Referential and propositional metonymy (Warren, Reference Warren, Panther and Radden1999, Reference Warren, Dirven and Porings2003, Reference Warren2006)

One of the most influential models of metonymy is Warren’s (Reference Warren, Panther and Radden1999, Reference Warren, Dirven and Porings2003, Reference Warren2006) model, which distinguishes between ‘referential’ metonymy and ‘propositional’ metonymy. Of the two types, ‘referential’ metonymy is the most typical form of metonymy. An example of a referential metonymy is as follows:

In this example, the noun ‘Shakespeare’ refers metonymically to the noun phrase ‘plays written by Shakespeare’. Warren postulates a distinction between ‘referential metonymies’ such as this (where one entity is related to another) and ‘propositional metonymies’ (where one proposition is related to another). An example of a propositional metonymy is:

In this example, the fact that Rosalind ‘raised her eyebrows’ triggers the proposition that she was surprised. Superficially, referential metonymies tend to violate truth conditions (it is not possible to ‘eat’ Shakespeare), whereas propositional metonymies relate one proposition to another via an ‘if–then’ relationship (if she raised her eyebrows then she must be surprised). Furthermore, referential metonymies are nouns (with a few rare exceptions) whereas propositional metonymies can suggest notions represented by adjectives, such as surprised in the example here, but may also involve other parts of speech.

The focus of Warren’s work is on referential metonymy. She provides a detailed account of the phenomenon, based on the analysis of a large number of examples taken from the literature on metonymy. In this analysis, she discusses the rationale for the use of referential metonymy, along with an account of its semantics and its syntactic behaviour. She contends that the interpretation of referential metonymies reveals a modifier-head construction, in which the modifier is the explicit noun and the head the inferred element. For instance, the potatoes are boiling is understood to mean ‘the water which contains the potatoes is boiling’. In other words, the interpreter amends the expression so that it harmonises with the predicate and the non-literacy is dissolved. She also contends that, like the other modifier-head combinations, referential metonymies express a limited number of ‘semantic patterns’. These patterns correspond, in nature, to the ‘metonymy types’ proposed by Radden and Kövecses, but are described at a much more general level. They comprise: causation, location in space and time, possession, composition, and representation.

One observation she makes is that while in metaphor there is thought to be more than one possible mapping between the vehicle and its referent, in referential metonymy there is usually only one, although the reader or listener can use their world knowledge to access more complex meanings. So, for example, in the metaphor Love is a Journey there are many ways in which a love affair might be like a journey (i.e. both have a beginning and end; they are both often thought to be purposeful, or ‘going somewhere’ in both it is possible to take a wrong turning, and so on), whereas in metonymy there is usually only one mapping. We can see how this contrasts with metonymy by looking at the following sentence:

Here, Warren would argue that linguistic knowledge gives us ‘people with anoraks’, exemplifying a part-whole metonymic link, and the fact that they are ‘geeky’ depends on world knowledge. This account differs slightly from the account that a cognitive linguist would give, as in cognitive linguistics no distinction is drawn between denotative and connotative knowledge; both types of knowledge are thought to be subsumed within the broader concept of ‘encyclopaedic knowledge’, which involves the activation of relevant ICMs (see Chapter 1). This example illustrates the ways in which metonymic mappings can be complexand highly context dependent. This will become even more apparent in subsequent chapters, which explore metonymy in different types of data.

A further difference between metaphor and referential metonymy, according to Warren, is that metaphor can be involved in a form of language play called ‘zeugma’, where two contrasting readings are activated at once, as in, for example, ‘the colonel took his hat and his leave’. Here there is incongruity in the fact that a literal and metaphorical reading of the same source item appear in the same phrase, and this incongruity leads to a humorous interpretation. In contrast, if literal and metonymic readings of the same source term appear in the same phrase, this does not lead to zeugma. For example, it is not incongruous, or even particularly marked, to say something like ‘I found the Metonymy book quite heavy-going so I put it down for a bit and watched TV’.

In both metaphor and metonymy, these ‘double framings’ can be exploited for rhetorical effect. In metonymy however, because of the reduced role of incongruity (i.e. zeugma), the effect tends to be more subtle than in metaphor. Here is one of Warren’s (Reference Warren, Dirven and Porings2003) examples illustrating how double-framing in metonymy is involved in language play:

Here, Warren argues that the image of ‘the gentle hand of a loving mother’ is contrasted with ‘the firm grip of a strong-willed, ambitious person’ (ibid.; 31). She goes on to point out that the mother of a ruler will, as a result of the care that she has given him (or her) as a baby, exert influence over the way in which he (or she) rules the country.

Warren argues that metaphor can operate above the phrasal level while referential metonymy does not. Metaphorical expressions from the same source domain can work together to form ‘themes’ which extend across large sections of text, whereas this does not generally happen with metonymy (Warren, Reference Warren, Dirven and Porings2003: 117). However, as we will see in subsequent chapters, when we take a broader definition of metonymy we see that it can operate at the level of the text where it can form themes that contribute to the overall coherence of the text.

