p.160
18
Some Consequences for Theories of Conceptual Structure
Any adequate theory of the human conceptual system will have to give an account of how concepts are (1) grounded, (2) structured, (3) related to each other, and (4) defined. So far we have given a provisional account of grounding, structuring, and relations among concepts (subcategoriza-tion, metaphorical entailment, part, participant, etc.) for what we take to be typical cases. We have argued, moreover, that most of our conceptual system is metaphorically structured and have given a brief account of what that means. Before we explore the implications of our views for definition, we need to look at two major strategies that linguists and logicians have used to handle, without any reference to metaphor, what we have called metaphorical concepts.
The two strategies are abstraction and homonymy. To see how these differ from the account we have offered, consider the word buttress in "He buttressed the wall" and "He buttressed his argument with more facts." On our account, we understand buttress in "He buttressed his argument" in terms of the concept buttress, which is part of the building gestalt. Since the concept argument is comprehended partly in terms of the metaphor an argument is a building, the meaning of "buttress" in the concept argument will follow from the meaning it has in the concept building, plus the way that the building metaphor in general structures the concept argument. Thus we do not need an independent definition for the concept buttress in "He buttressed his argument."
p.107
Against this, the abstraction, view claims that there is a single, very general, and abstract concept buttress, which is neutral between the building "buttress" and the argument "buttress." According to this view, "He buttressed the wall" and "He buttressed his argument" are both special cases of the same very abstract concept. The homonymy view takes the opposite tack. Instead of claiming that there is one abstract and neutral concept buttress, the homonymy view claims that there are two different and independent concepts, buttress1 and BUTTRESS2. There is a strong homonymy view, according to which buttress) and BUTTRESS2 are entirely different and have nothing to do with each other, since one refers to physical objects (building parts) and the other to an abstract concept (a part of an argument). The weak homonymy view maintains that there are distinct and independent concepts buttress1 and BUTTRESS2 but allows that their meanings may be similar in some respects and that the concepts are related by virtue of this similarity. It denies, however, that either concept is understood in terms of the other. All it claims is that the two concepts have something in common: an abstract similarity. On this point, the weak homonymy view shares an element with the abstraction view, since the abstract similarity would have precisely the properties of the core concept that is hypothesized by the abstraction theory.
We would now like to show why neither the abstraction nor the homonymy theory can account for the kinds of facts that have led us to the theory of metaphorical concepts---in particular, the facts concerning the metaphorical types (orientational, physical, and structural) and their properties (internal systematicity, external systematicity, grounding, and coherence).
The abstraction theory is inadequate in several respects. First, it does not seem to make any sense at all with respect
p.108
to up-down orientation metaphors, such as happy is up,
CONTROL IS UP, MORE IS UP, VIRTUE IS UP, THE FUTURE IS
up, reason is up, etc. What single general concept with any content at all could be an abstraction of height, happiness, CONTROL, MORE, VIRTUE, THE FUTURE, REASON,and north and would precisely fit them all? Moreover, it would seem that up and down could not be at the same level of abstraction, since up applies to the future, while down does not apply to the past. We account for this by partial metaphorical structuring, but under the abstraction proposal up would have to be more abstract in some sense than down, and that does not seem to make sense.
Second, the abstraction theory would not distinguish between metaphors of the form A is B and those of the form B is A, since it would claim that there are neutral terms covering both domains. For example, English has the love is a journey metaphor but no journeys are love metaphor. The abstraction view would deny that love is understood in terms of journeys, and it would be left with the counterintuitive claim that love and journeys are understood in terms of some abstract concept neutral between them.
Third, different metaphors can structure different aspects of a single concept; for example, love is a journey, LOVE IS WAR, LOVE IS A PHYSICAL FORCE, LOVE IS MADNESS. Each of these provides one perspective on the concept love and structures one of many aspects of the concept. The abstraction hypothesis would seek a single general concept love abstract enough to fit all of these aspects. Even if this were possible, it would miss the point that these metaphors are not jointly characterizing a core concept love but are separately characterizing different aspects of love.
