1In recent years, there has been increasing recognition of the fact that metaphor is not simply a literary device designed to embellish language. Rather it is a cognitive phenomenon that pervades all types of language from the most literary to the most mundane (Lakoff 1993, Lakoff & Johnson 1980). We tend to describe abstract concepts metaphorically in terms of more tangible physical experiences. This means that metaphor is a pervasive feature of everyday language. For example, a mundane-sounding expression such as I’m a bit low on cash reflects an underlying conceptual metaphor LESS IS DOWN; she finally let out her frustration reflects the conceptual metaphor THE BODY IS A CONTAINER FOR THE EMOTIONS; and I’m looking forward to tomorrow reflects the conceptual metaphor MOVING THROUGH TIME IS FORWARD MOTION. When language learners come across these types of expressions for the first time, unless there is a direct equivalent in their own language, they may need to think metaphorically in order to work out their meaning. The pervasiveness of metaphor therefore means that it is useful to know what kinds of strategies language learners use to work out the meanings of metaphorical language in the target language.
2In previous studies (Littlemore 2002, Littlemore in press, Littlemore under review), I have empirically investigated the strategies used by students to understand metaphors in highly controlled environments. I have found that they make use of the context and the literal meaning of the words with varying degrees of success, and that the strategies that they use vary according to the contexts in which the metaphors appear, the transparency of the metaphors, and the learning styles of the students. The problem with these studies is that they have tended to look at metaphor interpretation strategies by individual students, or by groups of students in highly controlled, fairly unnatural environments. None of them have observed the naturally occurring behaviour of students dealing with metaphor on an informal basis in an authentic classroom setting.
3In this article, I will describe a small-scale case study in which I used a goal-directed interactive think-aloud technique (Cameron 2003) to determine the metaphor comprehension strategies used by a small group of Japanese students of English. The aim of this study was to see how these students worked together in the group to decipher the meaning of metaphors in the target language. A further aim of the study was to relate these strategies to current theories of metaphor interpretation. I did this so as to gain some insight into the mental processes that the students might actually be employing when they are interpreting the metaphors. These theories also provided a framework within which I could discuss the different types of strategies that the students used.
4In section 2, I will give an overview of the current theories on metaphor interpretation. In section 3, I will outline the experimental approach used in the study. In section 4, I will give examples of the various metaphor interpretation strategies that were employed by the students, relating these strategies to the theories of metaphor comprehension outlined in section 2. Finally, in section 5, I will discuss the implications of the findings for language teaching, and outline ways in which language learners might be helped to interpret metaphors in the target language.
5All metaphors consist of a source domain and a target domain. In a metaphor, such as science is witchcraft, the term ‘witchcraft’ is used to show the author’s opinion of the field of science. In this example the word ‘science’ and its various connotations form the target domain of the metaphor, whereas the word ‘witchcraft’ and its various connotations form the source domain. In order to comprehend this metaphor, one must ‘map’ aspects of the source domain onto the target domain. Appropriate mappings for this metaphor might be that both science and witchcraft are difficult to comprehend, both involve a certain amount of pre-determined procedure, both might involve the mixing together of substances in a container, both involve power and control and neither are to be trusted.
6Metaphor comprehension also involves an appreciation of differences between domains. For example in order to comprehend the metaphor science is witchcraft, not only must one perceive the correlations between the source and target domains (that they are both difficult to understand etc.), but one must also be able to perceive the differences between them (science is ‘good’ whereas ‘witchcraft’ is ‘bad’; ‘science’ is ‘rational’ whereas ‘witchcraft’ is ‘magical’, etc.). The differences between the source and target domains provide the metaphoric tension, which can only be resolved by seeking out correlations between them. Metaphoric thought therefore requires not only an ability to perceive correlations between apparently disparate domains, but also an ability to use the differences between these domains.