According to Warren, the vast majority of referential metonymies are noun phrases, although there are some examples of adjectives functioning as referential metonymies. This occurs when a metonymy is a modifier. We can see this in the following example:

Here, the term ‘red pen’ refers metonymically to a pen that has red ink in it. The pen itself may be any colour.

It is possible for a single phrase to be both referential and propositional, as we can see in the following excerpt, from Act 1 of Arthur Miller’s 1955 play A View from the Bridge. In this play, Eddie, Catherine’s uncle, is deeply suspicious of Catherine’s boyfriend, who he thinks is simply using her to gain American citizenship:

Here, Eddie refers to Catherine as her boyfriend’s ‘passport’, and goes on to explain that the boyfriend is only interested in marrying her so that he can remain in America. At first sight, this use of the word ‘passport’ appears to constitute an example of Warren’s ‘referential’ metonymy (one would not normally bow to a passport). However, an if–then relationship is also implied (if he marries her then he will obtain an American passport), which suggests that it also has elements of propositional metonymy.

Warren’s position represents the traditional view of metonymy, and before the 1990s, this restrictive view of what constitutes metonymy was widely shared amongst linguists. However, since the 1990s, with the increasing attention that cognitive linguists have paid to metonymy, broader, more encompassing definitions have been proposed. It is in this broader, more encompassing framework that this book is situated, as such a view is helpful in explaining the complex, and sometimes contradictory characteristics that metonymy can exhibit when it is found ‘in the wild’. All of the subsequent models discussed in this chapter adopt this broader view of metonymy.

Referential, predicational and illocutionary metonymy (Panther and Thornburg, Reference Panther and Thornburg1998)

Panther and Thornburg (Reference Panther and Thornburg1998) offer a slightly more fine-grained distinction than Warren, and identify two broad types of metonymy: ‘propositional’ and ‘illocutionary’. ‘Propositional’ metonymies are further broken down into the two sub-types: ‘referential’ and ‘predicational’.

To a large extent, referential metonymies are identical to those identified by Warren in that they involve relationships between entities, as in the following example, where ‘the buses’ refer to the drivers of those buses:

Predicational metonymies mostly involve relationships between events. An example of a predicational metonymy is:

Here the term ‘able to tell me’ metonymically stands for the fact that ‘he did tell me’. In these sorts of metonymies, a potential event expressed through a form of modality (e.g. the ability, possibility, permission or obligation to undertake an action) is metonymically linked to its occurrence in reality. Metonymies such as these would be classed as potential for actualin Radden and Kövecses’ framework.

Illocutionary metonymy involves pragmatic inferencing. For example, in the following example, the question ‘Have you got a fiver?’ is linked through an illocutionary metonymy to the question ‘Please can you lend or give me five pounds?’

Rather than relying on frame or ICM-based relationships, illocutionary metonymies rely on scenario-based relationships. That is to say, they rely on the speaker and hearer’s knowledge of a ‘typical scenario’ (in the above example the ‘typical scenario’ would involve lending a friend some money in order to help him or her pay for something). One part of the scenario is used to refer to other parts of it. These might include the necessary preconditions for an event, the event itself and the consequences of the event. Thus, in the above example, a precondition (i.e. that the hearer is in possession of a five-pound note) is used metonymically by the speaker to ask if he or she can borrow it. According to Panther and Thornburg, the ‘literal’ meaning of a metonymy is always relevant to the interpretation of a metonymic expression; it is not the case that metonymy simply involves the substitution of one word for another.

Panther and Thornburg’s model makes two key contributions to the field of metonymy studies. These are its focus on illocutionary metonymy and the attention that it pays to the role played by metonymy in grammar. These mark metonymy out as something far more encompassing than a lexical phenomenon. Their model shows how typical metonymic relations (such as cause for effect, result for action, producer for product and so on) operate as natural inference schemaswithin our long-term memory, and that they can be rapidly retrieved to help us access the intended meaning of the utterances that we hear. This is an important contribution as it allows for a more systematic study of the role of metonymy in pragmatics and grammar, allowing analysts to investigate different metonymic relationships that are involved in different kinds of pragmatic inferencing. Their model will be referred to frequently in subsequent chapters.

‘Source in target’ metonymy and ‘target in source’ metonymy (Ruiz de Mendoza Ibáñez and Diez Velasco, Reference Ruiz de Mendoza Ibáñez, Diez Velasco, Dirven and Porings2002)

Another model of metonymy which, like those discussed above, has made a key contribution to the metonymy literature, but which may need to be developed to account for some types of real-world data, is that proposed by Ruiz de Mendoza Ibáñez and Diez Velasco (Reference Ruiz de Mendoza Ibáñez, Diez Velasco, Dirven and Porings2002). They believe the distinction between referential and propositional metonymy to be irrelevant, and focus instead on the relationship between the metonymic expression and its referent. They argue that all instances of metonymy can be described as one of two types: either as a ‘target in source’ metonymy, where the metonymic term is part of its referent, or as a ‘source in target’ metonymy, where the referent is part of the metonymic term.