Fourth, if we look at structural metaphors of the form A is B (e.g., LOVE IS A JOURNEY, THE MIND IS A MACHINE, IDEAS ARE FOOD, AN ARGUMENT IS A BUILDING), we find that B (the defining concept) is more clearly delineated in
p.109
our experience and typically more concrete than A (the defined concept). Moreover, there is always more in the defining concept than is carried over to the defined concept. Take ideas are food. We may have raw facts and half-baked ideas, but there are no sauteed, broiled, or poached ideas. In an argument is a building only the foundation and outer shell play a part in the metaphor, not the inner rooms, corridors, roof, etc. We have explained this asymmetry in the following way: the less clearly delineated (and usually less concrete) concepts are partially understood in terms of the more clearly delineated (and usually more concrete) concepts, which are directly grounded in our experience. The abstraction view has no explanation for this asymmetry, since it cannot explain the tendency to understand the less concrete in terms of the more concrete.
Fifth, under the abstraction proposal there are no metaphorical concepts at all and, therefore, no reason to expect the kind of systematicity that we have found. Thus, for example, there is no reason to expect a whole system of food concepts to apply to ideas or a whole system of building concepts to apply to arguments. There is no reason to expect the kind of internal consistency that we found in the time is a moving object cases. In general, the abstraction view cannot explain the facts of internal systematicity.
Abstraction also fails to explain external systematicity. Our proposal accounts for the way that various metaphors for a single concept (e.g., the journey, building, container, and war metaphors for arguments) overlap in the way that they do. This is based on the shared purposes and shared entailments of the metaphorical concepts. The way that individual concepts (such as core, foundation, cover, shoot down, etc.) mix with each other is predicted on the basis of shared purposes and entailments in the entire metaphorical system. Since the abstraction proposal does not have any metaphorical systems, it cannot explain why metaphors can mix the way they do.
Sixth, since the abstraction proposal has no partial
p.110
metaphorical structuring, it cannot account for metaphorical extensions into the unused part of the metaphor, as in "Your theory is constructed out of cheap stucco" and many others that fall within the unused portion of the theories are buildings metaphor.
Finally, the abstraction hypothesis assumes, in the case of love is a journey, for example, that there is a set of abstract concepts, neutral with respect to love and journeys, that can "fit" or "apply to" both of them. But in order for such abstract concepts to "fit" or "apply to" love, the concept love must be independently structured so that there can be such a "fit." As we will show, love is not a concept that has a clearly delineated structure; whatever structure it has it gets only via metaphors. But the abstraction view, which has no metaphors to do the structuring, must assume that a structure as clearly delineated as the relevant aspects of journeys exists independently for the concept love. It's hard to imagine how.
Inadequacies of the Homonymy View
Strong Homonymy
Homonymy is the use of the same word for different concepts, as in the hank of a river and the hank you put your money in. Under the strong homonymy theory of the kinds of examples we have been considering, the word "attack" in "They attacked the fort" and "They attacked my argument" would stand for two entirely different and unrelated concepts. The fact that the same word, "attack," is used would be considered an accident. Similarly, the word "in" of "in the kitchen," "in the Elks," and "in love" would stand for three entirely different, independent, and unrelated conceptsand again it would be accidental that the same word was used. According to this view, English has dozens of separate and unrelated concepts, all accidentally expressed by the word "in." In general, the strong
p. 111
homonymy view cannot account for the relationships that we have identified in systems of metaphorical concepts; that is, it views as accidental all the phenomena that we explain in systematic terms.
In the first place, the strong homonymy position cannot account for any of the internal systematicity that we have described. For example, it would be possible, according to this view, for "I'm feeling up" to mean "I'm happy" and, simultaneously, for "my spirits rose" to mean "I got sadder." Nor can this position account for why the whole system of words used for war should apply in a systematic way to arguments or why a system of food terminology should apply in a systematic way to ideas.
Second, the strong homonymy view has the same problems with cases of external systematicity. That is, it cannot account for the overlap of metaphors and the possibility of mixing. It cannot explain, for example, why the "ground covered" in an argument can refer to the same thing as the "content" of the argument. This holds in general for all the examples of mixing that we have given.
Third, the strong homonymy view cannot explain extensions of the used (or unused) portion of a metaphor, as in "His theories are Gothic and covered with gargoyles." Since that theory has no general metaphors like an argument is a building, it must view such cases as random.
