Once More: Colors and the Synthetic A Priori Arthur Pap The Philosophical Review, Vol. 66, No. 1. (Jan., 1957), pp. 9499. Stable URL: http://links.jstor.org/sici?sici=00318108%28195701%2966%3A1%3C94%3AOMCATS%3E2.0.CO%3B2E The Philosophical Review is currently published by Cornell University.
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, noncommercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/sageschool.html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
JSTOR is an independent notforprofit organization dedicated to and preserving a digital archive of scholarly journals. For more information regarding JSTOR, please contact
[email protected] http://www.jstor.org Sat May 12 00:16:13 2007
ONCE MORE: COLORS AND
THE SYNTHETIC A PRIOR1
s AGAINST my earlier contention that such statements as "Nothing is a t the same time both blue and red all over" cannot be analyticthough being unquestionably a prioribecause of the unanalyzability of color predicates,l Hilary Putnam has recently attempted an elegant proof of their analyticity., He succeeds in the proof of the analyticity of the completely general proposition that nothing can have two colors at the same time by proceeding as follows. He selects "indistinguishable with respect to color" as the sole primitive of a constructional system dealing with colors. While it is an empirical fact that this relation is not transitive, we can define "exactly identical in color" on the basis of this primitive in such a way that the relation designated by it is demonstrably not only symmetrical and reflexive but also transitive:
A
Ex, y = ( t ) (It,x = Iz, y) DlNext we define the second level predicate "color" on the basis of "E" as follows: F is a color = there is something, y, such that a thing is F if and only if it is exactly identical in color with y. Formally:
c (F) = (3y) (x) (Fx E Ex, y) D2. I borrow the symbol "E" from R. Carnap's Meaning and Necessity to express logical equivalence. I t seems to me impossible to give the intended definition in extensional terms (material equivalence) for this reason: Suppose that F were a quality other than a color, for example, a shape. I t might happen that the predicates "blue" and "round," for example, had the same extension. Now, if a is the standard patch with reference to which "blue" is defined, then in our hypothetical universe anything which has E to a is round and conversely. But then D, would justify the conclusion that roundness is a color if its definiens were extensional. Next let us formalize the theorem to be proved on the basis of "Logical Nonsense," Philosoply and Phenomenological Research, IX ( I 948), 269283; "Are All Necessary Propositions Analytic?" Philosophical Review, LVIII (1949), 299320; Elements of Analytic Philosophy (New York, 1949), ch. 16 b. 2 "Reds, Greens, and Logical Analysis," Philosophical Review, LXV (1956)~ 206217.
COLORS AND SYNTHETIC A PRIOR1
these (intuitively adequate) definitions; namely, that nothing has two colors a t once:3 (F) (G) [C (F)
. C (G) . F #
G 3 (x) (Fx 3  Gx)]
T.
Putnam is right in saying that T is formally provable on the basis of D, and D,with the help of functional logic, without presupposing any extralogical postulates (the apparatus of functional logic needs to be supplemented, however, by a modal operator, since the proof presupposes that F = G if and only if (x) [Fx Z Gx]). And I admit that I was not aware of this possibility when I argued against the analyticity of the special cases of T. I shall show, however, that the claim that the special cases of I, like "nothing is both blue and red," are analytic involves us in a paradox which no empiricist can afford to accept. Let us not be overhasty and say: since it has been formally proved that nothing can have two colors at the same time, no special proof is required for the proposition that nothing can be both blue and red at the same time; for surely you will not question the analyticity of "Blue and Red are colors." I n the first place, what is at issue is analyticity in the strict sense of "formal provability with or without help of adequate definitions of extralogical terms", not analyticity in the broad sense of "true ex vi terminorum", which I believe means nothing else than oldfashioned, selfevident necessity. I never denied that statements of color incompatibility are true ex vi terminorum, in the sense that it is impossible to question them if one understands them; nor would I put "Red is a color" into a different category. But secondly, observe that the substitution instance of T with "Red" substituted for "F" and "Blue" substituted for "G" reads:
C (Red) . C (Blue) . Red f Blue 3 (x) [Red (x) 3  Blue (x)]
t.