The example that they give of a ‘target in source’ metonymy is the Pill to refer specifically to the contraceptive pill, as in:

Here, the word Pill is used to refer to a specific type of pill, which means that the target of the metonymy (the contraceptive pill) is a subset of the domain covered by the general word ‘pill’.

The example that they give of a ‘source in target’ metonymy is the use of ‘hand’ in expressions such as:

In this example, the ‘hands’ refer to the sailors who are doing hard physical work so the hands are simply part of the domain.

Another way of describing their approach is to say that in ‘target in source’ metonymies the referent is a sub-domain of the metonymic vehicle, whereas in ‘source in target’ metonymies the metonymic vehicle is a sub-domain of the referent. Ruiz de Mendoza Ibáñez and Diez Velasco argue that ‘source-in-target’ metonymies involve ‘domain expansion’, while ‘target-in-source’ metonymies involve ‘domain reduction’, and that these are two fundamental processes. They adopt a very broad definition of what is meant by the ‘source’ and ‘target’, which allows them to classify a wide range of metonymies as ‘target-in-source’ metonymies. For example, in their model, the following would be classified as a ‘target-in-source’ metonymy:

This is because the source domain (IBM) includes everything that a speaker might know about this domain, including the fact that it has managers and that those managers have the capacity to hire and fire. The fact that IBM as a whole is used to refer to this particular aspect of the company means that the metonymy involves ‘domain reduction’.

If we think back to the taxonomy of metonymy types proposed by Radden and Kövecses (Reference Radden, Kövecses, Panther and Radden1999) which we saw in Chapter 2, we will notice a marked difference between their approach and the approach employed by Ruiz de Mendoza Ibáñez and Diez Velasco. Whereas Radden and Kövecses distinguish between part/whole and part/part metonymies, Ruiz de Mendoza Ibáñez and Diez Velasco groupall metonymies under the first of these headings. They do this by extending the notion of the ICM. So for example, the sentence ‘she married money’ would be a trait for personmetonymy under Radden and Kövecses’ approach, whereas it would simply be a part for whole metonymy under Ruiz de Mendoza Ibáñez and Diez Velasco’s approach, as wealth would be seen as a possible part of the ICM for an ‘ideal spouse’.

Ruiz de Mendoza Ibáñez and Diez Velasco’s approach constitutes a useful tool for metonymy analysis, especially when we look at the role of metonymy in genres and registers that are specific to particular discourse communities. The process of ‘domain reduction’ appears to be particularly productive in the meaning-making processes in which individual discourse communities engage. As we will see in Chapter 4, discourse communities often take fairly commonly used words and give them their own particular, much more specific meanings, thus creating a feeling of in-group membership. For example, Deignan et al. (Reference Deignan, Littlemore and Semino2013) show how in the language used by a group of supporters at a children’s football club, the instruction to ‘kick it’ has a narrowed-down, metonymic meaning of ‘kick it really hard, with conviction’. It does not simply mean ‘kick it’. The concept of domain reduction is also useful when speakers wish to be deliberately vague, for example when they want to avoid singling out a particular individual within an organisation who may have been responsible for a particular action. The very broad definition of domains that they adopt is also useful in explaining anaphoric reference and textual cohesion involving metonymy. For these reasons, I will make regular references to their model when discussing the functions of metonymy in real-world data.

Some metonymic expressions can involve both a source in target mapping and a target in source mapping. Ruiz de Mendoza Ibáñez and Diez Velasco illustrate this with the sentence:

They point out that this sentence involves both a target in source relationship (where ‘Shakespeare’ is used to refer to ‘books written by Shakespeare’) and a source in target relationship (where Shakespeare’s work is used to refer to the actual books in which this work appears).

When we look at examples of metonymy in real-world data and take contextual and pragmatic meanings into account, we see that this combination of target in source and source in target relationship within a single metonymy is arguably more widespread than Ruiz de Mendoza Ibáñez and Diez Velasco suggest. Let us begin by exploring their two earlier examples that were used to illustrate the different relationships. As we saw above, they point out that the use of the term ‘the Pill’ to refer only to the contraceptive pill is clearly a target in source metonymy as it refers only to a specific pill. However, on a more general level, it could also be said to be a source in target metonymy as it provides metonymic shorthand to a much wider set of events, including the taking of the Pill, and the set of societal changes that occurred because of the Pill, as we can see in these examples:

The example ‘all hands on deck’, which they provide as an example of a source in target metonymy, can also be said to involve both types of relationship. In their model, the hands stand for the bodies of the sailors but it could be re-analysed as a target in source metonymy, as the hands stand for one just particular thing that we do with our hands (i.e. work). Indeed, in the Bank of English, the expression ‘all hands on deck’ occurs forty-three times, and not once is it used to refer to sailors. It refers in nearly all cases to the idea of working together and ‘mucking in’, and the hands do not stand for sailors, but what people do with hands (i.e. work). We do not know whether the expression ever did refer to sailors or whether it was ever used by sailors. The fact that we do all sorts of things with our hands besides working suggests that the referent is narrower than the metonymic vehicle, ‘hands’.

