The topic of metaphor has a history dating back at least to
Aristotle, but not until the 20th century did it come to be regarded seriously as a problem relevant to the general study of language and thought.
Aristotle took the use of metaphors to be evidence of a superior intellect. At the same time, he seemed to believe that their use was primarily the prerogative of poets and politicians. For philosophers and scientists, they were potentially too misleading. In limiting the use of metaphors to the ornamental, rather than concentrating on their communicative value, Aristotle essentially elevated metaphor from the prosaic to the esoteric. The unfortunate result was that attention to the topic came to be largely restricted to rhetoricians. Indeed, during the 19th century, the principal activity of rhetoricians became the interpretation of particular metaphors (and other tropes) in literary texts — an obsession that was largely responsible for the demise of rhetoric as a serious discipline.
In spite of limiting what he presumably considered to be the legitimate uses of metaphor, Aristotle's analysis of the underlying logic of metaphor has always dominated thinking on the topic. He believed metaphors to be implicit comparisons based on the principles of analogy. It can be argued that this view has had the undesirable effect of blurring the distinctions of metaphor, analogy, and similarity — more will be said of this later.
Around the turn of the century there appeared an English translation of Michel Bréal's
Essay de sémantique. In this influential work, Bréal claimed that metaphor was a linguistic device, widespread in its use, and of great importance in linguistic change. Then, in 1936, metaphor was revitalized in the study of the language of literature by I. A. Richards. Richards introduced useful terminology for talking about metaphors (topic/tenor, vehicle, ground, and tension) — terminology that has come to be fairly standard.
In 1962 Max Black's book
Models and Metaphors appeared. In it, Black rejected the Aristotelian 'comparison' view wherein metaphors are merely elliptical comparisons, and he questioned the generality of the 'substitution' view wherein metaphors are merely ornamental substitutes for literal language. Instead, elaborating on the views of Richards, he proposed an 'interaction' view. On this view, the
tenor (or topic, or subject) of a metaphor is seen as interacting with the metaphorical
vehicle to produce a kind of emergent meaning for the entire sentence — one that could not have resulted from the combination of the tenor with some other predicate, literal or metaphorical. Thus, in the metaphor 'man is a wolf', 'man' is viewed, as it were, through a 'wolf' filter. Our interpretation of both terms is altered.
By the mid-1970s, an active interest in the topic had spread to all sorts of disciplines, most relevant of which, in the present context, was cognitive psychology. Psychologists became interested in questions such as: does the comprehension of metaphors involve special processes not normally involved in the comprehension of literal language? Do metaphors play a role in the development of language? Why do people use metaphors? What is the interrelationship of metaphor, analogy, and similarity? Many of these questions have their counterparts in philosophy, but the empirical techniques that psychology brings to bear may throw more light on them.
The principal dispute over comprehension mechanisms is about whether or not metaphors are understood by first unsuccessfully attempting to impose a literal interpretation. According to one view, when people encounter metaphors they recognize that a literal interpretation is incompatible with the context and then proceed to reinterpret the metaphor figuratively. According to the other view, ordinarily people interpret the metaphorical meaning directly, without even entertaining a literal interpretation. Flatly stated in this way, each view has its problems. One problem with the reinterpretation account is that it does not specify the basis upon which a reinterpretation is made. Many linguistic forms require 'reinterpretation': sarcasm ('That was a clever thing to do,' said of an obviously foolish act), hyperbole ('There were millions of people at the party,' said to indicate that there were many — but obviously not millions), and indirect speech acts, such as indirect requests ('Do you know what time it is?', meant as a request to be told the hour of the day), are just some of the many examples. Reinterpretation theories need to be able to characterize the differences between these (and other) kinds of non-literal uses of language. They need both to specify the different rules that underlie the comprehension of different types, and to offer an account of how people are supposed to know in advance which rules to employ on a particular occasion: it hardly seems likely, for example, that the rules for reinterpreting sarcasm are applied, or even considered, in the ordinary interpretation of a metaphor.
The direct comprehension view finesses this last question. However, as stated, it suffers from underspecification. It amounts to little more than a statement to the effect that, when people encounter metaphors, they understand them; on its own, it has no answer to the question, how? A more general theory of language comprehension is needed to answer this question. The theory (or theoretical framework) that is most often appealed to is one based on the notion of
schemas (or scripts, or frames). The idea is that comprehension in general proceeds by finding a schema that 'fits' the input. Then, the argument goes, the difference between understanding literal and metaphorical language turns out to be simply one of quality of fit.
Data from experiments suggests that elements of both views are correct. Experiments measuring (for example) the time taken by people to indicate that they have understood a sentence (either metaphorical or literal) show that, with sufficient preceding context, subjects do not require the additional time to process metaphors that the reinterpretation theory predicts. On the other hand, when a metaphor is encountered with very little prior context, the predicted increase in
reaction time is found. In such cases, it seems, people have to 'work out' what the sentence means, whereas in other cases what they do seems more like confirming contextually generated expectations. Of course, 'working out the meaning' is something that is often required for literal uses of language too. Furthermore, there are many 'frozen' metaphors in the language, often idioms, that, once known, never require special processing. Thus, the evidence suggests that extra processing is required by language that is not well integrated into the context, be it literal or metaphorical, and that under appropriate circumstances metaphorical language is processed as quickly and easily as literal language.
Observant parents, as well as developmental psychologists, have often noticed that quite young children (2–3 years of age) appear to be very creative in their use of language. When the young child tries to express something for which he has not learned the conventional word, he often uses some other word that succeeds in realizing his communicative intentions, even though the choice may not be literally appropriate. This activity is described by developmental psychologists as 'semantic overextension' (see
language: learning word meanings), but some have claimed that this behaviour in fact reflects the use of metaphor by young children, and that the use of metaphor is a fundamental ingredient in language development.
