The role played by metonymy in language change

It is well established that metonymy plays an important role in language change and that it therefore has a diachronic as well as a synchronic aspect. This presents the analyst with a problem when it comes to identifying it in text. If metonymy has been the motivating force behind the development of meaning then when do we draw the line deciding whether it still is metonymy or not?

There is a significant body of work on the role of metonymy in grammar and language change. For example, Bartsch (Reference Bartsch, Dirven and Pörings2002) shows how the word ‘noise’ has been metonymically extended to refer to ‘interference’, and the word ‘mother’ has been metonymically extended to refer to typical ‘mothering-type’ behaviour. Essentially, metonymic relationships play an integral role in shaping grammatical meaning even though by and large, through processes of conventionalisation, the metonymies become hidden and do not need to be accessed upon interpretation. Taylor (Reference Taylor, Dirven and Porings2002, Reference Taylor2003) explains this phenomenon in terms of ‘category extension’. According to this approach, words have prototypical meanings which generally relate to concrete experience. They then acquire more abstract meanings via the processes of metonymy and metaphor. Of the two processes, metonymy is the more pervasive. For example, Taylor (Reference Taylor2003: 129) shows how the French word ‘chasser’ has an original meaning of ‘to pursue an animal with the aim of catching/killing it’. The second, more recent, sense of this word (‘to chase someone away’) throws into perspective the common knowledge that if we pursue an animal, it will run away. According to Taylor, there is a metonymic relationship between these two meanings because we can establish a mental connection between them in the given ICM of the hunt.For the metonymy analyst this presents a problem as metonymy has been involved in the development of this word’s meaning, but it is very difficult to say whether or not it is still present as both senses of the word can be considered ‘basic’, and the fact that one is historically older is neither here nor there when it comes to language use. This reflects the underlying fact that metonymy is best described as a process rather than as a static phenomenon that can be identified with certainty.

A form of grammatical change that has a particularly strong association with metonymy is noun–verb conversion, where for example the noun ‘eye’ becomes the verb ‘eyeing’ in the following expression:

Here, the fact that ‘eyeing’ means ‘looking at’ involves an object for action metonymic relationship. Dirven (Reference Dirven, Dirven and Pörings1999) argues that in English, the metonymic relationships that are involved in noun–verb conversion in English can be explained by three different types of schema: the ‘action’ schema, the ‘location’ schema and the ‘essive’ schema.

‘Action’ schemas involve object for action metonymies that focus on the object, such as:

They also involve instrument/manner for action metonymies, such as:

‘Location’ schemas involve place for action metonymies, such as:

Finally, ‘essive’ schemas involve status for action metonymies, such as:

Some of these metonymies are so conventional that many analysts might prefer to label them as literal. It is difficult for the analyst to draw the line in these cases. As has been observed for metaphor, metonymy can exist at different levels of analysis, and a metonymy that is ‘dead’ in the text or in the language as a whole may or may not be dead in the mind of the beholder. Like metaphors, metonymies can be ‘dead’, ‘alive’, ‘sleeping’ or ‘waking’ for different interlocutors in different contexts of use (Müller,Reference Müller2008).

Instrument for action metonymic relationships appear to be particularly productive in English, which means that we still find relatively novel uses, such as the following, which were found by Littlemore and Tagg (in preparation) in a corpus of text messages:

Here the words training and wowing are motivated by instrument for action and effect for cause metonymic relationships.

A further role played by metonymy in language change involves ‘grammaticalization’. This is the process whereby ‘lexical’ or ‘content’ words acquire a grammatical meaning over time and, in some cases, start to lose their content meaning. It is thought to be one of the principal processes underlying language change (Hopper and Traugott,Reference Hopper and Traugott1993) and is uni-directional (i.e. ‘content’ words tend to become ‘grammatical’ words, but not the other way round). Three different theories have been proposed to explain grammaticalisation (see Evans and Green, Reference Evans and Green2006). These are ‘metaphorical extension’, ‘invited inferencing theory’ and the ‘subjectification approach’. Of these three, the latter two invoke metonymy, as they involve changes in perspective and internalisation.