These examples show that it can be very difficult to decide whether a particular metonymic expression is a ‘target in source’ metonymy or a ‘source in target’ metonymy, and that when we look at the contextual meanings of metonymic expressions, we find evidence of both relationships occurring at once. This is probably due to the broad definition of ICMs that Ruiz de Mendoza Ibáñez and Diez Velasco adopt. The more broadly we define an ICM, the more difficult it becomes to say whether one is part of another or vice versa. This, however, does not detract from the usefulness of the model when accounting for the functions played by metonymy in real-world data. As mentioned above, we will see in Chapters 4 and 5 how the notion of domain reduction helps to explain the ways in which particular discourse communities metonymically narrow down the meaning of word until it means something very specific to that community. It is also useful in accounting for metonymy-based textual cohesion. We will also see how the notion of domain expansion is useful for showing how metonymy can be used to discriminate against large groups of people who happen to share a single characteristic.

Peirsman and Geeraerts’ (Reference Peirsman and Geeraerts2006a) varying strengths of contiguity

Peirsman and Geeraerts’ model of metonymy constitutes a radical departure from the preceding models as it treats metonymy as a radial category. The basic idea of a ‘radial category’ is that whenever we put things into categories, we are able to identify those which are core ‘prototypical’ members of the category, and those which are less central. So if we were to set up the category of ‘pets’, a cat or a dog might be seen as ‘prototypical’ (at least in the UK). Other animals, such as snakes and rats would be seen as less prototypical, and lions and tigers even less so. The latter would probably shade into the category of ‘zoo animals’ though they might also be seen as pets at a stretch. Thus at this point the category of ‘pets’ shades into the category of ‘zoo animals’ and the boundary between them is fuzzy. So radial categories can be said to have prototypical and less prototypical members and fuzzy boundaries between them. There are very few things in life that exist in discrete categories with clear boundaries, and it would be wrong to try and categorise language in this way too. The idea that language exists with radial categories is a fundamental tenet of cognitive linguistics, and radial category approaches have been applied to word senses, phonological features, intonation patterns, linguistic constructions and grammar rules (Taylor, Reference Taylor2003). It thus makes sense to see metonymy as a phenomenon which exists within a radial category.

In their radial category model of metonymy, Peirsman and Geeraerts emphasise the importance of contiguity between the vehicle and its referent; in other words, they are interested in how close the term used is to its intended meaning. For them, prototypical metonymies involve cases where the term is part of its metonymic meaning, or vice versa. An example of such a metonymy would be:

Here, a roofover my head is part of its intended meaning, a house, so the example would be classed as prototypical. An example of a metonymy that lies between the core and periphery of the category would be:

Here the theatre refers to the people who were sitting in the theatre. There is a contiguous relationship between them (as the people were actually in the theatre at some point) so the metonymy would be classed as being midway between the core and the periphery of the category. An example of a highly peripheral metonymy would be as follows:

Here there may not be an actual ‘round table’ in the room. The metonymy is based on the idea that people ‘might’ be sitting round a round table and thus all participating equally in the discussion, but they might not. Because there is not necessarily a round table in the room, the term and its referent cannot be said to be contiguous in any way, so it is a peripheral example of metonymy.

Peirsman and Geeraerts also argue that metonymies lying towards the edge of the category are less ‘bounded’ than those at the centre. Bounded entities have a clear cut-off point, whereas unbounded entities do not. A metonymy involving two bounded entities, which would thus be seen as prototypical, is:

The metonymy pair of legs would be described by Peirsman and Geeraerts as prototypical as both the term (a pair of legs) and its referent (a woman) are bounded entities. Metonymies where either the term or its referent (or both) are unbounded, would be seen as lying towards the periphery of the category. An example of such a metonymy would be:

Here, although Hollywood is bounded (there is probably a city line somewhere in the suburbs) its intended referent is not. It is difficult to say exactly what it means and the concept of a Hollywood film is probably itself a radial category that has fuzzy boundaries with other genres.

Finally, Peirsman and Geeraerts see concrete metonymies as lying at the centre of the category and more abstract metonymies lying towards the periphery, so for example the following would be seen as prototypical:

Here both a set of wheels and its referent, a car, are concrete entities. Lying towards the periphery are metonymies involving more temporal or abstract entities, such as:

The referent of this metonymy, ‘getting married’ is less concrete than ‘a car’ because you cannot see it, get in it or be run over by it.

