire, meaning "shoot," is almost never used metaphorically, except for when it appears in the expression fire a warning shot, which is always metaphorical. The concordance of gun/guns showed a similar lexical development; the word is rarely used metaphorically, but in the collocation guns blazing, metaphorical citations predominate. The following citation is typical:

The Foreign Secretary came out all guns blazing yesterday.

These patterns raise questions about the development of lexical phrases unique to the target domain, questions that cannot be tackled here. Overall, however, the linguistic evidence is consistent with a mapping of source domain structure onto the target domain.

5.3. HAPPINESS IS LIGHT / SADNESS IS DARK

The second corpus analysis focused on words from the source domain of light and darkness that are used to talk metaphorically about happiness and sadness. Kovecses gives elicited and intuitive data in support of the mapping HAPPINESS IS LIGHT ("Happiness: A Definitional Effort"). Barcelona, however, discusses the other side of the metaphor, A NEGATIVE EMOTION IS DARK, and argues that it has a metonymic basis:</p> <pre>

Light is likely to arouse a feeling of confidence, safety, liveliness or happiness etc and physical well-being, which is positively valued, whereas dark tends to bring about a feeling of insecurity, melancholy and physical unease, which is negatively valued. (40) </pre> <p>Nonetheless, Barcelona considers the mapping to be metaphorical, a specific submetaphor of the mapping NEGATIVE IS DARK. The lexicographical perspective discussed above also suggests that the mapping is metaphorical, albeit metonymically based, in that for most or all figurative uses of light and dark, a literal interpretation is not possible and so there is a semantic gulf, not a continuum, between the senses. By the criteria developed earlier in this paper then, the mapping is an instance of metonymy-based metaphor.

As for the above study, I examined samples of the concordances of the central terms and analyzed the semantic relations between their literal and metaphorical meanings. This was more complicated than for ARGUMENT IS WAR, because the source domain of light and dark is mapped onto a number of target domains; those appearing to lexicalise happiness and sadness were studied in detail. There is convincing corpus evidence that dark is used to talk about deep unhappiness. Its main metaphorical use in this target domain means something like "haunted" and often describes a grim, hidden side of human behavior and personality, in citations such as the following:</p> <pre> ... but there is a dark side to him as well. Denise, 31, explores the dark worlds of mental illness, drugs, alcoholism and child abuse. </pre> <p>The "unhappy" meaning of dark also shows a strong tendency to collocate with nouns referring to periods of time such as hours, days, and years:</p>

<pre> Visiting the patients in the same hospitals where I had spent endless dark days. [His] joy at sharing so much with [her] will help him through the dark years without her. </pre> <p>There are no corresponding uses of light. Light collocates with years, but the expression light years is a hyperbolic way of expressing metaphorical distance and is unrelated to HAPPINESS IS LIGHT. Bright is sometimes used to describe happiness, but has a different collocational pattern from dark. It does share the relatively frequent collocate side, in expressions such as "look on the bright side," but the expression is not clearly antonymous to "a dark side," which generally refers to someone's personality rather than a situation more generally. Light is used to talk about lack of seriousness, in expressions such as light entertainment and on a lighter note, but it seems likely that these are mappings from another literal meaning of light, the antonym of heavy, as there is a metaphorical use of heavy meaning something like "dull" or "over-serious." Verbal light is mapped onto happiness in the phrasal verb light up:

... the way laughter could light up the faces of so many children.

But because it is a phrasal verb, this is a separate lexical item from light. It appears to be antonymous to a metaphorical sense of darken, which is also used to describe the effects of emotion on people's eyes and faces. Where metaphorical light occurs as a noun, its usual meaning appears to be mapped from KNOWING IS SEEING, through which light stands for understanding, in expressions such as shed light on and bring to light.

Gloomy, which in its literal sense is a near synonym of literal dark, is used with a meaning that is similar to, but less strong than, the metaphorical meaning of dark:

The mood is quite gloomy.

This use appears to echo the source domain structure, though metaphorical gloomy has a different range of collocates from metaphorical dark, never collocating with words used to talk about periods of time, for instance.

As Barcelona points out, there are a number of linked mappings from the source domain of light and darkness, and some words from the domain are used metaphorically with meanings that are related to but not synonymous with happiness, or are positive in a more general way. For instance, glowing is used to connote strongly positive and possibly happy feelings, and also to mean "positive" more generally:</p> <pre> There was love glowing in her eyes. [He] had no hesitation in offering a glowing reference. </pre> <p>Shadow is a frequent metaphor, and sometimes seems to lexicalise SADNESS IS DARK, in citations such as this:

Rich and poor alike live under the shadow of a disintegrating social fabric.