The obvious general inadequacy of the strong homonymy view is that it cannot account for any of the systematic relationships that we have found in metaphorical concepts because it sees each concept as not only independent but unrelated to other concepts expressed by the same word. The weak homonymy view is superior to the strong view precisely because it does allow for the possibility of such relationships. In particular, it holds that the various concepts expressed by a single word can in many cases be
p.112
related by similarity. The weak homonymy view takes such similarities as given and assumes that they are sufficient to account for all the phenomena that we have observed, though without the use of any metaphorical structuring.
The most obvious difference between the weak homonymy position and ours is that it has no notion of understanding one thing in terms of another and hence no general metaphorical structuring. The reason for this is that most of those who hold this position are not concerned with how our conceptual system is grounded in experience and how understanding emerges from such grounding. Most of the inadequacies we find in the weak homonymy position have to do with its lack of concern for issues of understanding and grounding. These same inadequacies will, of course, apply also to the strong version of the homonymy position.
First, we have suggested that there is directionality in metaphor, that is, that we understand one concept in terms of another. Specifically, we tend to structure the less concrete and inherently vaguer concepts (like those for the emotions) in terms of more concrete concepts, which are more clearly delineated in our experience.
The weak homonymy position would deny that we understand the abstract in terms of the concrete or that we understand concepts of one kind in terms of concepts of another kind at all. It claims only that we can perceive similarities between various concepts and that such similarities will account for the use of the same words for the concepts. It would deny, for example, that the concept buttress, when part of the concept argument, is under stood in terms of the physical concept buttress as used in building. It would simply claim that these are two distinct concepts, neither of which is used to understand the other but which happen to have an abstract similarity. Similarly it would say that all of the concepts form or up are not way of understanding concepts partly in terms of spatial orientation but, rather, are independent concepts related by
p.113
similarity. On this view, it would be an accident that most of the pairs of concepts that exhibit "similarities" happen to consist of one relatively concrete concept and one relatively abstract concept (as is the case with buttress). In our account the concrete concept is being used to understand the more abstract concept; in theirs, there would be no reason for there to be more similarities between an abstract and a concrete concept than between two abstract concepts or two concrete concepts.
Second, the claim that such similarities exist is highly questionable. For example, what possible similarities could there be that are shared by all of the concepts that are oriented up? What similarity could there be between up, on the one hand, and happiness, health, control, consciousness, VIRTUE, RATIONALITY, MORE, etc., On the other? What similarities (which are not themselves metaphorical) could there be between a mind and a brittle object, or between ideas and food? What is there that is not metaphorical about an instant of time in itself that gives it the front-back orientation that we saw in our discussion of the time is a moving object metaphor? On the weak homonymy view, this front-back orientation must be assumed as an inherent property of instants of time if expressions like "follow," "precede," "meet the future head on," "face the future," etc., are to be explained on the basis of inherent conceptual similarity. So far as we can see, there is no reasonable theory of inherent similarity that can account for any of these cases.
Third, we have given an account of metaphorical grounding in terms of systematic correspondences in our experience, for example, being dominant in a fight and being physically up. But there is a difference between correspondences in our experience and similarities, since the correspondence need not be based on any similarity. On the basis of such correspondences in our experience, we can give an account of the range of possible metaphors. The weak homonymy position has no predictive power at all and
p.114
seeks none. It simply tries to provide an after-the-fact account of what similarities there are. Thus, in the cases where similarities can be found, the weak homonymy position still gives no account of why just those similarities should be there.
To our knowledge, no one explicitly holds the strong homonymy position, according to which concepts expressed by the same word (like the two senses of "buttress" or the many senses of "in"), are independent and have no significant relationships. Those who hold the homonymy position tend to identify themselves as holding the weak position, where the interdependencies and interrelationships that are observed between concepts are to be accounted for by similarities based on the inherent nature of the concept. However, to our knowledge, no one has ever begun to provide a detailed account of a theory of similarity that could deal with the wide range of examples we have discussed. Although virtually all homonymy theorists espouse the weak version, in practice there seem to be only strong homonymy theories, since no one has attempted to provide the detailed account of similarity necessary to maintain the weak version of the theory. And there is a good reason why no attempt has been made to give such a detailed account of the kinds of examples we have been discussing. The reason is that such an account would require one to address the issue of how we comprehend and understand areas of experience that are not well-defined in their own terms and must be grasped in terms of other areas of experience. In general, philosophers and linguists have not been concerned with such questions.