To sum up, according to Peirsman and Geeraerts, metonymy types radiate out from prototypical instantiations in three different ways: they gradually become less contiguous, less bounded and less concrete. This model is useful as it shows that metonymies can differ from one another in more than one way. However, one problem with the model is that, when faced with metonymies from real-world data, it is sometimes difficult to see which category they fit into. We can see this by looking more closely at some of the examples discussed above. The expression ‘have a roof over my head’ refers, in very basic terms, to the idea of actually having a house, so at first sight there appears to be a high degree of contiguity between the vehicle term and its referent. However, the expression also carries an extra meaning of having enough money to provide for one’s basic needs, and here we start to see less continguity between the vehicle and its referent. The exact degree of contiguity between a vehicle and its referent can therefore be very difficult to identify when metonymies are used in context, as there can be more than one referent. Moreoever, it is not always clear which referent(s) is/are intended as they are often vague and underspecified. The same can be said for boundedness. We have already seen, in Chapter 2, that ‘a pair of legs’ does much more than simply refer to a woman, and that it carries all sorts of connotations about a woman’s sexuality and availability. These connotations are not easily covered by the notion of a ‘bounded entity’. Again, because the metonymy does not have a clear, single referent, it is difficult to measure the extent to which this referent is ‘bounded’. Finally, it can be difficult to determine the level of concreteness of some metonymic expressions. We saw this in Chapter 1 with the ‘brothers needed some muscle’ example. Here, we could say that ‘some muscle’ simply refers to a strong man, in which case the referent is highly concrete, or it could refer to a build-up of forces prior to a fight, which is a slightly less concrete phenomenon. The main problem with this model is that it assumes a one-to-one correspondence between the vehicle term and its referent, and in reality this is not always the case. Having said that, it is useful to see metonymy as a radial category, and Peirsman and Geeraerts are right to say that different types of metonymy can radiate out from the centre along different axes. For this reason, the model is useful when analysing the different senses of ambiguous metonymies in context.

Langacker’s (Reference Langacker1993) focus on active zones and what this means for models of metonymy

One of the most influential cognitive linguistic theories of metonymy is that proposed by Langacker (Reference Langacker1993). He argues that metonymy is involved virtually every time a different aspect of word’s meaning (or the domain that it represents) is highlighted by its use in a different context. This, he argues, is because a shift in profile is generally involved. For example, there is a metonymic relationship between the use of the word ‘university’ to refer to the actual buildings and the use of the word to refer to the institution or a university sports team, as shown in the following three examples:

Each of these senses ‘profiles’ a different aspect of the university and therefore involves metonymy. His argument is based on the premise that words alone are incapable of fully specifying the meaning intended by the speaker and that there is always a degree of indeterminacy. The information that is provided in the actual words that we read or hear does not itself establish the precise connections that are understood by the speaker and hearer. In Langacker’s words:

These ‘other means’ frequently involve metonymy. In each case, a different part (or ‘facet’) of our knowledge of the referent is brought to the fore. In other words, it becomes the focus of our attention. This facet is then ‘profiled’ and becomes an ‘active zone’. Individual words can only ever point to a broad area of meaning, and different facets of that meaning become active in different contexts. Metonymy is therefore a reflection of our fundamental ‘reference point ability’, which refers to our capacity to decide which facet is being profiled in any particular instance of language use. To illustrate further, each of the following sentences profiles a different understanding of the word ‘town’:

These examples refer respectively to the town centre, the people who live in the town, the physical location of the town and the town football club. The metonymic ‘work’ that needs to be done to interpret these examples involves drawing on the various ICMs that are triggered by the word ‘town’. Equally, different meanings of the word ‘film’ are profiled in each of these examples:

The meaning of the word ‘film’ that is being profiled in each of these three examples gradually becomes more concrete, reflecting the context in which it occurs. Langacker (Reference Langacker, Panther, Thornburg and Barcelona2009: 54) suggests that another way to help us understand metonymy is to look at compound nouns, such as ‘butterfly net’ and ‘mosquito net’, which can be seen in the following examples:

In each case, a different facet of the word ‘net’ is profiled by the noun that it compounds with. The first is talking about a net that is used to catch butterflies, whereas the second is talking about a net that is used to protect the user from mosquitoes. The relationship between the two nouns is different in each compound even though the grammatical structure is the same. The reason for these different interpretations is that different ICMs are invoked in each case: in the case of the butterfly net, the idea is that we like to catch butterflies because they are pretty and in the case of mosquitoes, the idea is that we need to protect ourselves from them because they bite and carry diseases such as malaria. These examples are of course, highly conventional so we do not need to engage in active metonymic thinking in order to understand them. In other words, they are ‘conventional’ metonymies. When faced with novel metonymies, we need to engage in ‘metonymic thinking’ in order to understand them. Langacker (Reference Langacker, Panther, Thornburg and Barcelona2009: 54) illustrates this nicely with his example of a made-up compound: ‘elephant table’. This could be a table made out of an elephant’s foot, a table covered in elephant ornaments, a table shaped like an elephant, and so on. Because the metonymic connection is not fixed we are free to form our own connections and therefore understand the term in different ways. Metonymy can therefore be described as a thinking process that allows us to make sense of utterances by profiling a particular facet of a given phenomenon. The active zone is different in each of these cases. Empirical support for Langacker’s theory is provided by Onysko and Degani (Reference Onysko and Degani2012), who presented participants with a range of entirely novel English compounds (such as ‘voice canoe’, ‘bucket philosopher’ and ‘word truck’) and asked them to make sense of them. They found that the meanings produced always involved metonymic links between the constituents of the compounds, although they sometimes interacted with metaphor too.