7Most theories of metaphor interpretation fall between two general views. The more traditional views of metaphor interpretation assume that in order to interpret metaphors listeners and readers first need to analyse and reject their literal meanings. On the other hand, adherents of the more recent ‘direct access view’ maintain that listeners and readers do not necessarily need to access the complete literal meaning of metaphors in order to understand them, as adequate clues are usually provided in the context to point them in the right direction (Gibbs 2001: 318). The nature of the role played by context is therefore debatable. Does it come into play after all possible interpretations have been activated, enabling the listener to choose between them? Or does it come into play right from the start, restricting all contextually-inappropriate interpretations from being activated? The main theories that are currently used to explain metaphor comprehension are outlined below. These are the graded salience hypothesis, interaction theory, blending theory, the career of metaphor theory, and the class-inclusion model.
8According to the graded salience hypothesis (Giora 1997), highly salient meanings of both the source and target domains are automatically processed during the initial stages of metaphor interpretation. When individuals are interpreting familiar metaphors, both literal and metaphoric meanings are likely to be activated. When they are interpreting unfamiliar metaphors, they may only initially activate the literal meanings as these are most salient. Peleg et al. (2001) argue that two mechanisms are at work here: a linguistic processor produces the immediate activation of salient word meanings, whereas a contextual processor independently shapes the overall message. They found that clues in the context sometimes enable the listener to select the appropriate meaning without having to activate all the features of the source domain.
9The interaction theory (Black 1962) says that people use the target domain of a metaphor as a filter to organize and highlight certain aspects of the source domain while hiding other aspects. According to this theory, there may not be any a priori similarity between the two domains, but new meanings appear as a result of the juxtaposition of the source and target domains. For example, Gineste et al. (2000) point out that when asked to interpret the metaphor a kiss is a fruit, informants come up with concepts such as pleasure, paradise, adoration, need, a token, and coming back for more. None of these concepts are particularly central to the meaning of either kiss or fruit, but they do represent metaphorical or allegorical associations of those domains. In many cases, they reflect the conceptual metaphor SENSUAL PLEASURE IS EATING. When the metaphor is being interpreted, these non-literal associations are activated and highlighted, whereas the more literal senses of the source and target domains tend to be somewhat downplayed.
10Emergent features can also be explained by blending theory, which, in many ways, can be seen as an extension of interaction theory. Under blending theory (Fauconnier & Turner 1998), as we saw above, metaphorical mapping alone does not provide a crucial element of a person’s interpretation of the metaphorical statement. Metaphor comprehension is seen as a process of ‘blending’ or ‘conceptual integration’ of domains. This may result in novel ‘mental spaces’ with their own emergent structure, i.e. they may include elements that are not intrinsically part of either source or target. For example, the proverbial expression Vanity is the quicksand of reason has its own emergent structure: whereas in the physical domain one is likely to be aware of the quicksand when one is trapped by it, in the mental domain one may be unaware of being misguided by vanity (Turner 1998: 65).
11The career of metaphor theory (Gentner & Bowdle 2001) proposes that metaphorical mappings can be accomplished through either comparison or categorization processes, and that there is a shift from comparison to categorization as metaphors become more conventionalised. For example, when a metaphor such as the eye of the needle is encountered for the first time, the listener may have to compare an eye to a needle in order to work out its meaning. However, a listener who is familiar with the expression will simply access the sense of the word ‘eye’ that refers to the hole in the end of a needle. So, when it becomes a conventional metaphor, it has several meanings, and the listener simply needs to access the one that is most appropriate to the context. Only novel metaphors require comparisons to be made.
12According to the class-inclusion model (Glucksberg et al 2001), the two domains of a metaphor are put into a single category containing the attributes that they both share. The listener then activates his or her knowledge of the target domain and fits the source domain concept into this knowledge framework. For example, upon hearing a metaphor such as my lawyer was a shark, a listener will first activate their knowledge concerning the various types of lawyers that there are. He or she will then use the source domain ‘shark’ to choose between these types of lawyer. Research has shown that when a source domain does not match the schema for the target domain, then informants experience significant difficulties in interpreting expressions. For example, the combination ‘unframed planet’ is difficult to interpret, because ‘framedness’ is not a normal characteristic of planets (Glucksberg et al. 2001). The class-inclusion view gains support from the finding that metaphor-irrelevant information is actively inhibited during the comprehension process (ibid).