Invited inferencing theory (Traugott and Dasher,Reference Traugott and Dasher2002) emphasises the usage-based nature of language change. An invited inference is one that is suggested by the context, based on the hearer’s world knowledge, their own point of view and their own set of expectations about what they think the speaker is likely to be telling them. Let us look, for example, at three citations form the Bank of English which contain the word since:

The first of these citations has a clear temporal meaning, the third has a causal meaning and the second arguably has both meanings. The temporal meaning of ‘since’ pre-dates the causal meaning, and Traugott and Dasher (Reference Traugott and Dasher2002) argue that the meaning change came about because the temporal meaning, in certain contexts, ‘invites’ a causal interpretation. Over time, this interpretation has become part of the meaning of ‘since’, giving the word a degree of polysemy. Because the two meanings are linked conceptually, one can argue that the meaning change has a metonymic basis.

The subjectification approach (Langacker,Reference Langacker1999) sees meaning change as being motivated by changes in perspective. We can see this in the two citations containing the string ‘down the road’:

In the first of these examples, down the road is literal and involves movement as the monkeys are walking down the road. The second refers to the place that one would get to if one walked down the road, but no walking is involved, so ‘down’ becomes more grammatical and simply operates as a marker of location. The focus is on the end point of the journey. Thus the focus changes and begins to incorporate the subjective viewpoint of the listener. This involves a metonymic shift of meaning within the ‘perception’ ICM.

It is not always easy to distinguish between the two approaches, as we can see in the following two citations containing ‘must’ from the Bank of English:

The deontic meaning in the first citation of ‘must’ involves direct obligation whereas the epistemic meaning in the second citation is more along the lines of ‘it must be the case that’. This meaning also contains a degree of obligation and appears to be a case of category extension. However, the meaning is also internal to the hearer and focuses on the end point of the ‘must’ relationship, so we can see evidence of both the invited inferencing theory and subjectification in this example. Regardless of which theory one uses to explain the process, we can see that there is a metonymic relationship between the two meanings involving the ‘perception’ ICM, but it would be very difficult for the analyst to identify every instance of this type of metonymy as it is largely a historical phenomenon.

We see the same process at work in other cases where modal verbs have taken on particular meanings that are different from their core deontic or epistemic meanings. In her analysis of written academic feedback that undergraduate students receive on their essays, Lee (Reference Lee2013) notes that the expression ‘can be’ is used as a hedge as in expressions such as:

These uses of ‘can be’ highlight one aspect of the epistemic meaning of ‘can’, namely that it is ‘sometimes the case’. This use of ‘can’ involves a potential for actual metonymic relationship.

The role played by metonymy in grammar and language change also becomes apparent when we look at sign language. For example, Sutton-Spence and Coates (Reference Sutton-Spence and Coates2011) point out that the British Sign Language (BSL) sign for Preston North End football club in the UK involves miming the wearing of a pair of glasses. This sign refers (via a salient property for category metonymic relationship) to the fact that in 1922 goalkeeper J. F. Mitchell became the first and last player to wear glasses in an FA Cup Final. This shows how metonymic vehicles can involve historical facts which for most people are long-forgotten, and which are not particularly salient for the individuals using the language. This is an example of metonymic chaining where the glasses metonymically represent the goal keeper and the goal keeper metonymically represents the club as a whole. Again, we encounter the difficult question of where to draw the line between metonymy and ‘literal’ language.

The interaction and overlap between metonymy and metaphor

A further difficulty inherent in metonymy identification is that it frequently co-occurs with and overlaps with metaphor, and it can be difficult to separate out the metonymy component in these cases (Kövecses, Reference Kövecses2013). For example, in the following sentence both Washington and Tehran could be viewed as metonymies for the politicians who work in those places, or they could be seen as personification metaphors, in which Washington and Tehran take on human characteristics:

A similar problem of overlap can be found in the following sentence, where ‘get hot under the collar’ means to get stressed and angry:

This expression can be viewed as an example of an effect for causemetonymy, in that a particular symptom (getting hot under the collar) refers back to its cause (getting stressed and angry). However, usually when the expression is used, there is no real suggestion that the person is actually getting hot under the collar, and in this case it is perhaps more accurately described as a ‘metaphor from metonymy’ (Goossens,Reference Goossens1990), a phenomenon which is discussed in more detail below. Equally, although the ‘pencil in’ example that we saw in Chapter 1 involves metonymy, in many cases, such as the following, there is not necessarily an actual diary or pencil involved:

