If, What-If, and So-What:
Daniel H. Cohen
Life is not really a bowl of cherries, and kangaroos do have tails. We know this full well, and yet are willing to entertain their negations in one way or another--the former as metaphor, the latter as counterfactual hypothesis. These are common moves to make in trying to understand the world indeed, characteristically philosophical moves but they are very different in kind. Their similarities, differences, and conflicted interaction reveal something about metaphors and conditionals, but mostly about philosophy itself.
Specifically: (1) "Philosophy" is a term that covers an extraordinary array of discourses, but with the possible exception of completely formal exercises in logic, philosophy is thoroughly metaphorical and largely conditional. (2) Some conditionals and metaphors, however, are incompatible. There are limits to how metaphors can occur in conditionals, and how conditionals can themselves be metaphors. Thus, (3) there is a tension at the heart of philosophy between two moments to philosophy or modes of philosophizing, a speculative, radical-revisionary mode that is metaphorical, and an analytic-explanatory mode that is conditional. Philosophers have special reason not to mix their metaphors (at least not with their conditionals). In this paper, I will focus on that linguistic tension.
§1. Metaphors in Philosophy: Metaphors are invitations to see things in a new way Communism as a spectre haunting Europe, philosophy as a series of footnotes to Plato, or all the world as a stage. Life is a bowl of cherries, for example, is an easy invitation to decline but it still manages to hang around the bad penny of metaphors. Mind as machine, the marketplace of ideas, and language games are all rich metaphors that could once be regarded as incredible or even paradoxical, but that have since been taken as profound truths, the seed for vital philosophical theories, and for some now, common coin of the realm and almost inescapably entrenched ways of thinking.
At its best, philosophy is all about the production of these rich metaphors. Metaphors are the organizing schemes and effective vehicles for philosophical understanding. (1) Although Platos Forms, Augustines Chain of Being, Hegels Dialectic, and other great philosophical theories may have been intended as literal truths, it is as metaphors that they can continue to grip our understanding and imagination if not thoroughly inform our world. (2) Philosophy can be a radical, "revisionary" enterprise, completely reforming the world we live in. When it is, it is by metaphors like these. It is not condescension towards philosophys past, therefore, to regard its products as metaphors. Rather, it is an honesty that is harder to summon up with respect to our own theorizing.
Metaphors are ubiquitous in language and integral to philosophizing. How did it happen, then, that the most language obsessed, methodologically self-conscious of all philosophers this centurys analytic philosophers were able to ignore metaphors so thoroughly and for so long? The blind spot resulted from two sets of assumptions, one about the workings of language, the other about the work of philosophers. Language was conceived as essentially literal and descriptive, with meaning a function of reference and truth-conditions. (3) Philosophys self-image was as a search for ultimate truths about the world, whose proper method was linguistic analysis. Contrary to Hegelian charges, analysis need not be "falsification" because careful analysis yields analytic truths that do not materially change the analysands.
There is a contradiction implicit in this metaphilosophy. Classical empiricism presented itself as a significant body of truths about that special part of the world concerning human knowledge and understanding. Analysis seemed to be the right tool for the job, but analysis is sterile. Analytic truths can be neither profound, ultimate, nor generally descriptive in any satisfactory way. The implicit metaphilosophy of analysis, made explicit by the Positivists, meant that philosophy had to be content with linguistic police-work or theoretical custodial work.
The unpalatability of that conclusion was a crisis for philosophy. To respect the deep sense that this is not what philosophy is all about, a semantics for philosophical discourse is needed. What makes a text philosophical? What is the meaning of the peculiarly philosophical element in philosophical texts? To answer this, the Linguistic Turn needs another (metaphorical) turn, a Metaphorical Turn, for it is philosophys peculiar metaphors and its peculiar use of metaphors that distinguish philosophical discourse, from both ostensibly more literal scientific discourses and ostensibly more figurative literary discourses. (4)
§2. Metaphors and Conditionals: In asking us to imaginatively recreate the world, metaphors are similar to conditionals, for conditionals, too, may ask us to imagine a different world. When we are told that If kangaroos had no tails, they would topple over (5) we are bidden to imagine a world with tail-less kangaroos. That creative task is completable in many ways, limited mostly by our own imaginations. Do we rewrite the evolutionary history of kangaroos to make them tail-less four-footed crawlers? Do we concoct a story about Aboriginal de-caudation rituals? Why not a tale of international black markets with kangaroo tails, not rhinoceros horns, as aphrodisiacs?
