Summary of Results

As we have seen, the contemporary theory of metaphor is revolutionary in many respects. To give you some idea how revolutionary, here is a list of the basic results that differ from most previous accounts.

• Metaphor is the main mechanism through which we comprehend abstract concepts and perform abstract reasoning.
• Much subject matter, from the most mundane to the most abstruse scientific theories, can only be comprehended via metaphor.
• Metaphor is fundamentally conceptual, not linguistic, in nature.
• Metaphorical language is a surface manifestation of conceptual metaphor.
• Though much of our conceptual system is metaphorical, a significant part of it is nonmetaphorical. Metaphorical understanding is grounded in nonmetaphorical understanding.
• Metaphor allows us to understand a relatively abstract or inherently unstructured subject matter in terms of a more concrete, or at least a more highly structured subject matter.

• Metaphors are mappings across conceptual domains.
• Such mappings are asymmetric and partial.
• Each mapping is a fixed set of ontological correspondences between entities in a source domain and entities in a target domain.
• When those fixed correspondences are activated, mappings can project source domain inference patterns onto target domain inference patterns.
• Metaphorical mappings obey the Invariance Principle: The image-schema structure of the source domain is projected onto the target domain in a way that is consistent with inherent target domain structure.
• Mappings are not arbitrary, but grounded in the body and in everyday experience and knowledge.
• A conceptual system contains thousands of conventional metaphorical mappings, which form a highly structured subsystem of the conceptual system.
• There are two types of mappings: conceptual mappings and image- mappings; both obey the Invariance Principle.

• The system of conventional conceptual metaphor is mostly unconscious, automatic, and is used with no noticeable effort, just like our linguistic system and the rest of our conceptual system.
• Our system of conventional metaphor is alive in the same sense that our system of grammatical and phonological rules is alive; namely, it is constantly in use, automatically and below the level of consciousness.
• Our metaphor system is central to our understanding of experience and to the way we act on that understanding.
• Conventional mappings are static correspondences, and are not, in themselves, algorithmic in nature. However, this by no means rules out the possibility that such static correspondences might be used in language processing that involves sequential steps.
• Metaphor is mostly based on correspondences in our experiences, rather than on similarity.
• The metaphor system plays a major role in both the grammar and lexicon of a language.
• Metaphorical mappings vary in universality; some seem to be universal, others are widespread, and some seem to be culture- specific.
• Poetic metaphor is, for the most part, an extension of our everyday, conventional system of metaphorical thought.

These are the conclusions that best fit the empirical studies of metaphor conducted over the past decade or so. Though much of it is inconsistent with traditional views, it is by no means all new, and some ideas-e.g., that abstract concepts are comprehended in terms of concrete concepts-have a long history.

The evidence supporting the contemporary theory of metaphor is voluminous and grows larger each year as more research in the field is done. The evidence, as we saw above, comes from five domains:

• Generalizations over polysemy
• Generalization over inference patterns
• Generalizations over extensions to poetic cases
• Generalizations over semantic change
• Psycholinguistic experiments

I have discussed only a handful of examples of the first three of these, hopefully enough to make the reader curious about the field. But evidence is convincing only if it can count as evidence. When does evidence fail to be evidence? Unfortunately, all too often. It is commonly the case that certain fields of inquiry are defined by assumptions that rule out the possibility of counterevidence. When a defining assumption of a field comes up against evidence, the evidence usually loses: the practitioners of the field must ignore the evidence if they want to keep the assumptions that define the field they are committed to. Part of what makes the contemporary theory of metaphor so interesting is that the evidence for it contradicts the defining assumptions of so many academic disciplines. In my opinion, this should make one doubt the defining assumptions of all those disciplines. The reason is this: The defining assumptions of the contemporary theory of metaphor are minimal. There are only two. The Generalization Commitment: To seek generalizations in all areas of language, including polysemy, patterns of inference, novel metaphor, and semantic change. The Cognitive Commitment: To take experimental evidence seriously. But these are nothing more than commitments to the scientific study of language and the mind. No initial commitment is made as to the form of an answer to the question of what is metaphor. However, the defining assumptions of other fields do often entail a commitment about the form of an answer to that question. It is useful, in an interdisciplinary volume of this sort, to spell out exactly what those defining assumptions are, since they will often explain why different authors reach such different conclusions about the nature of metaphor.

I started this Chapter with a list of the false assumptions about literal meaning that are commonly made. These assumptions are, of course, false only relative to the kinds of evidence that supports the contemporary theory of metaphor. If one ignores all such evidence, then the assumptions can be maintained without contradiction. Assumptions about literality are the locus of many of the contradictions between the contemporary theory of metaphor and various academic disciplines. Let us review those assumptions. In the discussion of literal meaning given above, I observed that it is taken as definitional that: What is literal is not metaphorical. The false assumptions and conclusions that usually accompany the word literal are:

• All everyday conventional language is literal, and none is metaphorical.
• All subject matter can be comprehended literally, without metaphor.
• Only literal language can be contingently true or false.
• All definitions given in the lexicon of a language are literal, not metaphorical.
• The concepts used in the grammar of a language are all literal; none are metaphorical.