In this example, ‘pencilled in’ refers to a provisional arrangement in a hypothetical diary, and, as we will see below, might thus be said to be a ‘metaphor from metonymy’ (Goossens,Reference Goossens1990, Reference Goossens, Dirven and Pörings2003). At times it will be unclear whether or not a real diary and pencil were actually involved, and in some such cases it is difficult to tell whether we are dealing with metonymy or metaphor, or both. The same could be said for the term ‘handbagging’ (which was mentioned in the introduction) as it appears in the following example:

It is extremely unlikely that Ronald Reagan, the US President, was, in fact, struck with a handbag in this case, so the expression must be metaphorical, but it does have a metonymic basis so it could be labelled as containing both metaphor and metonymy. These examples underscore the difficulties encountered when metonymy shades into metaphor.

Metaphor and metonymy are intrinsically ‘slippery’ concepts, as are the criteria that are used to distinguish between them. Various researchers (e.g. Dirven, Reference Dirven, Dirven and Pörings2003; Radden, Reference Radden and Barcelona2000) have proposed continua ranging from the purely literal, through metonymy, to metaphor. Croft and Cruse (Reference Croft and Cruse2004: 220) also give examples that suggest intermediate possibilities between metaphor and metonymy, while also warning that what may appear to be intermediacy may be the result of combining distinctly different processes. Thus, in principle, an expression can never be said to be metaphorical or metonymic in any absolute sense, but only for a given user in a given context. The slipperiness of the metaphor/metonymy distinction has been studied in depth by Barnden (Reference Barnden2010). He looks at the two main grounds for differentiation: the idea that metaphor involves similarity while metonymy involves contiguity, and the idea that in metonymy links to the source domain are preserved while in metaphor they are not. He then looks at cases of metaphor and metonymy where these ‘rules’ appear to be broken. He begins by challenging the notion that metaphor relies solely on similarity, pointing out how many ‘primary metaphors’, such as knowing is seeing or more is up (Grady,Reference Grady1997) are in fact based on real-life experiential correlations which are essentially contiguous in nature. In other words, when we are first learning our language and its metaphors, knowing regularly involves seeing, and more regularly involves an ‘upward increase’. The relationship between the source and target domain therefore involves contiguity rather than comparison and is perhaps better described as ‘metonymy’.Primary metaphors are thought to underlie whole hosts of conceptual metaphors, so if they are essentially metonymic in nature then this also applies to the conceptual metaphors that they underlie. Barnden goes on to point out that when someone is feeling down, they may not actually be bent over or drooping but that we could perhaps imagine them to be in this state and this hypothetical reasoning could form part of our understanding of the metaphor. Thus we have to acknowledge that some metonymic processing is at least potentially involved in the comprehension of metaphors such as these. As he puts it, ‘contiguity is in the eye of the beholder’ (Barnden Reference Barnden2010: 10).

In addition to demonstrating how some metaphors can involve contiguity, Barnden also shows how some ‘metonymies’ can involve similarity. An example that he uses here is that of the name of a country, such as the USA to refer metonymically to a sports team that is representing the USA. He points out that the similarity derives from the fact that they only make sense within the context of a sports competition in which different teams play one another. There is therefore a one-to-one correspondence between the country and its team, and a structural analogy is set up within which competitive relationships between the teams correspond to competitive relationships between the countries. He argues that when seen in context, many metonymies in fact involve structural similarities between the source and target domain and can thus be viewed as metaphor.

Another criterion for distinguishing between metaphor and metonymy that Barnden finds problematic is the idea of ‘link survival’. It has been suggested that in metonymy the ‘literal’ meaning of the vehicle term is still very much part of the final meaning of the metonymy whereas in metaphor it is not. So, for example, we might contrast the following metaphor and metonymy, both of which involve the idea of a ‘mouth’:

In the first (metaphorical) expression, there is no actual mouth present, it is simply that the river opens into the sea, like a mouth, whereas in the second (metonymic) expression, the actual mouth is still very much present in the comprehension process as it is involved in the speaking, or in the imitation of speaking. Barnden finds this distinction problematic, citing the example of the metaphor ‘army ant’. Here the ants do exhibit the behaviour of real armies although they are not real armies as such. Reference to the behaviour of a real army is an important part of the comprehension process, particularly the first time one comes across the expression. Therefore the link survives in this metaphor, which makes it difficult to distinguish from a metonymy.