Of course, not all conditionals lend themselves to coherent stories this easily, not all the stories we can tell in response to this sort of conditional are equally compelling, and some of the best stories will go beyond the speakers original intent. Interpreting this sort of conditional is a creative endeavor.
Metaphors too can be interpreted indefinitely. Consider the metaphor, popular with such French authors as Mallarmé, Anatole France, Lacan, and Derrida, that words are coins in our commerce with the world. Even a few minutes reflection suffice to generate significant and interesting glosses: words are a culture-bound, local currency; they can suffer inflation; words are tokens; there are rare words treasured for their rarity, counterfeit malapropisms passing as real, and eponyms, the medallions of the linguistic world commemorating individuals; words are how we gain purchase on a phenomenon; and when no word is available, we coin a new one. Words, like coins, have been debased in the modern world. A pre-linguistic state is as imaginable but unrecoverable as a pre-monetary state. And, oh, the possibilities with a rich vocabulary to spend in the world!
Of course, not all metaphors lend themselves to coherent stories this easily, not all the stories we can tell in response to a metaphor are equally compelling, and some of the best stories will go beyond the speakers original intent. Interpreting metaphors can be even more of a creative endeavor than interpreting "imagine-that" conditionals and just as creative as creating metaphors!
Despite these substantial similarities, metaphors are not conditionals and conditionals are not metaphors. (6) While it might be helpful to juxtapose the categories at times, there are two important caveats. First, conditionals are not all of a kind. There are distinct kinds of conditionals, only some of which ask us to imagine other possible worlds. Second, and more important, what "imagine-that" conditionals ask us do is only similar to, not identical with, what metaphors ask us to do: Conceiving the world differently is not the same as conceiving a different world. The most important thing about this difference is its significance for the most important similarity between metaphors and conditionals: their centrality to philosophy.
§3. Conditionals in Philosophy: The semantic analysis of conditionals has a traditional place on the philosophical agenda because of the connections between conditionals and arguments, the focus of logic on argument analysis, and the centrality of logic for philosophy. Accordingly, the search for a formal theory of the logical "if-then " has always been recognized as a legitimate philosophical project. This century, three factors have given the formal study of conditionals greater urgency.
First, there were the great advances in formal logics, notably the development of rigorous axiomatic and algebraic methodologies and their deployment to the fields of multi-valued, modal, and relevance logics. This grew out of and reinforced the traditional concern with conditionals as representative of logical inferences.
Logical Positivisms concern with the language of science provided a second motivation, by bringing the subclass of counterfactual conditionals onto center stage. Positivism took on the twin tasks of unearthing the logical structure of scientific theories and providing them with empiricistically acceptable semantics. In practice, this meant that correspondence rules, or bridge principles, were needed to connect abstract terminology with observable data to ground the meanings of theoretical terms. Constructing a lexicon term by term proved intractable, and recourse to contextual definitions for dispositional properties invoked counterfactual conditionals, making the formal study of conditionals even more of a priority. Since the available models for conditionals were all based on an inferential paradigm the assumption that consequents are consequences that is how counterfactuals were first approached: as enthymematic arguments whose antecedents provided premises from which, in conjunction with laws and initial conditions, the consequents were deducible. (7)
The third factor the "linguistic turn" led to the conception of counterfactuals as revisioning, or "imagine that" conditionals. The enhanced arsenal of semantic, syntactic, and pragmatic machinery was now available for the challenge presented by counterfactuals. Possible worlds semantics for modal logics were exploited with notable success, (8) but one thing this success showed was that logical inference was not always the main ingredient in conditionals. Sometimes, the appropriate validating story for a conditional invokes an imaginative re-visioning of the world, not the technical reconstruction of a deductive argument. (9) The semantic difference is most evident between subjunctive and indicative conditionals: (10)
Differences also appear wholly within the class of subjunctive conditionals, and even within the very restricted class of assertible counterpossibles. E.g.:
both represent part of Euclids reductio proof for the impossibility of a greatest prime. The justification is provided by the deductive context. In contrast, consider these counterpossibles:
These are no less sensible, assertible, or true than either of the previous pair, but they are justified as sensible, assertible, or true differently. In the first case, we imagine a different world (albeit, not a possible world); in the second, we reconsider the actual world. (11)
In the event, the task of providing one semantics for all conditionals remained unmet. Truth conditional semantics stumble over counterfactuals, possible worlds semantics cannot accommodate counterpossibles, axiomatic and metalinguistic approaches are largely inadequate for the gamut of ordinary language conditionals, and pragmatic theories generally fail for the original job of validating logical inference.