13Each of the above theories offers a slightly different perspective on the main underlying fact that, for native speakers, metaphor comprehension requires the listener to identify some kind of relationship between the source and target domain. More often than not, this relationship involves non-central features or connotations, of the source or target domain. Most importantly, the successful comprehension of metaphor requires that the listener attends to the knowledge of the source and target domain that they share with the speaker, and this information can include obscure connotations of both domains. This means that, in order to interpret a metaphor, one must be aware of a wide range of connotations for both the source and target domains, and identify those that are being referred to by that particular speaker in that particular context.
14This process is not usually difficult for native speakers, as context and shared knowledge are usually sufficient to help them recover the intentions of the speaker. However, when language learners interpret metaphors, difficulties may arise from the fact that they may have different sets of connotations from their native speaker interlocutors, and that even when they do have similar sets of connotations, they may transfer the wrong ones. They need to access a wide range of connotations for the source domain, and use the context to help them decide which are the most relevant. For example, a student of business might come across the metaphorical use of the word ‘target’ in the expression: companies need to target their customers. Here, the literal sense of the word ‘target’ constitutes the source domain, whereas the business context constitutes the target domain. The student needs to identify connotations of the word ‘target’ that are relevant to the domain of business. These might be, for example, that companies need to identify and focus on particular customers. Students may also lack the necessary cultural knowledge to help them understand metaphors. For example, in Gineste et al’s study (2000), which was referred to above, the informants came up with the concept of paradise in response to the metaphor a kiss is a fruit. This presumably reflects the presence of Christianity in their cultural mindset. Individuals from non-Christian cultures may interpret such a metaphor somewhat differently.
15It would be interesting to know how language learners go about accessing the connotations of the source domain: whether they rely on connotations that are common in their own culture, or whether they try to predict connotations that native speakers may have. It would also be interesting to discover at what point they try to use the context to refine their search. Do they rely on the context right from the beginning, or do they first try out a number of possible interpretations, then use the context to help them decide between them? And what factors lead them to adopt different strategies? In order to answer these questions, I carried out an exploratory study in which I observed the metaphor interpretation strategies used by a group of four students. The study is described in section 3 below.
16The study focused on four intermediate level Japanese learners of English. All of these students were language teachers themselves and were attending a 12-hour course in spoken English. The aims of this course were to improve the students’ own spoken English skills and to help them to reflect upon the ways in which they teach spoken English in their own classrooms. The classes therefore combined language practice with teacher training. During the classes, I regularly introduced metaphoric expressions, and observed the students work out their meaning. Four of the metaphors were deliberately included in the teaching materials (see examples 1, 2, 3 and 5 below), whereas one arose spontaneously during the teaching sessions (see example 4 below). All of the sessions were videoed so as to gain access to visual, as well as verbal indicators of their metaphor comprehension strategies. After the classes, I watched and transcribed the video tapes in order to reproduce the discussions that the students had about the metaphors. This is an adaptation of a ‘Goal Directed Thinkaloud Technique’ (Cameron 2003) in which the metaphor interpretation task forms part of a more general task, and the students verbalise their mental processes as they try to work out the meaning as a group. This technique is slightly limited, as the students are unlikely to verbalise all of the mental processes in which they are engaging. On the other hand, the fact that they are working out the meaning as a group means that they are obliged to share their thought processes to a significant degree.
17In this section, I will report five instances of metaphor interpretation by the students. These are particularly interesting cases as they show that there was considerable variation in the strategies used. They also go some way towards validating the theories of metaphor interpretation outlined in section 2 above.
18In one of our first classes, we were looking at the following extract (Example 1) from McDonough and Shaw (1993: 160) in order to discuss the characteristics of spoken English.
Example 1: ‘Cradle-work’
19Here, the students had difficulty understanding the word ‘cradle work’ which is used in line 1 to refer to the boxes that window cleaners go up tower blocks in, in order to clean the windows. The group discussion was as follows (cf. Discussion on Example 1).