One conclusion that Barnden reaches is that it may be possible to distinguish between metaphor and metonymy on the basis of the use that one makes of the links between the source and target domain, an approach which de-emphasises any intrinsic links that may exist between them. He concludes that rather than asking whether a particular expression is intrinsically metaphorical or metonymic, it is more useful to ask questions such as: what type of similarity does it involve, if any? What sort of contiguity does it involve, if any? Does it involve link survival? And is hypothetical reasoning involved? This approach to the study of metaphor and metonymy is much more refined than previous approaches, and provides a promising line of investigation. It could be refined even further if two further factors were included. These are the nature of metonymy as a radial category (as discussed in Chapter 3) and the fact that, in language at least, there are clues in the text or intonation that often indicate whether or not a particular expression should be understood as metaphor or metonymy.

At times, the distinction between metaphor and metonymy is left deliberately ambiguous, as we can see from the extract below from the novelGoodnight Mr Tom, by Michelle Magorian. At this point in the novel, Mr Tom is remembering the birth of his child, who died of scarlatina soon after being born along with his mother, Rachel, Mr Tom’s wife. Rachel had been fond of painting:

In this extract, the blue paint metonymically represents both the baby boy (as boys are invariably given blue clothes, blue bedrooms, and so on) and Tom’s wife, Rachel (as it reminds him of her love of painting). The fact that the paint is now locked in a box is a metaphor for the fact that Tom’s feelings are now ‘locked away’ and hidden from the outside world. However, if we consider some of the discussion above, we can see that the source domain is still very much present (the pot of blue paint is still in the box), so this metaphor might be said to contain shades of metonymy. The literal references to the colours red and blue at the beginning and end of the passage (‘the familiar colour of scarlatina’ and ‘his lips had turned blue’) strengthen the dramatic effect of the metonymic and metaphorical references to red and blue. This felicitous combination of metaphor, metonymy and literal meaning, and the fact that it is easy to slip between them, contributes to the overall effectiveness of the passage.

Goossens (Reference Goossens1990, Reference Goossens, Dirven and Pörings2003) has identified four main ways in which metaphor and metonymy interact, and refers to the overall process as ‘metaphtonymy’. The first way in which they interact involves metaphor from metonymy in which the experiential basis of metaphor is in fact metonymy. The ‘hot under the collar’ and ‘pencilled in’ examples mentioned above are what would be described as ‘metaphor from metonymy’ as they begin life as metonymy and then develop into metaphor in certain contexts. The second way in which they interact involves metonymy within metaphor. This occurs when a metonymy functioning in the target domain is embedded within a metaphor. For example, in the expression ‘she caught his eye and laughed’ (BNC), ‘his eye’ refers metonymically to the fact that they exchanged glances, but the catching implies a conduit metaphor. Deignan (Reference Deignan2005a) points out that metaphor from metonymy can lead to ambiguous utterances. For example, when we talk of ‘pencilling something in’, it could be the case that we have literally taken out a pencil and noted something in our diaries, or it simply might be the case that we have made a loose arrangement. With metonymy within metaphor there is no such ambiguity. In the above example it could not be the case that the woman literally ‘caught’ the man’s eye.

The third way in which they interact (which Goossens admits is very rare) involves demetonymisation within metaphor. The example that he gives of this is ‘paying lip service’, as in:

At first sight, this expression appears to involve a part for whole metonymy, where the lips stand for speaking, but the expression is only ever used in abstract senses and is therefore always a metaphor. In this metaphor, there is no part for whole metonymy as there is no actual speaking. The expression therefore loses its apparent metonymic element.

The fourth way in which metaphor and metonymy interact (also rare) involves metaphor within metonymy. The example that he uses to illustrate this relationship is ‘get up on one’s hind legs’, as in:

This expression, which means to ‘get up and say something in public’, is essentially a metonymy, but it relies on the metaphorical construal of a human being as an animal.

Researchers working with Goossens’ model have since revealed greater complexity. For example, Wojciechowska and Szczepaniak (Reference Wojciechowska and Szczepaniak2013) looked at idiomatic expressions containing the word hand whose motivation consists of metonymy mixed with metaphor. They identified different types of metaphor–metonymy interaction in the various expressions and found evidence for all four types of ‘metaphtonymy’ that were distinguished by Goossens (Reference Goossens1990, Reference Goossens, Dirven and Pörings2003). However, their data demonstrates more intricate patterns where cumulative and integrated metaphtonymies are combined. They found that the nature of the interplay varies according to whether or not the idioms are in their canonical form, which indicates the importance of phraseology in metonymy analysis. Their findings reflect Geeraerts’ (Reference Geeraerts, Dirven and Pörings2003) claim that ‘metaphtonymy’ is part of a more inclusive model, which allows for other ways in which metaphor and metonymy interact, including, for example, successions of metonymies followed by metaphors, metaphors followed by metonymies, and metonymies followed by metonymies. Geeraerts’ model provides a more comprehensive description of the semantics of composite expressions such as idioms and compounds.