Although sadness is a possible interpretation, however, shadow here probably has a wider, more generally negative meaning. Other citations suggest that it connotes feelings of fear at least as often as sadness. In yet other citations, different aspects of literal shadows are mapped, although there is still the suggestion of negativity:</p> <pre>

He had come out from the shadow of his mentor, Mitterand. ... still a shadow of the player he was a couple of years ago. </pre> <p>This study suggests that the semantic relations from the source domain of light and dark are partially mapped onto the target domain of happiness and sadness but that there are a number of inconsistencies and gaps in the linguistic correspondences; centrally, metaphorical light is not used antonymously with metaphorical dark, although there is a limited correspondence where verbal light up is used antonymously to darken. Overall, the impression is one of a bundle of related mappings, which can all be interpreted within a general statement that positive emotions are associated with lightness and negative ones with darkness but which lack the overall structural coherence of ARGUMENT IS WAR.

5.4. EMOTIONS ARE TEMPERATURES

I now discuss the domain of temperature, widely used as a source of emotion metaphors (Lakoff and Kovecses; Gibbs, The Poetics of Mind). Some of the expressions generated by this mapping are instances of metaphor from metonymy, though the central terms, which are discussed here, are instances of metony-mybased metaphor.

Because the four central temperature terms, hot, cold, warm, and cool, are highly polysemous, I analyzed 1,000 citations of each. I examined concordances of their comparative and superlative forms, and of nominal and verbal inflections where applicable, but these did not yield any significant additional patterns, so they are not reported here. I also examined the concordances of a number of less frequent terms from the source domain of temperature, such as tepid and chilly. A full account, showing all the senses found in the data, is in an earlier paper (Deignan, "Metaphorical Polysemy and Paradigmatic Relations"); here the main findings are summarized.

A completely systematic mapping of the domain of temperature onto the domain of emotion would imply that metaphorical hot should describe feelings similar to but more intense than those described by metaphorical warm. The citations show that this is not the case. Although both lexemes have metaphorical senses that are used to talk about feelings, they do not seem to occupy adjacent positions on a single scale. Hot is used to talk about anger and, much less frequently, about sexual desire. Warm, on the other hand, is used to talk about friendliness. In their evaluative orientation, the main metaphorical senses are opposed. Hot when used to describe anger is often negative, while the "friendly" sense of warm evaluates positively, as the following citations show:</p>

<pre> He has little patience with the child and, while he's never been physically abusive, Joanne worries that his hot temper will lead to violence. Where her own parents were warm and friendly and did their utmost to make strangers feel welcome, his were haughty and distant. </pre> <p>Cool and cold have a relationship of loose synonymy when used with their literal senses. Each has a number of metaphorical senses, several of which appear to be related in the same way as the literal senses. Cool is used to talk about a lack of friendly feelings in citations such as the following:

The audience is likely to give the president a cool reception.

Cold has a sense that seems to denote similar but more intense feelings or behavior, maintaining the relationship between the literal meanings of the lexemes, as the following citation shows:

The next time you call him, you'll get a cold reception. You'll lose him as a friend.

Both cold and cool also have a conventional metaphorical sense denoting control over emotions, and the relationship between these senses seems to be loosely synonymous. As in their literal senses, these senses evaluate differently, cool expressing approval and cold being negative:</p>

<pre> The timing of his tackles and his cool assurance were first-class.

He is cold and calculating. </pre> <p>Other metaphorical senses of cold and cool seem to be unrelated. Cold is used to talk about fear in citations such as the following:

I shivered as cold chills of fear crept over me.

There is no corpus evidence that cool is used with a similar meaning. It is used with an unrelated meaning, to signify approval in citations such as the following:</p> <pre> There will be days when things aren't cool, but if you look back over your life you'll realise that problems do eventually get worked out. </pre> <p>There is no corresponding meaning of cold.

The literal senses of both hot and cold are clearly antonymous in the corpus, and occasionally, there also seems to be antonymy between metaphorical hot and cold. For instance, just as hot is used to talk about sexual desire, a conventionally metaphorical sense of cold is used to talk about lack of sexual feelings. The following citations show this apparent opposition:</p> <pre> The couple are tipped to steam up the screen with red-hot love scenes. What I adore about her is the fact that sexually she's rather cold and remote. </pre> <p>However, the collocations of the two senses are different: hot used in a sexual sense collocates with words that refer to texts or pictures, such as film and photo, while cold collocates with words that refer to people. There is also some antonymy between metaphorical warm and cool, but as is the case for hot and cold, there are many nonliteral senses where this relationship does not hold.

Overall, the data suggest that a few of the metaphorical senses of the four central temperature adjectives are related to each other by loose synonymy or by antonymy, echoing the relationship which exists in the source domain, but that the majority of their nonliteral senses exist independently.