While some still seek the Holy Grail of a unified theory of conditionals, reasoned consensus advises abandoning the quest. There are irreducibly different kinds of things that are said and done with conditional locutions. Even within the restricted subclasses of counterfactual and subjunctive conditionals these divisions are still visible. Only some conditionals involve reference to alternative possibilities. But it is just these conditionals, rather than the logical conditionals of implication and entailment, that are important for speculative philosophy and for speculating about philosophy. For much of philosophy, what we mean by "what-if" is more telling than what we mean by "what follows."
§4. Different tasks: The second caveat is that seeing the world differently is not the same as seeing a different world. Metaphors ask us to do one thing; what conditionals ask us to do is often something else. Sometimes the tasks are nearly congruent; other times they are incompatible.
One difference is clear from the round square examples: imagining a different world (If there were round squares, ) is not the same as reconsidering this world (If there are round squares, ). What metaphors ask us to do is more like the latter, yet different still. The conditional asks to consider, at least hypothetically, a change in what we believe about a world. A metaphor may do this too, but mostly it asks us to consider a change in how we see the world, not what we see in the world. It is more of a call for a change in attitude than a change in belief. (12)
Suppose we heard someone call Ludwig Wittgenstein a "round square." What would that mean? How could we interpret it? Because it is manifestly absurd, we would surely take it as metaphor, but that just labels the phenomenon rather than explains it. We do not imagine a world in which both Wittgenstein and geometry are different to interpret that utterance. Nor would we necessarily have to change anything at all about our picture (in a Tractarian sense!) of this world, as if this were a new datum for his biographers. Instead, we might take another look at what we already know about that philosopher, trying to see him as a round square. Perhaps we would see him as two complete but incompatible figures rather than as a single coherent one. Or we might see him as impossibly inconsistent rather than just frustratingly enigmatic. Or, if we are versed in his technical vocabulary, we might recognize the characterization as senseless (sinnlos) but not nonsensical (unsinnig) and conclude the point must be that the nature of the man is something unsayable. There are, remember, lots of stories that can be told, limited only by our own cleverness.
Although Wittgenstein was not literally a round square, he can be seen as such. The fact that there are no round squares makes it no more difficult to see him as one than the absence of ogres, angels, or dragons prevents us from describing other people in those terms. To see Wittgenstein as a round square is not to imagine a world in which he is a round square. Indeed, seeing him as a round square is not merely different than imagining a world in which he is one, they are incompatible. One precludes the other.
§5. Philosophical Incompatability: Metaphors are proposals of a sort, suggesting that we see part of the world in a certain way. (13) Thus, regardless of whether they make the same proposals that counterfactuals do, metaphors should not be able to occur in the conditioning clauses of conditionals. This parallels a problem that plagued the emotivist program for explaining ethical discourse. Proposals, like imperatives, promises, and other illocutionary acts, can be conditioned but cannot themselves serve to condition other speech acts. (14) Although a promise can be the condition for other acts If you promise to do x, then I will do y the locution "If I promise, " cannot be used to make a promise. We can condition an imperative If you are cold, shut the door but we cannot condition with an imperative: the attempted converse If shut the door, then you are cold does not pass grammatical muster.
Nevertheless, metaphors apparently do occur as the antecedents of conditionals. The following conditionals, with metaphors in both the antecedent and consequent positions, are viable:
The issue is not whether these say something intelligent about their subjects, but whether they represent linguistic possibilities. If so, then whether or not they make sense, we can make sense of them. Still, there is something odd about such putative "counterliterals."