Langacker’s notion of active zones is by no means limited to language. It is equally relevant to other forms of communication and is particularly common in the medium of cinema, which is inherently metonymic because of the way in which different camera angles can be used to change the angle, perspective and focus of the ‘shots’. This idea is discussed in depth in Chapter 5, which looks in detail at the different functions served by metonymy in multimodal forms of expression. The notion of active zones has also been used by researchers to show how metonymy is central to grammar and what might loosely be described as grammar ‘rules’ as well as to language change. A whole volume of papers has been dedicated to this idea (Panther et al., Reference Panther, Thornburg and Barcelona2009), and this work will be explored in Chapter 6.

Langacker’s view of metonymy presents a serious challenge to metonymy scholars as it suggests that, in some respects, just about all instances of meaning and communication are metonymic, as meaning comprehension nearly always involves highlighting some aspects of a phenomenon and putting others in the shade. Although this is interesting from a philosophical perspective, it presents a practical problem to those interested in defining and investigating metonymy, as it implies that there is nothing ‘special’ about examples such as those discussed in the Introduction, and that they are no different from any other form of language. However, there clearly is something interesting and different about metonymy that makes it worthy of investigation. The fact that a great deal of humour derives from the contrast between metonymic readings of words and their more basic senses means that humans do recognise it as a discrete phenomenon, albeit subconsciously. Metonymy scholars have therefore sought to reconcile the apparently contradictory views of ‘metonymy as a pervasive unremarkable phenomenon’ and ‘metonymy as something special’. Even Langacker agrees that there is something of a cline from prototypical ‘stands for’ metonymy and the more nebulous ‘active zone’ metonymy discussed in his work.

Barcelona’s (Reference Barcelona, Cuyckens, Berg, Dirven and Panther2003b) progressive membership constraint model and Handl’s (Reference Handl2012) radial category approach to metonymy

Barcelona (Reference Barcelona, Cuyckens, Berg, Dirven and Panther2003b, Reference Barcelona, Benczes, Barcelona and Ruiz de Mendoza Ibáñez2011) rises to the challenge presented by Langacker’s work by proposing a radial category approach to metonymy that accounts for both the prototypical examples of metonymy that are based on contiguity, and instances of metonymy that simply involve domain highlighting. Prototypical metonymies at the centre of the category are the sorts of referential metonymies discussed in Chapter 1 that could, in some cases, be described as reflecting ‘stands for’ relationships. Those at the periphery of the category involve domain highlighting, of the type discussed by Langacker. Metonymies in the ‘half way’ category are referred to as ‘typical’. The model is shown in Figure 3.1.

Figure 3.1 Barcelona’s progressive membership constraint model

Barcelona’s prototypical model of metonymy is a useful development in metonymy theory but it is not without its problems: it does not provide much detail about the exact nature of ‘typical’ metonymies that lie between prototypical and peripheral examples, and it is difficult to see why it gives such preferential treatment to referential metonymies.

Handl (Reference Handl2011) improves upon Barcelona’s model of metonymy by bringing in the notion of ‘underspecified meaning’. Her study of metonymy in the British National Corpus led her to conclude that a very large number of metonymies have a meaning that is in fact ‘underspecified’. In these metonymies, the basic sense of the vehicle term is retained and contributes to the contextual meaning of the expression, meaning that both the basic meaning and the contextual meaning are present. In order to illustrate this type of metonymy, she uses two examples:

1. (1) The White House isn’t saying anything.

2. (2) An earlier ferry had got me a front seat in one of the waiting buses.

She argues that the first of these examples constitutes a ‘typical’ case of metonymy in that the vehicle recedes fully into the background as soon as the target has been identified. However, the second example is more complex. In this example, both the buses and the drivers of those buses (who are doing the waiting) are part of the intended meaning; the buses do not simply provide ‘mental access’ to the drivers. Both the buses and their drivers are being referred to at the same time. The sentence therefore lacks the kind of directionality that is inherent in some of the earlier models of (prototypical) metonymy. The metonymic meaning comprises both the target and vehicle and it is ‘underspecified’ in that it is not entirely clear what is being referred to. It is not clear what direction the metonymic mapping should go in because the weight of its constituent parts is relatively balanced. The same can be said of the following example.

In this example, the BMW refers to both the car and the driver. It does not simply refer to one or the other, so its exact referent can never be specified.

Handl thus proposes the radial category approach to metonymy shown in Figure 3.2, which is a more refined version of Barcelona’s earlier model.