Discussion on Example 1
20The approach taken by the students to work out the meaning of ‘cradle work’ appears to provide support for the graded salience hypothesis. The most salient association of the word ‘cradle’ (baby) was suggested first, even though this association has nothing to do with the context at hand. However, when the students looked more closely at the context, they were able to think of other aspects of a cradle that might explain its use in this context (the fact that it rocks, and that it has high sides). Interestingly, there was a lot of miming during this discussion, which appeared to help the students to work out the meaning of the metaphor. This suggests that they might have been employing some kind of imagery in order to help them understand the metaphor. The miming may also have been a reflection of embodied cognition (Johnson 1987, Lakoff & Johnson 1999). According to the theory of embodied cognition, cognitive processes are deeply rooted in the body’s interactions with the world. In other words, we tend to think about abstract concepts in terms of corresponding physical processes. For example, we think of going forward in time in terms of physically moving forward, and this relationship is often reflected in the types of gestures that people use when talking about future events. The mime observed in this example suggests that the student was actually experiencing a rocking motion, and that this was helping him to work out the meaning of the metaphor.
21In the second example, we were looking at an extract from Carter and McCarthy’s (1997: 85) ‘Exploring spoken English’, in order to study some of the features of informal spoken English.
Example 2: ‘Toss up’
22In this example, I wanted to check whether the students had understood the expression ‘toss-up’. The group discussion was as follows (cf. Discussion on Example 2).
Discussion on Example 2
23This process corresponds roughly to the career of metaphor theory. Student A had learned that the meaning of ‘a toss up’ is to make a decision, and had stored this as one of the meanings of the expression. However, when asked to explore the metaphor in more depth, he was able to deepen his understanding of the expression. When he placed one hand alternately on top of the other, he was probably trying to convey the idea of equal weighting. This less central feature of coin tossing is highly relevant to the target domain. Thus the student appears to have activated his knowledge of the target domain, and fitted associations of the source domain into this knowledge framework. This approach appears to correspond to the class-inclusion model where the two domains of a metaphor are put into a single category containing the attributes that they both share. As with example 2, the miming provides support for the theory of embodied cognition.
24The third example also arose whilst we were looking at the extract from Carter and McCarthy (1997: 85). The part of the extract in question was as follows (Example 3).
Example 3: ‘Pig out’
25In this example, I wanted to check whether the students had understood the meaning of the expression ‘to pig out’. The group discussion was as follows (cf. Discussion on Example 3).
Discussion on Example 3
26In this example, the students worked out the meaning of the expression very quickly. This is somewhat surprising given the number of interpretations that they could have made (behave as messily as a pig; smell like a pig, and so on). Presumably they were able to work out the meaning quickly because of the presence of strong contextual clues. They knew that the people participating in the dialogue were eating cakes and chocolate, and that the topic of conversation had previously revolved around food. This use of the context to limit the number of possible interpretations appears to correspond to the class inclusion model.
27In this example, we were having a discussion about the explicit teaching of grammar in spoken English classes. I used the expression ‘skirt around’ which proved to be problematic for the students (cf. Discussion on Example 4).
Example 4: ‘Skirting the hardest topics’
28In this example, Student A began by picking up on one of the salient features of a skirt: that it serves to hide, or cover, what is underneath. He then used mime in an attempt to work out the meaning. Interestingly, although this strategy did not appear to help him work out its meaning, it did appear to help the student sitting opposite him, who suggested the meaning ‘go round’. This student had picked up on the idea that a skirt is an outer-garment and that it goes round the knees. He then transferred this idea to the context of teaching. Another student continued this line of thought by suggesting the word ‘avoid’. In this example, a more salient meaning of skirt was suggested first (a garment that stretches from waist to knees), then a less central meaning was suggested (an outer garment with a circular hem) as the students attempted to relate the expression to the context. This strategy appears to correspond to the graded salience hypothesis. Inferring the meaning of idioms such as this from their more literal source domains is likely to be a useful language learning strategy. It has been demonstrated that if teachers systematically draw the attention of language learners to the source domains of idioms, then the learners’ depth of knowledge for those idioms, and their ability to retain them improve significantly (see for example Boers 2000). If language learners are able to do this sort of thing for themselves, then this should enable them to work more autonomously and to process the target language at a deeper cognitive level.