Significantly, the metaphors in the antecedents of (4) - (6) are related to the metaphors in their respective consequents. The conditionals serve to explore, extend, or as in (6), by way of an enthymematic modus tollens reject the antecedent metaphor. In each case, the metaphor occupying the antecedent, is "antecedent" in another sense: it must conversationally or dialogically precede the conditional. It must already be on the table so that it can be operated on, as if it were literal.
Similarly, the antecedent position readily accommodates metaphors that are obvious or have a pre-established usage:
An established metaphor is being applied. It is incidental that the trope, "down in the dumps," is a metaphor. Its metaphoricity is set aside for the purposes of the conditional. A literal paraphrase could be substituted without any real damage something not generally true for metaphors. Pre-existent conventions prescribing their use, or conceptual truisms determining their use, obviate any need for the kind of interpretive creativity that characterizes fresh, "imaginative" metaphors. Conditionals like these are only apparently "counterliterals."
These roles, then, are open to conditionals with metaphors as antecedents: applying an established, obvious or pre-existent metaphor and extending or exploring a conversationally given metaphor. In neither case does the conditional ask us to imagine an alternative world. Thus, they are closer to the "what-follows" use of conditional locutions than to the "imagine that" use. The antecedent-consequent connection is argumentative, if not exactly deductive.
The square peg that does not fit any of the available slots is the combination of imaginative metaphors with imagine-that conditionals. They would put too great a strain on the semantic imagination. Consider the following conditionals:
Like (3a) and (3b), these say different things: (8a) extends the metaphor, (8b) resists it, but the difference is not a difference in kind. It is not a question of providing an argument versus telling a possible-worlds justifying story. Possible worlds semantics can shed great light on counterfactuals, but they are not available for counterliterals: (8b) does not ask us to imagine an alternative world in which Wittgenstein is a round square. Conditionals can do that, however, when the antecedent metaphor has a literal use or counterpart:
Metaphors are by nature subject to multiple interpretations, but processing a conditional requires some reasonably fixed interpretation of the antecedent in order to evaluate the consequent and, consequently, the entire conditional. Thus, the conditional in (8a) presupposes a conversational context hospitable to the antecedent metaphor. Otherwise, Gricean maxims of both clarity and relevance would be violated.
On the face of it, the antecedent metaphor Wittgenstein is a round square would violate Gricean maxims, even in discussions of Wittgensteins oddities, because round squares are irrelevant and the metaphor is obscure. There is an important difference, however. On its own, the metaphor does not merely violate Grices maxims, it flouts them, triggering the mechanisms of conversational implicature. Since conditionals antecedents can be false without problem the locution is designed for that maxims of quality cannot be violated or flouted. Maxims of relevance and clarity, however, can be, so in what-follows conditionals, the antecedent has to be acceptable as a working hypothesis, and flouting either of those maxims would undermine the conditional altogether. (16)
Imagine-that conditionals establish their own conversational contexts. That is what their antecedents do in invoking possible worlds. But antecedents can serve that function only by having some fixed semantic content. That content need not be detailed or complete the picture of tail-less kangaroos has many blank spots but it must be at least relatively determinate. This is something metaphors are not. Fresh metaphors are not pictures of the world, so they cannot be blurry or clear, complete or incomplete, determinate or not, nor, for that matter, accurate or inaccurate. Thus, fresh metaphors cannot do what the antecedents of imagine-that conditionals need to do. Metaphors and logical conditionals are tools for philosophers. One is for the speculative task of revisioning the world, the other is for explanation and argumentation. The tasks are generally complementary so the difference can be ignored with impunity, but if we do not respect that difference, we may find ourselves analyzing metaphors and seeing logical analyses as metaphorical and thus missing the point on both fronts.
(1) See Lakoff and Johnson 1980.
(2) Philosophys richest, most vital metaphors are not intended as metaphors, but philosophys history is well-understood as a history of metaphors. Danto 1983 suggests this; D. Cohen 1996 develops it further.
(3) Russell expressed this well in his Introduction to Wittgensteins Tractatus: "The essential business of language is to assert or deny facts." Wittgenstein 1961. p. x.
(4) See D. Cohen 1996
(5) Lewis 1973 opens with this example.