Figure 3.2 Handl’s (Reference Handl2012) radial category approach to metonymy

In this model, the relatively rare ‘typical metonymies’ involve completely distinct sub-domains. Washington is not actually the capital of America, it is the capital of the United States of America. ‘Underspecified metonymies’ involve both the vehicle and the target in their meaning and ‘domain highlighting’ metonymies have no distinct sub-domains whatsoever, just differences in construal or viewpoint that result from the juxtaposition of the different words.

Handl found that underspecified metonymies tend to involve some sort of containment relation. She argues that this is because people rarely need to differentiate between a container and its contents and that they tend to perceive them in a unified way because they operate as a ‘functional unit’ (Reference Handl2012: 4). In other words, the metonymic term does not simply ‘stand for’ the referent in a metonymy, but rather, it stands for both itself and the referent. Her model stresses the fact that linguistic metonymies are not simply another way of expressing an already extant idea, but that they represent an altered conceptualisation of the idea.

As well as corpus data, Handl (Reference Handl2011, Reference Handl2012) provides experimental evidence indicating that many metonymies are in fact often understood in an underspecified manner. She found that the informants in her study usually thought that the metonymy was referring to both the term and its referent. She assigned different weights to the two parts of the metonymy based on frequency of response and was thus able to provide a differentiated description of metonymies in terms of the constituents’ weights. In prototypical metonymies, the weight of the vehicle is relatively high. That is to say, the actual words of the metonymy play an important role in its meaning, whereas, when we move out towards the periphery of the radial category, the vehicle recedes into the background. In the more intermediate types of metonymy, both the vehicle and its referent play an equally important role. The fact that Handl’s model of metonymy is based on real-world data makes it particularly appropriate for the analysis of the communicative functions of metonymy that will be presented in the next two chapters.

Metonymy and blending theory

Blending theory is an approach to language and communication which holds that the construction of meaning involves the integration of more than one ‘mental space’, and that the resultant meaning is more than the sum of its parts (Fauconnier and Turner, Reference Fauconnier, Turner, Panther and Radden1999). A good account of the role of blending theory in metonymy comprehension is provided by Coulson and Oakley (Reference Coulson, Oakley, Panther and Thornburg2003), who illustrate their argument by exploring the expression:

They argue that this expression involves much more than a straightforward mapping from the source domain of ‘grave digging’ to the target domain of ‘getting oneself into difficulties’. In order to understand this expression, they argue that one needs to form a single conceptual blend in which the person who is getting themselves into difficulties is digging his or her own grave and then presumably climbing into it and dying. The expression cannot be understood by any ‘logical’ analysis; rather, its meaning is facilitated by the result for action metonymy of one’s grave standing for one’s own death. Once this metonymic relationship is activated the expression makes sense. The digging of one’s own grave and the act of dying are compressed into a single utterance. Metonymy is used to unpack meaning from the compressed elements in the blended space. Coulson and Oakley also argue that metonymy holds together the network of mental spaces that need to be activated in order for reasoning to be sustained over an extended length of time. They go on to show how metonymy comprehension involves two key concepts that are involved in the formation of a conceptual blend: ‘compression’ and ‘vital relations’. For example, the metonymy ‘a warm coat’ involves the compression of the cause of wearing a coat with the effect of being warm. The vital relation between these two is one of property. That is to say, the ‘warmth’ that the coat endows to the wearer is a property of coat within this blend.

We can see this principle at work in what at first sight appears to be a non-figurative extract from Arthur Miller’s play A View from the Bridge, which was also referred to above. In this play, an Italian immigrant couple (Beatrice and Eddie) who live with their niece (Catherine) in New York temporarily offer shelter to two cousins who are illegal immigrants and who are thus hiding from the authorities. Before the cousins arrive, Eddie utters the following to his wife and niece:

In the italicised segments, he means that they must act as if they do not have guests staying with them in order to avoid attracting undue attention from neighbours and potential spies. In order to understand what is meant here one has to create a blend between the actual non-noticing of the guests and the ability to behave as if one has not noticed them. In order to unpack the blend and to make sense of it, one needs to evoke the metonymic effect for causerelationship between actually seeing something and behaving as if one has seen it. One issue that has been raised with respect to blending theory is that it is extremely hard to prove or disprove (Gibbs, Reference Gibbs2000) as the theory itself is basically infalsifiable and it is difficult to test it out against alternative hypotheses. Having said this, it does present an intuitively appealing explanation for the more difficult-to-explain aspects of metonymy.