Many of the problems concerning the close relationship between metonymy and metaphor also apply to its identification in other modes of expression. We saw in Chapter 5 that a number of the metonymies identified in advertising, art,film and music shade into metaphor in much the same way as they do in language. Metaphor and metonymy have been found to interact in complex ways in advertising (Pérez-Sobrino, Reference Pérez-Sobrino2011, Reference Pérez-Sobrino, Gola and Ervas2013a) and sign language (Kaneko and Sutton-Spence,Reference Kaneko and Sutton-Spence2012; Taub,Reference Taub2004), and it is often very difficult to extricate one from the other. Moreover, in art, sign language, poetry and dance it is possible for many metonymies to be used simultaneously, and they interact in ways that are somewhat different from language.

In her study of multimodal advertisements, Pérez-Sobrino (Reference Pérez-Sobrino2011, Reference Pérez-Sobrino, Gola and Ervas2013a) discusses the term ‘Greenwashing’, which is a compound word involving the ideas of ‘whitewashing’ and ‘being green’. This is a form of advertising which actively promotes the (often artificial) perception that a company’s policies or products are environmentally friendly. She looks at two advertisements in which the metaphor product x is a green product interacts with the metonymic complex green for nature for nature-friendly product to misleadingly convey a positive image of the product.

The first advertisement is for Saab cars. This advertisement features the following text, accompanied by a picture of a red Saab:

Pérez-Sobrino also points out that the fact that the Saab in the picture is actually red combined with the fact that the word Grrrrrreen resembles the roaring of a lion lends the advertisement a degree of irony, ensuring that it also appeals to drivers who like high-performance cars and may not have much sympathy with the ‘green’ movement. She argues that a full understanding of this advertisement involves a conceptual blending process, in which the boundaries of metaphor and metonymy are blurred.

The second advertisement is for Otis, a company that makes lifts, escalators and moving walkways. Here, the text is:

The word ‘green’ is written in green in this advertisement, and the text is accompanied by the image of several dots building up in number from left to right and changing colour from blue to green as they do so. In this advertisement, the metonymy green for nature for nature-friendly product is combined with the metaphors more is up, good is up and change is forward motion. Again, the advertisement relies on the reader’s ability to blend these metaphors and metonymies in order to reach a full understanding of the advertisement. It is very difficult to tease the metaphor and metonymy apart in these advertisements and to do so would diminish their rich meanings.

Also problematic for the identification of metonymy in other forms of expression is the fine line that exists in some cases between metonymy and more ‘literal’ forms of expression. If we are to see metonymy as a reference point phenomenon as Langacker does, then in a film every camera angle is a metonymy of one kind or another. The problem is that it is impossible to film something without having a camera angle, which makes metonymy synonymous with perspective. Finally, in art, as with language, it is not always easy to see where a particular instance of metonymy begins and ends. We saw this in some of the artworks that were discussed in Chapter 5. A possible way round these problems is to focus on creative or marked forms of metonymy in other forms of expression, insofar as this is possible, in much the same way as with language.

Signalling

One possible way of using linguistic information to help with the automatic identification of metonymy might be to pay attention to the ways in which it tends to be signalled in texts. Two influential studies have identified the use of particular signals for metaphor in spoken discourse. Cameron and Deignan (Reference Cameron and Deignan2003) found a number of signalling devices – such as ‘just’, ‘like’ and ‘sort of’ – that are used to indicate metaphor in spoken discourse. They use the term ‘tuning devices’ to label these devices as this reflects their interactional nature and focuses on their function, which is to alert the hearer to a possible problem in interpretation and to suggest that a metaphorical meaning is intended (Reference Cameron and Deignan2003: 150). Tay (Reference Tay2011) observed a strong co-occurrence of discourse markers such as ‘you know’, ‘and’ and ‘right’ and the use of extended metaphors in his corpus of psychotherapeutic consultations.

