December 12, 1993
In his recent paper, "Nominalism and Transference: Meditations on Goodman's Theory of Metaphor", Jonathan Cohen concludes that Goodman's metaphysical commitment to nominalism prevents him from giving a plausible account of metaphor. This conclusion is drawn from two arguments. According to the first, Goodman equivocates between two conceptions of nominalism, and only one of them is compatible with his account of metaphor. Unfortunately, that conception of nominalism suffers from two drawbacks: it is incompatible with Goodman's conventionalism and it does not cogently explain how words get their meanings. In his second argument, Cohen points to a more general explanatory deficiency affecting any nominalist program. He suggests that an aversion to non-extensional elements of meaning deprives the nominalist of the resources necessary for a complete 'hermeneutics' of metaphorical meaning. I will examine these two arguments in turn.
Cohen is correct detecting an equivocation in Goodman's account, and he is correct in saying it involves a blurred distinction between precept and practice. But, I think this equivocation does not involve two conceptions of nominalism. For Goodman, nominalism is an ontological position about the existential status of abstract objects and non-individuals. This ontological position compels a certain conception of the relationship between words and things, but it does not entail any position about how words come to be related to their extensions. In that sense, Cohen is right in saying that a nominalist can believe either that symbolic association comes through precept or that association comes through practice. But this choice is equally available to philosophers with more lenient ontologies. The precept/practice equivocation is not rooted in or dependent on Goodman's ontology. Instead, the equivocation infects his conception of convention.
Goodman recognizes that convention is a "dangerously ambiguous" term.{2} When we say that something is conventional we can mean one of two things: first, we might mean that it is ordinary, usual or consonant with common practice; second, we might mean that that it is artificial, arbitrary, optional or invented. Taking some liberties, we can call these two meanings, 'normal-convention' and 'norm-convention' respectively. The difference between normal- and norm-convention can be detected in the conflicting implications of the phrases 'very conventional' and 'highly conventional.'. But the conflict is not absolute. Something which is norm-conventional can become normal-conventional. For example our practice of going on green and stopping on red reflects an arbitrarily stipulated rule which we all learn at a very young age. But our adherence to this rule has become a matter of normal practice. It is a regrettable consequence of this overlap that we can never get a full grasp on what Goodman means when he calls something conventional.{3}
Cohen's argument is helpful in this regard. The distinction between norm-convention and normal-convention seems to parallel the distinction between precept and practice. A rule of association is norm-conventional if it is stipulated or handed down in the form of a precept, while a rule is normal-conventional if it is determined by practice. When Goodman espouses his conventionalism, he seems to have both of these things in mind. But, Cohen's argument raises some difficulties for Goodman. If a conventional rule of association is determined by practice (i.e., by normal-convention) alone, then the label governed by that rule cannot be said to apply to an object prior to its being used to apply to that object. There is no matter of fact, independent of practice, which determines applicability for expressions whose meanings are normal-conventional. Thus, Goodman cannot say that a novel metaphorical application of a predicate is metaphorically true or false if the association rule of that predicate is normal-conventional. He cannot say it is both novel and familiar, surprising and correct. Consequently, Goodman cannot concurrently claim that a predicate's association rules are normal-convention and that the predicate has a relationship of fit when it is inventively applied. Cohen's critique shows us that Goodman is committed to norm-conventionalism --- the view that association rules are determined by precepts extending from an unspecified source.
However, this conclusion allows Goodman to circumvent the dilemma implied Cohen's argument. Goodman's account of metaphor is perfectly compatible with (one form of) conventionalism. But he can only circumvent the dilemma if he disavows his remarks which are sympathetic to normal-conventionalism and becomes a full-fledged norm-conventionalist. It is not clear that Goodman is willing to do this. Thus, Cohen's argument still has considerable force.