And the analyticity oft which follows from the analyticity of Tis a sufficient ground for asserting the analyticity of the consequent of t only if the antecedent of t is likewise analytic. That the three statements forming the antecedent of t are all analytic is readily shown if we construct explicit definitions of "Red" and "Blue" which these definitions will conform to Putnam's definition of "color"but I n order to keep symbolic complexity at a minimum, I omit the time variable in the formalization. The range of the individual variable may be conceived as consisting of spatially extended entities, whether physical things or sense data is immaterial. If x is a physical thing, then "red ( x ) " is to be read as: the whole surface of x is red.
ARTHUR PAP
make use of proper names, which is the beginning of the paradox I am about to develop. Let Red ( x ) = df E (x, a) (D,), and Blue (x) = df E (x, b) (D,), where a and b are different particulars. And let identity of properties, as usual, be defined in terms of necessary equivalence. Then we may begin by proving the selfcontradictoriness of "Red = Blue," and thereby the analyticity of "Red # Blue," as follows: (Red
= Blue)
(x) (Red(x) E Blue(x))
Z (x) (E(x,a)
E(x,b)).
But the latter necessary equivalence entails the identity of a and b; for it cannot be selfcontradictory to suppose that something has exactly the same color as a without having exactly the same color as b unless a and b are identical. Since "a = b" is selfcontradictory ex hyjothesi, "Red = Blue" is selfcontradictory. Next we tackle "C (Red)"but omit the analogous proof for "C (Blue)". O n the E(x, a)). Hence: (sy) basis of D,, we can assert: (x)(Red(x) (x) (Red (x) 5 Exy), which by D, means that Red is a color. This is all very exhilarating for the constructionist, until he pauses to draw fiom D, and D, the startling consequence that the atomic statements "a is red" and "b is blue" are analytic and that his formal definitions therefore warrant an "ontological leap" to the existence, not of God, but of red and blue things. But how could "a is red" be analytic? The first alternative is that "a" is an ostensive proper name, a symbolic equivalent of "this" or "that," and therefore meaningless outside of a "tokenproducing" context. The statement "a is red" then says that that which a certain speaker is a t a given time pointing at is red; but I discover no selfcontradiction in the negation of such a statement. The second alternative is that "a" is short for a coordinate description. But surely the atomic statements of a Carnapian coordinate language are supposed to be synthetic; it is a contingent fact that such and such a quality occupies a specified spacetime point or region. Finally, "a" may be short for a characterizing description in a thing language, something like "the mail box ) to be analytic, in front ofmy house." Now, in order for "Red ( ( ~ xPx)" P must analytically entail Red.* But I do not see how a simple quality like Red ( I mean an absolutely specific shade of Red) can be analytically entailed by P unless P is a conjunction containing Red. 4 Those who accept the theory of descriptions would have to hold that even this is not a suficient condition, but I would not rest a philosophical argument on so controversial a theory.
96
COLORS AND SYNTHETIC A PRIORI If so, the proof of analyticity will rest on a definition of "P" in terms of "Red," and since we have supposed "a" to be defined in terms of "P", D, then becomes circular. Since the definitions in question, if noncircular, give rise to the consequence that there are contingent existential statements which are analytic, they must be rejected by an empiricist for whom c c analytic" is incompatible with "contingent." Indeed, even intuitively these definitions are not meaning analyses. They are formalizations of ostensive definitionseven fictitious ostensive definitions, for one learns the meanings of color predicates before one knows the meaning of "the same in color," and then only by abstracting a common quality ostensive definition is not meaning from several particularsbut analysis. "Red" and "Blue" are what C. G. Hempel and P. Oppenheim, as well as Carnap, have called purely qualitative predicates, which means that no reference to a specific particular is necessary to explain their meaningsan idea flatly contradicted by the criticized definitions. Surprisingly, then, my earlier argument that, because of the unanalyzability of "blue" and "red," the necessary statement "nothing is both blue and red" is not analytic has not been refuted by Putnam's clever demonstration of the analyticity of Twhich demonstration, be it noted, makes use only of definitions that do not contain individual constants. The philosophical moral I wish to draw from this is that the analyticsynthetic distinctionnot to be confused with the necessasycontingent distinction!has no epistemological significance a t all; that is, no essential difference of cognitive faculties