At first sight, this appears to offer a promising avenue for metonymy identification. However, on closer inspection, it turns out not to be as productive as one might hope. The following utterance is taken from a recording of an academic working in the International Development Department at a British university when explaining two management models to a student in that department, who is from Kazakhstan:

As she utters these words, the lecturer points to a diagram containing four quadrants, each of which represents a different style of management. By pointing to a particular part of the model, she is telling the student that the Government of Britain has a particular style of management, represented by that quadrant. This involves metonymic relationships between the place in the diagram and the style of management that it represents, and between the actual government and its ‘place’ in the diagram. At first sight, we may conclude that ‘sort of’ may be a good candidate for signalling device indicating metonymy in this conversation. However, when we look at the lecturer’s use of ‘sort of’ in the entire conversation, we realise that it is used to indicate many other things besides metonymy:

1. 1. The government of Britain is sort of there but if you were in a kind of economic task force

2. 2. The sort of performance management kind of goal oriented type of organisation

3. 3. But em painted in a sort of Kazakh way rather than a Chinese way

4. 4. beautiful little dishes, they were sort of er almost like Chinese balls Ma but

5. 5. So these are sort of opposites, right?

6. 6. Figure out what is happening here, what sort of culture it has

7. 7. Think about what sort of organisation you might be wanting to work for

8. 8. We can think of open systems, the sort of organisation usually small ones where

The first two uses of ‘sort of’ in this list are indeed used to signal metonymy. However, in citations three to five, it is used to indicate vagueness and in citations seven and eight it is used to indicate a ‘type of’. It is not simply used to signal metonymy. However, these findings are interesting nonetheless. We have already discussed the relationship between metonymy and both vagueness and part–whole relationships in previous chapters, so the use of ‘sort of’ here appears to be indicating some sort of radial category that includes metonymy along with the semantic and pragmatic meanings with which it is often associated. However, it is insufficiently refined to serve as an indicator of ‘pure’ metonymy. It could be that the search term simply needs to be more specific than this and that the corpus search needs to be more refined. The first instance of metonymy in this list is signalled by the string: ‘is sort of’. Could it perhaps be the case that the string ‘is sort of’ might be more indicative of metonymy? Unfortunately, a British National Corpus (BNC) search for this string reveals that this is not the case. Of the first fifty lines to appear in a search for ‘is+sort+of’, only three contain metonymy. This example suggests that this particular signalling device is by no means unique to metonymy. On the other hand, we can see in the eight instances above that the cluster of functions served by the string ‘sort of’ do seem to be related to one another in some sort of radial category, a fact which has the potential to be exploited in future, perhaps in combination with other features, for the automatic detection of metonymy or related linguistic features.

It could be that there are other signalling devices that are better predictors of metonymy. These could be verbal devices or they could involve gesture and intonation. It would be interesting to investigate whether this is indeed the case. Or it could simply be that metonymy does not tend to be signalled as much as metaphor. This could well be the case. As we will see in Chapter 7, from a psycholinguistic processing point of view, metonymy is closer to literal language than metaphor is, and it could well be that speakers are simply not generally aware of the fact that they are using metonymy. This would not be surprising as, unlike metaphor, metonymy is rarely studied in schools and many people do not know what it is.

Formal features of metonymy

A more promising way of identifying metonymy might be to focus on its formal features. An analyst using this approach would need to identify ways in which the grammatical patterns surrounding metonymy differ from those that accompany ‘literal’ language and other types of language such as metaphor. It has already been observed that when words are used metaphorically or metonymically there is often some sort of change in terms of their formal features, marking their figurative usage as being different from their meaning in literal contexts (Deignan,Reference Deignan2005a, Reference Deignan2005b). These can take the form of changes in spelling, word class, grammatical patterning and phraseology. It would also be useful to invoke findings from construction grammar (Goldberg,Reference Goldberg2006). These features could usefully be explored, and possibly exploited, in the search for a robust method of metonymy identification.