Besides these central four terms, there are several other lexical items which are used to talk about literal temperature, each of which has a relationship of loose synonymy with one of the four central terms. For example, in the source domain, both lukewarm and tepid describe a temperature slightly below that described by warm, and their metaphorical senses parallel this source domain relationship. The following citations illustrate their metaphorical use:</p> <pre> They showed at best a lukewarm attitude and at worst a positive hostility. His nomination has received tepid support in the Senate. </pre> <p>Similarly, icy and chilly have metaphorical senses denoting unfriendliness that are loosely synonymous with a metaphorical sense of cold, echoing the relationship between the literal senses of the words:</p> <pre> ... enduring icy silences. ... his chilly relationship with Stephens. </pre> <p>It seems likely that the very patchy semantic patterns found in the target domain of EMOTIONS are the result of the mappings being primarily metonymic rather than metaphorical. As Lakoff and Kovecses have noted ("The Cognitive Model"), there is a general mapping of emotions onto temperatures, which can be attributed to a cause-effect relationship; we perceive changes in our body temperature as a result of experiencing certain emotions. However, this is not a structural mapping of domain onto domain such as ARGUMENT IS WAR. Work by Kovecses (Metaphor and Emotion) and others suggests that each emotion is individually mapped onto a part of the temperature scale, according to its physiological effects. The effects of emotions on our bodies are not organized following the semantic relations within the domains of temperature: we may feel cold when afraid, but we do not usually notice sensations of heat when not afraid. This is supported by the corpus data examined here; it explains why metaphorical senses of cold and hot are not found on a cline which has lack of fear at one extreme and intense fear at the other.

At the level of detail, some literal near-synonyms have corresponding meanings in the target domain, such as tepid/warm/lukewarm. These could be the result of a structural mapping; that is, we use lukewarm as a mild antonym of warm when describing friendships or welcomes because there is a mapping of this subdomain of emotion onto the subdomain of temperature, which has created correspondences. Alternately it is possible that each separate emotion has been mapped. This would imply that the consistency between source and target domain meanings of warm and lukewarm is the result rather than the cause of the mappings. The first possibility would be closer to a metaphorical process, while the second would be metonymical.

6. Summary and Conclusion

In the first half of this article, I put the case for metaphor and metonymy to be seen as ends of a cline. This cline has certain points that have been identified by researchers such as Goossens and that were identified here through corpus analysis. In the second half of the paper, I examined some target domain relationships between expressions motivated by metaphor and those motivated by an interaction between metaphor and metonymy. The corpus studies were not exhaustive, in that they did not analyze in detail every mapping and did not include quantitative comparisons. Given these limitations, they tended to suggest that mappings that are, from a theoretical perspective, primarily metaphorical rather than metonymic lead to coherent semantic networks in the target domain. Such mappings show semantic relationships that, in the main, echo those found in the source domain. By contrast, expressions that, according to the literature, have a strong element of metonymy, such as TEMPERATURE metaphors, seemed to show small-scale areas of coherence at the most detailed level, but at the larger level only very limited evidence of coherent relationships of antonymy and synonymy. From this data, it seems that the linguistic realizations of metonymy-based mappings form loosely related patches of small scale mappings, within a much more general framework, rather than a consistent one-to-one mapping of domain onto domain at all levels of specificity. Given that metonymy is now believed to underlie much, possibly most, nonliteral language, it is possible that systematic, coherent mapping is relatively rare.

These three corpus studies suggest that there is some correlation between target domain semantic patterns and the type of motivation underlying the mapping. If this finding could be confirmed over a wider range of target domains, combined with more detailed analyses, researchers would have another tool for the analysis of different types of nonliteral language: lexical patterns in the target domain could be examined for insights into the types of mapping involved. More generally, this paper has attempted to demonstrate how the rigorous examination of naturally occurring language can give new insights into metaphorical and metonymic mapping, insights that are potentially more detailed and far-reaching than those obtainable through elicited language data.

Works Cited

Barcelona, Antonio. "On the Plausibility of Claiming a Metonymic Motivation for Conceptual Metaphor." Metaphor and Metonymy at the Crossroads: A Cognitive Perspective. Ed. Barcelona. Berlin: Mouton de Gruyter, 2000. 31-58.

Cameron, Lynne. "Working with the Complexity of Language in Use: The Case of Metaphor." Annual Meeting of the British Association for Applied Linguistics, University of Brimingham, 1997.

Charteris-Black, Jonathan. "Speaking with Forked Tongue: A Comparative Study of Metaphor and Metonymy in English and Malay Phraseology." Metaphor and Symbol 18.4 (2003): 289-310.

Deignan, Alice. "Corpus-Based Research into Metaphor" Researching and Applying Metaphor. Ed. Lynne Cameron and Graham Low. Cambridge: Cambridge UP, 1999. 177-99.

--. Metaphor. Collins Cobuild Guides to English. London: HarperCollins, 1995.