is correlated with the difference between analytic and synthetic ~ t a t e m e n t s . ~ Anybody who knows that nothing can have two different colors at the same time thereby knows that nothing can be both blue and red at the same time, and the way he knows the more special proposition is surely not different from the way he knows the more general proposition; yet the latter can be plausibly argued to be analytic and the former to be synthetic.Vn a way this "moral" is implicitly recognized by Carnap when he broadens what Putnam 5 The reader will find it elaborated in my forthcoming book, Semantics and Necessary Truth (New Haven, Yale University Press, I 957). I should not be misunderstood to have argued that a synthetic proposition may be formally deducible from an analytic proposition. The statement t is formally deducible from T and is just as analytic as T. But the consequent o f t is not formally deducible from T alone but only from T in conjunction with thein my view likewise syntheticantecedent of t. it
A R T H U R PAP
calls the notion of "good old" analyticity, namely, formal provability by means of adequate explicit definitions alone, as follows: a statement is analytic in L if it is Limplied by the meaning postulates of L.' The meaning postulates (like the consequent o f t ) are trivially analytic by this definitionevery meaning postulate Limplies itselfbut they are not analytic in the "good old" sense; that is why they are needed as postulates. Carnap's broadened conception of analyticity incorporates the synthetic a priori within the analytic. But all that has happened is that synthetic a priori propositions in the narrower sense of "synthetic," namely, not formally provable with the help of adequate explicit definitions alone, have been renamed "meaning postulates."s Perhaps this terminological maneuver is justified by the reflection that after all "all bachelors are unmarried" and "nothing is both blue and red" have the important feature in common that nobody who understands what they mean could seriously doubt them. But it should be clear, then, that the modern controversy among analytic philosophers as to whether there are synthetic a priori propositions after all is far removed from the historic battlefield on which empiricists and rationalists met each other in deadly combat. If we reject D, as a meaning analysis, as we surely must, then "Red is a color" is not analytic in the "good old" sense. Yet, is it not unlikely that Kant, had he reflected on this kind of a priori proposition, would have postulated a new kind of "pure intuition" in order to explain how the predicate can be a priori associated with the subject? Again, consider the statement that I is a symmetrical relation, which presumably would occur as a "meaning postulate" in a system based on "I" as a primitive. Clearly, it cannot be analytic in the good old sense, for "I" is a primitive, and the statement is not true by virtue of its form (many statements of the form "for any x and y, Carnap, "Meaning Postulates," Philosophical Studies (1952). I t should not be overlooked that according to the old, narrow conception of analyticity, which left room, so to speak, for the synthetic a priori, a statement is analytic only if a contradiction is formally deducible from its negation, i.e., without using extralogical postulates. For this reason the following proof of analyticity, presented to me once by a famous empiricist, is invalid: x is red = the color of x resembles Red more than any other color; x is green the color of x resembles Green more than any other color; hence " x is green and red" entails "the color of x resembles Red more than Green, and Green more than Red," which is a contradiction. But the latter statement is not formally contradictory; it only contradicts the extralogical postulate that resembling (in a specified respect) y more than z is an asymmetrical relation with respect t o y and z.

COLORS AND SYNTHETIC A PRIORI
if xRy, then yRx" are false). But what capital could a rationalist make of such synthetic a priori knowledge? For notice that insofar as "if xly, then yIx" differs from a mere substitution instance of "if p, then p," its necessity, though "synthetic," has a linguistic origin: it so happens that the sentences by which we communicate thoughts are sequences of signs. This sequential character is irrelevant to some propositions, such as the proposition that an individual x has I to an individual ythough it is relevant to propositions about relations that are not symmetrical. The postulate "if xly, then ylx" merely expresses this irrelevance of the sequential character of sentences of '.the form "xly." If we had a tonal language, with simple tones symbolizing individuals and differences of pitch symbolizing relations between the individuals, the postulate might be redundant: the proposition that aIb might be expressed by sounding, say, a major third containing the tones symbolizing a and b, and this nonsequential