(6) There may well be value in these identifications as metaphors. Martinich 1984 proposes interpreting metaphors as the premises of enthymematic arguments. Reading them as the antecedents of elliptical conditionals is similar.
(7) Goodman 1965, while no Positivist, is the locus classicus for this approach.
(8) Lewis 1973 and Stalnaker 1981 are prime contributors.
(9) There is an irony in this history, since modal logics had their origins as logics of entailment, not logics of necessity and possibility, yet they provided the vehicle for seeing counterfactuals as conditionals of an altogether different sort.
(10) Adams 1970. There is some controversy as to whether (1b) is the subjunctive counterpart of (1a). The conditionals given below in (3a) and (3b) escape those objections, so the point about the need for different semantics remains.
(11) The conditional in (3b) is "open," in the terminology of Mackie 1973, since it makes no presuppositions about the antecedent, so it is not really a counterfactual--or counterpossible. There is a different semantic story to tell in its behalf.
(12) T. Cohen 1996 argues for this.
(13) This is worked out in Loewenberg 1975.
(14) The conditioning clause is the antecedent in the "if-then" locution for conditionals, but the consequent in the "only if" locution. Imperatives can be logical antecedents when they are not conditioning clauses, e.g., "Sound the fire alarm only if no one else has" (a very different set of instructions from "If no else has, sound the alarm," in whichthe lack of alarm is sufficient motive, even in the absence of a fire!). See Castañeda 1975, chs. 3 and 4.
(15) A question asked by the columnist Erma Bombeck, in a book with this as its title.
(16) Consider the assertion "If wishes were horses, beggars would ride." The enthymematic modus tollens, as in (6), yields an unconditional denial of the antecedent. The phenomenon is enshrined in the use of the non-metaphorical locution "If p, then Im a monkeys uncle" to emphatically deny p. ("Monkeys uncle"conditionals have "million-times" counterparts "If Ive told you once, Ive told you a million times" that, by way of modus ponens, emphatically and unconditionally affirm their consequents.)
Adams, E., "Subjunctive and Indicative Conditionals," Foundations of Language, 6 (1970), pp. 39-94.
Castañeda, Hector-Neri, Thinking and Doing (Dordrecht: D. Reidel, 1975).
Cohen, Daniel "Schoolhouses, Jailhouses, and the House of Being: Philosophy and Metaphors;" Conference on Narrative and Metaphor Across the Disciplines, Auckland, NZ, July 7-10, 1996.
Cohen, Ted "Metaphor and Feeling," Conference on Narrative and Metaphor Across the Disciplines, Auckland, NZ, July 7-10, 1996.
Danto, Arthur, "Philosophy As/And/Of Literature," APA Eastern Division Presidential Address, Dec. 28, 1983. Reprinted in Post-Analytic Philosophy, C. West and J. Rajchman, eds. (New York: Columbia U. P., 1985).
Goodman, Nelson, Fact, Fiction and Forecast, 2nd ed. (Indianapolis: Bobbs-Merrill, 1965).
Grice, H. P., "Logic and Conversation," in Syntax and Semantics, vol. 3, Speech Acts, P. Cole and J. Morgan, eds. (New York: Academic Press, 1975).
Lakoff, George and Mark Johnson, Metaphors We Live By (Chicago: University of Chicago Press, 1980).
Lewis, David, Counterfactuals (Cambridge: Harvard U.P., 1973).
Loewenberg, Ina, "Identifying Metaphors," Foundations of Language, 12 (1975).
Mackie, J. L., Truth Probability and Paradox, (New York: Oxford University Press, 1973).
Martinich, A.P., "A Theory for Metaphor," Journal of Literary Semantics 13 (1984), pp. 35-56. Reprinted in A. P. Martinich, The Philosophy of Language, 3rd edition (New York: Oxford U.P., 1996).
Stalnaker, Robert "A Theory of Conditionals," in Ifs, W.L. Harper, R. Stalnaker, and G. Pearce, eds. (Boston: D. Reidel Publishing Company, 1981).
Wittgenstein, Ludwig, Tractatus Logico-Philosophicus, D.F. Pears and B. McGuinness, trans. (London: Routledge and Kegan Paul, 1961)