--. Metaphor and Corpus Linguistics. Amsterdam: Benjamins, 2005.

--. "Metaphorical Polysemy and Paradigmatic Relations: A Corpus Study." Word 50 (1999): 319-38.

Deignan, Alice, and Liz Potter. "A Corpus Study of Metaphors and Metonyms in English and Italian." Journal of Pragmatics 36.7 (2004). 1231-52.

Feyaerts, Kurt. "Refining the Inheritance Hypothesis: Interaction between Metaphorical and Metonymic Hierarchies." Metaphor and Metonymy at the Crossroads: A Cognitive Perspective. Ed. Antonio Barcelona. Berlin: Mouton de Gruyter, 2000. 59-78.

Gibbs, Raymond W. The Poetics of Mind: Figurative Thought, Language and Understanding. Cambridge: Cambridge UP, 1994.

--. "Speaking and Thinking with Metonymy." Metonymy in Language and Thought. Ed. Klaus-Uwe Panther and Gunter Radden. Amsterdam: Benjamins, 1999.61-76.

Gibbs, Raymond W, and Nicole L Wilson. "Bodily Action and Metaphorical Meaning." Style 36.3 (2002): 524-41.

Goossens, Louis. "Metaphtonymy: The Interaction of Metaphor and Metonymy in Expressions for Linguistic Action." Cognitive Linguistics 1.3 (1990): 323-40.

Kittay, Eva Feder. Metaphor: Its Cognitive Force and Linguistic Structure. Oxford: Oxford UP, 1987.

Kovecses, Zoltan. "Happiness: A Definitional Effort." Metaphor and Symbolic Activity 6.1 (1991): 29-46.

--. Metaphor: A Practical Introduction. Oxford: Oxford UP, 2002.

--. Metaphor and Emotion: Language, Culture and Body in Human Emotion. Paris: Cambridge UP, 2000.

Lakoff, George. "The Invariance Hypothesis: Is Abstract Reasoning Based on Image-Schemas?" Cognitive Linguistics 1.1 (1990): 39-74.

--. Women, Fire, and Dangerous Things: What Categories Reveal about the Mind. Chicago: Chicago UP, 1987.

Lakoff, George, and Mark Johnson. Metaphors We Live By. Chicago: Chicago UP, 1980.

Moon, Rosamund E. Fixed Expressions and Idioms in English: A Corpus-Based Approach. Oxford: Clarendon, 1998.

Nunberg, Geoffrey. "The Non-Uniqueness of Semantic Solutions: Polysemy." Linguistics and Philosophy 3 (1979): 143-84.

Radden, Gunter. "How Metonymic Are Metaphors?" Metaphor and Metonymy at the Crossroads: A Cognitive Perspective. Ed. Antonio Barcelona. Berlin: Mouton de Gruyter, 2000. 93-108.

Sinclair, John. Collins Cobuild English Dictionary. London: HarperCollins, 1995.

--. Corpus, Concordance, Collocation. Oxford: Oxford UP, 1991. Sunday Mirror 6 May 2005.

Sweetser, Eve. From Etymology to Pragmatics: The Mind-As-Body Metaphor in Semantic Structure and Semantic Change. Cambridge: Cambridge UP, 1990.

Alice Deignan

University of Leeds 

Table 1. The Cline from Metonymy to Metaphor

Label              Characteristics

Metonym            Nonconventional
                   Intra-domain mapping
                   Interpretation dependent on
                   context.

                   Conventional
                   Intra-domain mapping.
                   Interpretation not dependent on
                   context.
                   Metonymic use is an established
                   sense of lexeme.

Metaphor           Metonymy clearly discernible.
from metonymy      Mapping both intra-domain and
(Goossens)         cross-domain.
                   Interpretation often ambiguous
                   between literal and figurative
                   use: continua of meaning
                   between the two.

Metonymy           One component of a meta-
within metaphors   phorical multiword phrase
(Goossens)         appears to be metonymical.
                   Rare in the corpus data searched.

Metonymy-          Mapping both intra-domain and
based              cross-domain.
metaphor           Interpretation only very rarely
                   ambiguous between literal and
                   figurative use.

Metaphor           Cross-domain mapping
                   Normally no ambiguity between
                   literal and figurative use.

Label              Example

Metonym            Ham sandwich
                   The ham sandwich is getting
                   impatient for his check.
                   (Nunberg)

                   Shelve
                   A new school exam for 16 year
                   olds has been shelved by the
                   government

Metaphor           Turn one's nose up
from metonymy      ... you should never turn your
(Goossens)         nose up at inexpensive plants ...

Metonymy           bite one's tongue off'
within metaphors   (Goossens 333)
(Goossens)

Metonymy-          Heated
based              I found myself in heated
metaphor           argument the other day

Metaphor           shoot down
                   He shot down all of my
                   arguments (Lakoff and Johnson)