Let us start by looking at the role of spelling. Barnbrooket al. (Reference Barnbrook, Mason and Krishnamurthy2013) found interesting differences in the meaning of ‘blonde’ and ‘blond’ in the British tabloid press. They found that when the word ‘blonde’ was employed it was used to refer metonymically to a woman with blond hair, whereas when the word ‘blond’ was used it was more likely to be an adjective. Thus in British English (at least in this register), the word ‘blonde’ with its original spelling (carrying over the feminine ending from French) appears to have metonymically narrowed its meaning via a trait for person relationship, to ‘a woman with blond hair’ and the word ‘blond’ with the American spelling has been imported to fill the semantic gap that has been left behind. Thus in order to conduct a corpus search for metonymic uses of the word ‘blonde’ in the British tabloid press, one would need to focus on the older spelling.

We saw some examples of changes in word class that are associated with metonymy in Chapter 1. For instance, ‘pencil’ is never used as a verb, except when it has a metonymic meaning of ‘to pencil [something] in’; ‘muscle’ is not usually an uncountable phenomenon when talked about in the context of a living human being, outside the field of medical discourse; ‘brains’ are usually singular, not plural; ‘handbags’ do not usually appear in the gerund or ‘verbal noun’ form (and gerunds themselves are rarely countable in the way that ‘handbagging’ is); and ‘suits’ are not normally given agency.

As we can see in many of these examples, a grammatical change that has a particularly strong association with metonymy is that of ‘conversion’ leading to denominal verbs (Dirven,Reference Dirven, Dirven and Pörings1999). Denominal verbs can also be extended via a metonymic chaining process to develop ‘verbal nouns’ (Quirket al., Reference Quirk, Greenbaum, Leech and Svartvik1985) or ‘ing’ nouns (Collins COBUILD English Grammar, 2011), such as the expression ‘a good handbagging’, which we saw in the introduction. Verbal nouns (which often involve metonymy) tend to be used uncountably, although the words in the pattern would in themselves usually be countable or singular. Thus something that is normally construed as being countable (e.g. a handbag) becomes part of an uncountable phenomenon when it is used in this construction.

Denominal verbs are also found in Japanese, where they are particularly closely associated with borrowings from other languages. Tsujimora and Davis (Reference Tsujimora and Davis2011) identified a small subset of so-called ‘innovative verbs’. These are denominal verbs that have been derived from borrowing of nouns from other languages and used creatively in Japanese to mean something slightly different from what they mean in their original languages. Although Tsujimora and Davis do not focus on metonymy in their study, it is clear that in many cases the derivational processes involve metonymy, as we can see in the following examples, many of which are taken from English:

Tsujimora and Davis make a convincing case for the idea that innovative verbs such as these operate as a single linguistic ‘construction’. As we saw in Chapter 2, a construction is a grammar pattern whose meaning is not predictable from its component parts (Goldberg,Reference Goldberg2006: 5). Tsujimora and Davis contend that innovative verbs meet these criteria as they all share the same set of properties. Firstly, they tend to undergo a ‘clipping’ or shortening of the base nouns, and when they do, the resulting verbal root must be at least two syllables long, although in Japanese language in general, there are plenty of verbs whose root is only one syllable. Secondly, they all take the same conjugation pattern, in that they all end in ‘ru’ rather than the other, more formal verb ending in Japanese, ‘masu’, and in the past tense they all end in ‘tta’ rather than the more usual ‘ta’. Thirdly, they consistently display a particular accentuation pattern where the stress is on the final syllable of the root (for example, in the word ‘jazu-ru’, the stress is paced on the ‘zu’). This marks them out as being different from other Japanese verbs. Fourthly, they have certain semantic and pragmatic features in common. They are particularly flexible semantically and their meaning can vary considerably from speaker to speaker, depending on factors such as discourse community membership and the age of the speaker. Pragmatically, these are all considered to be playful, informal uses of language, used by the young, and they often constitute an in-group code. Tsujimora and Davis argue that the cluster of properties displayed by innovative verbs suggests that they should be viewed as a ‘construction’ as defined by Goldberg (Reference Goldberg2006) as they constitute a ‘form-meaning-function complex where phonological, morphological, semantic, and pragmatic information is encoded as a collective property’ (Tsujimora and Davis, Reference Tsujimora and Davis2011: 801).

Another grammatical feature that accompanies some types of metonymy, and that might also be considered to be a construction, is ‘predicate transfer’. Here, a particular property is reassigned to an object that would not normally have that property. We saw an example of this in Chapter 3, in the expression ‘I am parked out back’ (Nunberg,Reference Nunberg1995). A person cannot normally be parked, but via a series of metonymic links, the sentence makes sense as we naturally understand that it is actually the person’s car that is parked out back, not the person him/herself. Predicate transfer makes it possible for one to say ‘I am parked out back’ without implying that one has become a car or something that can be parked. It also allows for a sensible interpretation of the utterance ‘we are parked out back’, and for this to refer to a single car, for example in cases where the speakers travelled together.