There is one more consideration which Cohen brings to bear against Goodman's commitment to a stipulative account or association rules. He suggests that such an account leaves the origin of such rules mysterious. Goodman's never tells us who stipulates these rules or how they come to have normative force. Without answering these questions, the stipulative account seems less plausible than an account which explains content in terms of practice. These considerations intimate a second dilemma. If Goodman retains his theory of metaphor, he must adopt a norm-conventionalism; but, if he adopts norm-conventionalism, he cannot successfully explain the way words get their meanings. Fortunately for Goodman, this explanatory gap can be filled. Meaning theories which emphasize linguistic practice are commonly called use-theories. Use-theories offer a cogent explanation of how expressions acquire content, but they have recently come under attack. Opponents of use-theories typically admit that practice can affect meaning, but they insist that meaning is not a statistical measure. A word can be misapplied more often than it is applied correctly. But if linguistic practice is not the primary determiner of content, what is? The most popular answer is that many expressions are causally related to their extensions. Two causal theories of meaning are in vogue. On the first, the causal link between an expression and its extension is established in an initial baptism in which a semantic relation is stipulated by the person or people coining the expression.{4} On the second, the causal link is established by a nomological covariance between tokenings of an expression and appearances of members of its extension.{5}
Both of these theories of meaning are compatible with Goodman's account in three important ways. First, they hold that the applicability of a predicate to an object is a fact established independently of the day-to-day use of that predicate. Second, both theories qualify as conventional in one sense of that term. Third, neither theory is incompatible with nominalism. The availability of these two theories of meaning bolsters Goodman's account. They provide an answer to the questions of origin which ostensibly jeopardize norm-conventionalism. As Cohen points out, Goodman wants to have it both ways. He appeals to both precept and practice. But his appeal to practice is at odds with his realism about rules of association. Recent work in the theory in causal semantics provides a way of substantially weakening the role of practice in the determination of meaning. With these tools, Goodman could skirt Cohen's second dilemma.
Of course, nominalists can give a partial explanation of asymmetric dependence. As Cohen points out, Goodman offers a detailed account of how metaphors work. He conceives metaphor as a way of transferring one schema of labels to a foreign realm of objects. This transference imposes certain structural relations within the schema onto its new realm. Goodman acknowledges parallels between his view and the account of metaphor defended by Max Black.{6} According to that account, metaphor involves an interaction between a predicate and an object of predication. The properties associated with the object are reorganized in terms of the properties associated with the predicate (and vice versa). Of course, Goodman cannot talk of properties. But he can talk about the ordering or structure of a schema.{7} For example, in the schema containing temperature words, there is a polarity between 'hot' and 'cold'. When we transfer the temperature schema to a foreign realm, we must preserve this structure. Applying the expressions 'hot' and 'cold' to two objects in that realm will imply a polar relationship between those two objects. In this way metaphor imposes and is constrained by structural features of schemata and realms. If a metaphor cannot establish a structural isomorphism between a schema and a realm, that metaphor will be infelicitous. Thus, Goodman can describe metaphors as successful and unsuccessful without appealing to a comparison of properties. He can say that a painting is sad without saying it has any properties in common with a sad person. He only commits to a structural relationship between sad and happy paintings which parallels the relationship between happy and sad people. Ostensibly, this account answers Cohen's objection. But, I think the answer is insufficient.
Metaphors do not merely reorganize alien realms. If this was all they did, then structurally similar schemata would have the same metaphorical force. For example, the temperature schema (hot/cold), the size schema (big/small), and the height schema (tall/short) seem to be organized in similar ways. But metaphorical applications of labels within these schemata (e.g., 'a hot idea', 'a big idea', and 'a tall idea') convey very different information. It is not clear how the nominalist can explain this fact without appealing to properties. I am not sure whether this is the point that Cohen wants to make, but I think the moral is similar. In order to explain the meaning of a metaphor, it is advantageous to have a relatively rich ontology. Goodman gives us a method of identifying metaphors and some insight into how metaphors work, but he does not tell us why metaphors are useful in language or what particular metaphors mean. He explains how the mechanisms of transference allow us to call a picture sad, and he gives us a rough idea of what kinds of structural constraints that metaphor places on further applications of emotive labels to the realm of pictures. But he does not explain what makes a sad picture sad, or what a sad picture has in common with a sad person, or what the act calling a picture 'sad' conveys. It seems that all Goodman can tell us is that sad pictures and sad people both exemplify a label coextensive with 'sad' and that they can be contrasted with happy pictures and happy people respectively. Clearly, this is an impoverished account. Consequently, I am inclined to share Cohen's suspicion that nominalists forsake the most promising avenues towards an full account of metaphorical meaning.