We will begin with the philosophy of language. The Generalization Commitment and the Cognitive Commitment are not definitional to the philosophy of language. Indeed, most philosophers of language would feel no need to abide by them, for a very good reason. The philosophy of language is typically not seen as an empirical discipline, constrained by empirical results, such as those that arise by the application of the Generalization and Cognitive Commitments. Instead, the philosophy of language is usually seen as an a priori discipline, one which can be pursued using the tools of philosophical analysis alone, rather than the tools of empirical research. Therefore, all the evidence that has been brought forth for the contemporary theory of metaphor simply will not matter for most philosophers of language. In addition, the philosophy of language comes with its own set of defining assumptions, which entail many of the false assumptions usually associated with the word literal. Most practitioners of the philosophy of language usually make one or more of the following assumptions.

• The correspondence theory of truth.
• Meaning is defined in terms of reference and truth.
• Natural language semantics is to be characterized by the mechanisms of mathematical logic, including model theory.

These assumptions entail the traditional false assumptions associated with the word literal. Thus the very field of philosophy of language comes with defining assumptions that contradict the main conclusions of the contemporary theory of metaphor. Consequently, we can see why most philosophers of language have the range of views on metaphor that they have: They accept the traditional literal-figurative distinction. They may, like Davidson (1981), say that there is no metaphorical meaning, and that most metaphorical utterances are either trivially true or trivially false. Or, like Grice (1989, p. 34) and Searle (this volume), they will assume that metaphor is in the realm of pragmatics, that is, that a metaphorical meaning is no more than the literal meaning of some other sentence which can be arrived at by some pragmatic principle. This is required, since the only real meaning for them is literal meaning, and pragmatic principles are those principles that allow one to say one thing (with a literal meaning) and mean something else (with a different, but nonetheless literal, meaning). Much of generative linguistics accepts one or more of these assumptions from the philosophy of language. The field of formal semantics accepts them all, and thus formal semantics, by its defining assumptions, is at odds with the contemporary theory of metaphor. Formal semantics simply does not see it as its job> to account for the generalizations discussed in this paper. From the perspective of formal semantics, the phenomena that the contemporary theory of metaphor is concerned with are either nonexistent or uninteresting, since they lie outside the purview of the discipline. That is why Jerrold Sadock in his chapter in this volume claims that metaphor lies outside of synchronic linguistics. Since he accepts mathematical logic as the correct approach to natural language semantics, Sadock must see metaphor as being outside of semantics proper. He must, therefore, also reject the entire enterprise of the contemporary theory of metaphor. And Morgan (this volume), also accepting those defining assumptions of the philosophy of language, agrees with Grice and Searle that metaphor is a matter of pragmatics.

Chomsky's theory of government and binding also accepts crucial assumptions from the philosophy of language that are inconsistent with the contemporary theory of metaphor. Government and binding, following my early theory of generative semantics, assumes that semantics is to be represented in terms of logical form. Government and binding, like generative semantics, thus rules out the very possibility that metaphor might be part of natural language semantics as it enters into grammar. Because of this defining assumption, I would not expect government and binding theorists to become concerned with the phenomena covered by the contemporary theory of metaphor.

Interestingly, much of continental philosophy and deconstructionism is also characterized by defining assumptions that are at odds with the contemporary theory of metaphor. Nietzsche (see, Johnson, 1981) held that all language is metaphorical, which is at odds with those results that indicate that a significant amount of everyday language is not metaphorical. Much of continental philosophy, observing that conceptual systems change through time, assumes that conceptual systems are purely historically contingent-that there are no conceptual universals. Though conceptual systems do change through time, there do, however, appear to be universal, or at least very widespread, conceptual metaphors. The event structure metaphor is my present candidate for a metaphorical universal. Continental philosophy also comes with a distinction between the study of the physical world, which can be scientific, and the study of human beings, which it says cannot be scientific. This is very much at odds with the conceptual theory of metaphor, which is very much a scientific enterprise.

Finally, the contemporary theory of metaphor is at odds with certain traditions in symbolic artificial intelligence and information processing psychology. Those fields assume that thought is a matter of algorithmic symbol manipulation, of the sort done by a traditional computer program. This defining assumption puts it at odds with the contemporary theory of metaphor in two respects: First, the contemporary theory has an image-schematic basis: The invariance hypothesis applies both to image-metaphors and characterizes constraints on novel metaphor. Since symbol-manipulation systems cannot handle image-schemas, they cannot deal with image-metaphors or imagable idioms. Second, those traditions must characterize metaphorical mapping as an algorithmic process, which typically takes literal meanings as input and gives a metaphorical reading as output. This is at odds with cases where there are multiple, overlapping metaphors in a single sentence, and which require the simultaneous activation of a number of metaphorical mappings.

The contemporary theory of metaphor is thus not only interesting for its own sake. It is especially interesting for the challenge it brings to other disciplines. For, if the results of the contemporary theory are accepted, then the defining assumptions of whole disciplines are brought into question.

---

previous | Table of Contents | next