Metonymy and relevance theory

Relevance theory is an approach to meaning, proposed by Sperber and Wilson (Reference Sperber and Wilson1987, Reference Sperber, Wilson, Ward and Horn2004), which foregrounds the fact that in any communicative encounter the hearer or reader will assume that what is being said is relevant to him or her and will use this as a guiding principle in the interpretation process. Traditionally, relevance theorists have not dealt with metonymy, preferring to focus instead on metaphor and hyperbole. For them, metonymy simply constitutes a ‘loose’ way of speaking alongside metaphor and hyperbole, and the principle of relevance is sufficient for its interpretation (Carston, Reference Carston1997). More recently however, relevance theorists have begun to pay more attention to metonymy. Relevance theory approaches to metonymy tend to emphasise the role of world knowledge in metonymy comprehension and attempt to show how such world knowledge explains why some metonymic expressions are deemed much more ‘acceptable’ than others. For example, Nunberg (Reference Nunberg1995) takes the sentence: ‘I’m parked out back’ and argues that two conditions need to be satisfied before ‘I’ can be taken to mean ‘the car’. The first is that there needs to be a ‘salient correspondence’ between the properties of the person and the properties of the car. In this case, the ‘salient correspondence’ is that the person is the owner of the car and he or she has just parked the car ‘out back’. The second is that it has to be ‘either useful or interesting to know that these acquired or inherited properties apply to their carriers’ (Reference Nunberg1995: 192). In the case of the parking example, one would probably be interested to know that the person was ‘parked out back’ if one was going to be given a lift by their friend, or borrow their friend’s car, or if the two people were comparing notes on where they had parked. Papafragou (Reference Papafragou1996) provides further examples showing how relationships that are salient or relevant in contextplay a key role in determining whether and how a particular metonymic expression will be understood. She argues that there is no need to draw on lists of typical metonymy-producing relationships, as the relationships are generated in an ad hoc manner as the need arises. She describes this process as ‘interpretative’.

An advantage of relevance theory is that it emphasises the role of context in the comprehension process and this is useful when studying metonymy in real-world data. On the other hand, it appears to lean slightly too far in this direction, claiming that context is everything. When we analyse metonymy comprehension in more detail we can see that, as well as contextual cues, metonymy types such as those listed by Radden and Kövecses (Reference Radden, Kövecses, Panther and Radden1999) are also involved in the meaning-making process. For example the ‘I’m parked out back’ sentence referred to in the preceding paragraph clearly involves a possessor for possessedmetonymic relationship. Context plays an important role in activating and exploiting this relationship, but this does not mean that the metonymic relationship itself is absent. Indeed, recent work in cognitive linguistics has thus challenged relevance theory. For example, Ruiz de Mendoza Ibáñez and Pérez Hernández (Reference Ruiz de Mendoza Ibáñez, Pérez Hernández, Panther and Thornburg2003) show how metonymic mappings are repeatedly involved in the interpretation process. Hearers draw on their knowledge of typical metonymy types, alongside the principle of relevance, when interpreting metonymic uses of language. In order to illustrate their theory, they provide numerous examples of utterances whose illocutionary force has to be derived through pragmatic inferencing, and show how, in every case, the inferencing draws on a limited number of fundamental metonymic relationships.

Metonymy as an emergent phenomenon: the role of complex systems theory

Complex systems theory is an attempt to explain human behaviour by taking account of all the factors that are likely to have led to a particular event taking place (Larsen-Freeman and Cameron, Reference Larsen-Freeman and Cameron2008). Complex systems theory has been shown to provide a useful framework for discussing the interaction between linguistic and conceptual metaphor. Gibbs and Santa Cruz (Reference Gibbs and Santa Cruz2012) make a convincing case for the dynamic emergence of conceptual metaphor in discourse. They discuss how conceptual metaphors ‘unfold’ over time in conversation according to specific types of dynamics. They see conceptual metaphors as temporary instances of stability in the system, which are described in complex systems theory as ‘attractor states’. Other conceptual metaphors also act on the conversation at the same time, and tend to ‘pull’ it in other directions. Thus no single conceptual metaphor has complete control over the way in which an utterance is interpreted. In their words:

Gibbs and Santa Cruz go on to argue that conceptual metaphors will facilitate metaphor understanding to differing degrees, depending on the interaction of different variables, along multiple timescales, at any given moment in the conversation. This argument is taken even further by Gibbs (Reference Gibbs2013: 60), who shows how any metaphoric utterance, at any time, may be influenced by any or all of the following: evolutionary forces that lead to group cooperation to ensure basic survival, historical forces that have shaped the development of the language that they are speaking, previous topics of conversation that these particular interlocutors have engaged in, social forces determining what it is and is not appropriate to say in certain contexts, and physical sensations that are being experienced by the interlocutors at the time of speaking.

Complex systems theory has recently been applied to the study of metonymy by Biernacka (Reference Biernacka2013) in her study of focus group conversations about terrorism. She found that, in the same way as metaphors, metonymies partially shaped the participants’ understanding of the topic under discussion, and that more than one type of metonymy was exerting a force at any one time, pulling a particular conversation in different directions. Different types of metonymy were called upon to differing degrees depending on the context of the utterance and the timeframe within which it took place. A particular metonymy would be employed by a speaker or writer at a particular point in time for a myriad of reasons and it was impossible to separate the purely cognitive from the purely social. This complex systems approach to metonymy does not usurp the theories that have been outlined above. It simply enriches them by providing further information about when and why particular metonymies are employed. This idea is discussed in depth in Chapter 4, where among other things I explore the role played by metonymy in anaphoric reference and textual cohesion.