Attractive though this view might be, it seems to suffer from some rather serious drawbacks. First among these is the evidence that such young children are unable to understand metaphors. Estimates of the age at which children begin to understand metaphors properly vary considerably. Some investigators claim that it is not until early adolescence, while others claim that by the age of 7 or 8 many children can deal with metaphors. The variations in these estimates are the result of several factors, the principal ones being fluctuating and often rather atheoretical criteria for what is to count as a metaphor, and for what is to count as evidence of its comprehension. However, most investigators agree that 3-year-old children cannot understand metaphors. Since it is well established that children's language comprehension is far ahead of their language production, if 2- and 3-year-old children were producing metaphors it would be the only known example of behaviour contrary to the comprehension-before-production rule. A second problem is that the claim that such young children produce metaphors fails to recognize that an utterance could be metaphorical from an adult's perspective, but not from that of the child who produces it. In other words, the claim may be excessively 'adultocentric'.
Doubtless, part of the attractiveness of claiming that very young children produce metaphors is that an important function of metaphors is to permit the expression of ideas that might otherwise be (literally) inexpressible for a particular speaker in a particular language. In some cases the lack of words in the language has resulted in entire domains being mapped into other domains so that the language itself incorporates now unnoticed metaphorical means of description — this is one of the ways in which, as Bréal pointed out, metaphors are important in linguistic change. A familiar example of such a mapping is the use of temperature terms to describe personality characteristics — we talk of people in terms of their being warm, or cool, cold, or icy. When such metaphors are embedded in the language to fill systematic gaps in the lexicon, they often go unnoticed. But the same principle operates at the level of the individual speaker, except that, if the language does not supply conventional resources (be they literal or metaphorical), the speaker may have to create his own. In such cases he uses a novel metaphor.
The expression of the otherwise inexpressible is not the only communicative function that metaphors serve. They also achieve a certain communicative compactness, since all the applicable predicates belonging to the metaphorical vehicle are implied succinctly through the vehicle itself. Thus, even if what a metaphor expresses may have been more or less expressible without the metaphor, its use may be more economical and hence more effective than the long list of predicates that it entails.
The relationship between metaphors and similarity is a complex one. Without having to commit oneself to one of the various theories about how metaphors work, it is apparent that at some level, and in some way, metaphors capitalize on a similarity between the term used metaphorically (the vehicle) and the thing that the metaphor is a metaphor for. Thus, even though it may be incorrect to claim, as some have, that a metaphor is
merely a statement of similarity, it is probably not incorrect to say that a metaphor is
largely a statement of similarity. Clearly, if one says of jogging that it is a religion, the metaphor would not work if jogging and religion were not in some way similar. On the other hand, it is obvious that in many ways jogging is not in the least bit like a religion. It could be argued, in fact, that jogging is not really like a religion at all; if we want something that is really like a religion, a cult is. The interesting thing is that if one now considers the two similarity statements 'Jogging is like a religion' and 'A cult is like a religion', the latter, while appearing to be really true, has no metaphorical potential. This suggests that, if metaphors are based on similarity statements, only some similarity statements can fulfil the required role. The similarity statements that seem to fit the bill are those that themselves seem to be metaphorical. According to this view, that is why one can say 'Metaphorically speaking, jogging is like a religion', but not 'Metaphorically speaking, a cult is like a religion'. Now, if the only similarity statements that can form the basis of metaphors are metaphorical similarity statements, two important consequences follow. The first is that it is futile to attempt to
explain metaphors by reducing them to similarity statements because the statements to which they get reduced still have the characteristic of being metaphorical. The second is that, as a research strategy, the examination of similarity statements may be the best way to uncover the difference between the literal and the metaphorical.
Psychological discussions about the nature of metaphor often seem to use the terms 'similarity' and 'analogy' as though they were interchangeable. It is possible, however, to be more precise about the relationship between the two by arguing that an analogy is a similarity between relations rather than between single-place predicates. On this view, an analogy is a particular kind of similarity statement and, from a psychological perspective, whether a particular comparison is or is not an analogy may depend on the way in which the entities being compared are conceptualized or represented at the time. A simple example will illustrate the point. Suppose we are told that cigarettes are like time bombs. If we entertain this proposition in terms of a simple similarity statement, we might say that both cigarettes and time bombs share the property of (potentially) causing death after a delay. In other words, considered in this way we would have something to the effect: 'Being a cigarette is like being a time bomb.' On the other hand, suppose one conceptualizes the statement in the following way: 'People smoking cigarettes are like people exposed to time bombs.' Now what we have is something roughly equivalent in meaning, except that it is stated as a similarity between two relations — it is an analogy. Again, from a psychological perspective, how the terms in a similarity statement will be represented is likely to depend on the context, it is not fixed by the linguistic structure of the statement itself. Thus, again, the point is not whether metaphors are built on similarity or on analogy, since both are forms of comparison. The point is that metaphors are built on comparisons which are themselves metaphorical, be they analogical or not.
(Published 1987)— Andrew Ortony
Bibliography- Black, M. (1962). 'Metaphor'. In Black, M. (ed.), Models and Metaphors.
- Blasko, D. G. (1999). 'Only the tip of the iceberg: who understands what about metaphor?' Journal of Pragmatics, 31.
- Glucksberg, S. (2003). 'The psycholinguistics of metaphor'. Trends in Cognitive Sciences, 7/2.
- Lakoff, G., and Johnson, M. (1980). Metaphors we Live by.
- Ortony, A. (ed.) (1979). Metaphor and Thought.