One could look at construction grammar the other way round and identify metonymy in some of the classic examples that have been proposed by Goldberg. As well as the very obvious role played by metonymy in Goldberg’s well-known example ‘He sneezed the napkin off the table’ (Goldberg, Reference Goldberg1995: 55), we can also find more subtle uses in some of the other widely cited examples of constructions, such as the genitive construction and the ditransitive construction. Although they were not looking at metonymy per se, Wolket al.’s (Reference Wolk, Bresnan, Rosenbach and Szmrecsanyi2013) findings regarding the development of these two constructions are of interest here. They tracked the relative use of the genitive construction (labelled ‘a’ in the following pair) in comparison with a near synonymous construction (labelled ‘b’ in the following pair) over time from the seventeenth century to the present day:

Type (a) constructions have traditionally been more associated with animate objects, while the type (b) constructions have been more likely to occur with inanimate objects. Wolk et al. found that, since the seventeenth century, the relative frequency of type (a) constructions has grown over time while the relative frequency of type (b) constructions has decreased. Moreover, they found that type (a) constructions are increasingly being used with what might normally be seen as inanimate objects, such as companies, countries and machinery. So, for example, expressions such as ‘Spain’s Prime Minister’ or ‘IBM’s marketing strategy’ are much more common than they used to be.

Wolket al. also tracked the relative use of the ditransitive construction (labelled ‘a’ in the following pair) in comparison with a near-synonymous construction (labelled ‘b’ in the following pair) over the same time period:

Again, type (a) constructions tend to be more associated with animate objects, while the type (b) constructions are more likely to occur with inanimate objects. Wolk et al. found that the relative frequency of type (a) constructions has grown over time while the relative frequency of type (b) constructions has decreased, and that type (a) constructions are increasingly being used with what might normally be seen as inanimate objects. For example, expressions such as the following are much more common in English than they used to be:

In these examples, it could be said that inanimate objects are increasingly being viewed as being somehow ‘animate’ and that personification metaphors are on the increase in English. Alternatively, given the blurred boundary between metonymy and personification metaphor that was discussed above, it could also be said that there has been an increase in the metonymic use of inanimate objects to represent the behaviour of those objects or, in the case of organisations, the people who work for them. Therefore one way to begin the search for these particular types of metonymy might involve focusing on these constructions in language corpora.

We have seen in this section that different types of metonymy tend to occur in particular types of grammatical structures and/or constructions. This may well help the reader or listener to distinguish figurative senses from more literal ones and may play a role in identification processes. Future work could usefully identify further sets of constructions that are likely to signal particular types of metonymy. This work would have potential applications to automatic metonymy identification.

The role of genre and register

A final possible aid to the detection of metonymy is an awareness of the types of texts and contexts in which it is likely to be found. We saw in Chapter 4 that Deignanet al. (Reference Deignan, Littlemore and Semino2013), after studying the use of figurative language in a range of contexts, found that certain combinations of genre and register features tended to lead to higher levels of metonymy use. They found metonymy to be particularly prevalent in the language that is used by tightly knit discourse communities, arguing that this is primarily because of its reliance on shared knowledge. They also observed that metonymy serves a slightly more prominent role in spoken discourse than in written discourse. Metonymy was more likely to be prevalent in situations where the communication was primarily concerned with people and entities located within a shared physical space, and where sequences of actions were constrained by time pressures. Examples included busy ‘hands-on’ workplaces and sporting events. This is arguably because metonymy serves the needs of rapid, efficient communication in shared physical settings where interlocutors are under time pressure, and where language plays an ancillary role to extra-linguistic activities. These findings correspond to Harrison’s (forthcoming, Reference Deignan, Armstrong, Herrmann and Sardinha2015) identification of gestural metonymy in a fish-packing factory, where similar conditions prevailed (see Chapter 4). These findings indicate that researchers interested in identifying and exploring metonymy would do well to focus on spoken language employed by closely knit discourse communities in busy hands-on workplace settings. If machines are to successfully identify metonymy and the functions that it performs then they need to take genre and register into account.