29In this example, we were playing a game in which the students had to guess the meanings of idioms. I began by giving each of the students an idiom, along with its correct meaning. I then asked each student to think of two made-up meanings for that idiom. I asked them to read out all three possible meanings to the other members of the group, who had to guess the correct meaning. One of the students had been given the idiom ‘A wolf in sheep’s clothing’, along with its definition ‘a bad person pretending to be a good person’. The two extra meanings that he made up were ‘to feel awkward’ and ‘to cheat someone and steal money’. Interestingly, one of the students in the group thought that the correct meaning was ‘to feel awkward’, and another of the students thought that the correct meaning was ‘to cheat someone and steal money’. Neither of these characteristics (feeling awkward and stealing money) are things that one would naturally associate with a wolf; however they are characteristics that one might project onto a wolf it were found to be wearing sheep’s clothing, and compared to a person. The same is true of the actual meaning of the expression (a bad person pretending to be a good person). This interpretation, like those given by the students appears to involve emergent features, arising as a result of the juxtaposition of the source and target domains. This suggests a merging of the two domains in a third ‘blended space’ domain, which, as we saw in section 2, is a feature of interaction theory and blending theory.
30Thus, in this short study, the approaches taken by the students correspond to all of the main theories of metaphor interpretation. It would be simplistic to say that there was a one-to-one relationship between each of the strategies outlined above, and the five theories of metaphor interpretation. One strategy may reflect more than one theory of interpretation, and one theory of interpretation may be manifested in more than one strategy. Nevertheless, the findings in this study do suggest that the students tended to use not just one strategy but a range of strategies, to work out the meanings of these metaphors. Possible causes of this variation are discussed in section 5, and a number of teaching implications are presented.
31The type of strategy used appears to vary according to the richness of the context in which the expression occurred. If there were few contextual cues (as in example 4), or if they did not appear until after the expression (as in example 1), then the approach taken by the students corresponded most closely to the graded salience hypothesis. In other words, the students accessed the most salient features of the source domain first, and then attempted to apply them to the context. If, on the other hand, the contextual cues were fairly rich (as in examples 2 and 3), then the approach taken by the students corresponded more to the class inclusion model. In other words, they used the context as a defining framework within which to look for relevant features of the source domain. These findings lend support to both theories of metaphor interpretation, and suggest that the strategies employed by language learners are determined by the richness of the contextual clues.
32There was also evidence of blending. Due to the nature of the activity in example 5, there were no obvious contextual clues to help the students understand the metaphor. They simply had to infer a target domain for themselves. However, they may have used previous examples in the activity to infer that the expression was probably used to talk about the (usually negative) traits of people. In this way, the students may have provided their own contextual clues as to the possible nature of the target domain. The fact that the students identified so many emergent features in this example may have been a by-product of their search for an appropriate target domain. The presence of emergent features in their interpretations and in the actual meaning of the expression demonstrates how important it is, when interpreting metaphors, to go beyond the immediately obvious characteristics of the source and target domains.
33Finally, the fact that one of the students had stored the meaning of ‘toss up’ as ‘decision-making’ (example 2) provides support for the career of metaphor theory. Once he had learned its meaning, he did not need to engage in metaphoric thinking in order to retrieve it. However, when pushed to explore the metaphorical relationship between the ideas further, he was able to develop a much richer understanding of the expression.
34One of the most striking things about the students in this study was their extensive use of gesture to help them understand the metaphors. Often, a simple gesture was enough to trigger quite a sophisticated understanding of a metaphor. A particularly interesting finding was the way one student’s gesture triggered another student’s understanding, even though the student making the gesture may not have been able to interpret it him or herself (as in the ‘skirt’ example). This suggests that it may be worthwhile encouraging students to use gesture more when working out the meaning of figurative language, and to watch their peers closely in order to pick up involuntary gestures, which might provide clues as to the meaning of the vocabulary. If, as I said above, the use of such gestures does reflect embodied cognition, then this should prove to be a powerful learning strategy. By attending to this body language students may gain better access to the embodied schemata that lie behind the expressions (see Johnson 1987).
35In summary, the findings from this study suggest that the approaches taken by students toward metaphor comprehension vary according to the richness of the context. Teachers should encourage their students to use both clues in the context and the intrinsic features of the source domain when working out the meaning of metaphors. They should also encourage them to look beyond what is immediately obvious, in order to identify any relevant emergent features. Group work and mime are also key strategies in metaphor interpretation and students should be encouraged to bounce ideas off one another and to pay close attention to their peers’ use of body language.