by M. G. Yoes
Part I: Quine and the Natural Standpoint
This is not a naturalism of method. Quine adopts the content of science by taking at face value what it tells us about the world. The philosopher is no outside critic, no skeptic working at arm's length, but a member of the firm, and as such is entitled to use all available resources. The philosopher is expected to deal with the most general and far-ranging problems in terms of these resources: whether numbers or sets exist, for example, or whether dispositionals can be properly analyzed. Here there is neither room nor need for an additional first philosophy; for on this anticartesian view, science is its own first philosophy.
In «Two Dogmas of Empiricism» vertical reductionism is laid waste.Foot note 1 The empiricist dogma of confirmation of isolated hypotheses by isolated data together with its corollary, the unrevisability of universally confirmed (that is, analytic) sentences, are rejected. This dogma is replaced by the doctrine of the interconnectedness of all sentences -- and not merely all sentences of a given theory, but all sentences of the language as a whole. (This is how the language becomes theory laded and loses its neutrality.) The truth value of any sentence, however empirical or however analytic, is in principle revisable. There is no saying how any evidence will be or should be accommodated; indeed, there is no definable logical relation of confirmation between evidence and hypothesis. Under the new doctrine of interconnectedness of all sentences, the empiricist ideal of a vertical reduction of all statements to a base of sense data statements becomes an impossible dream.
These familiar Quinean doctrines have Quinean consequences. If the overall systematic virtues of conservative connections with the past and inherent simplicity are present and only a few predicted sensory collisions are missed, then a mountain of theoretical terms, or rather their surrogate quantificational apparatus, and their corresponding theoretical entities have a reason for being. Such terms are no longer embarrassing, forever unpaid promissory notes. They have their own respectable posited reality, and reduction to something more solid is neither possible nor desirable. Posits are in, reductions are out.
2. Yet in Word and Object and earlier in «The Scope and Language of Science», there appears a kind of horizontal reductionism.Foot note 2 Thus the philosopher, in the role of full scientific partner, seeks an implementation of scientific results in philosophical research. The philosopher looks for solutions in terms certified by accepted scientific theory, in particular physical theory. The goal is implementation of scientific theory in the task of solving philosophical problems. Indeed, all scientific results are available to the full partner, and, it appears, only such results are available.
Thus adopting the natural standpoint entails more than taking science as given and turning away from skeptical questions. It also means taking science as the ultimate source of a necessary certification of notions to be used in philosophizing from the natural standpoint. Again theoretical terms are the life's blood of science, and there can be no vertical reduction of them to sense data or stimulus meaning. But there is an intended or ideal horizontal reduction, or something approaching it, of all other notions to those of high science.
Again this is not a question of method. It is not enough merely to emulate scientific method, just as it is not enough merely to have access to scientific results. One must be bound by the results of existing science. There is metaphysics beyond the method. As in logical empiricism, the language of science dominates; but with a difference. In Quine's book there is no independent from-the-ground-up empiricical analysis of the language of science, no empirical explanation of why science makes sense and theology does not. Science just is, and it just is our only touchstone.
This feature of the natural standpoint helps explain a long-standing puzzle for many readers of Word and Object who have not forgotten the lessons of «Two Dogmas». Given that theoretical terms are an essential part of scientific procedure, why not treat «analytic», «synonymous», «meaning», and the other terms of traditional philosophical semantics as theoretical terms in a theory of translation or of a theory of language generally? Why is the author of «Two Dogmas of Empiricism», of all people, following a narrow behaviorism, a classical reductivist philosophy? Wasn't it Quine who delivered the fatal blow to reductionist empiricism? Traditional philosophical semantics may turn out to be inferior as a theory, but it seems that the merits of the case are obscured by an arbitrary behaviorism and that the case for philosophical semantics can not get a fair hearing. Anyway, indeterminacy of translation is hardly a startling result, one hears, given behavioristic assumptions.
The natural standpoint explains the unabashed behaviorism. For behaviorism is an obvious application of high science to the problem of translation, a piece of philosophical engineering. As such it is well motivated. We cannot merely introduce theoretical terms as part of a philosophical theory of some subject matter and call them legitimate, however closely scientific methods are followed in building up that theory. Those terms must first somehow be measured against the already legitimate notions of science, whatever those may be.
Chapter II of Word and Object is a careful and determined effort to take the measure of the terms of philosophical semantics in just this way. The outcome is negative but not forgone. The meaning of indeterminacy is that traditional philosophical semantics applied to the problem of radical translation cannot be cashed in scientific-philosophic currency. The evidence for this is that implementation of physics for translation in the behavioristic theory spelled out in Chapter II fails radically to model the traditional theory. The resulting indeterminacy of translation marks the limit of scientific questions about translation and meaning.
While behaviorism is an obvious implementation of the natural standpoint, it is not, as Quine should be the first to admit, uniquely determined by it. Legitimate philosophical questions are no more uniquely answered by the natural standpoint than any other scientific questions. Thus while the behavioristic theory of Chapter II may be an obviously acceptable implementation of physics for language theory, there is no proof that it is the only possible one. The indeterminacy of translation and other results, then, are relative both to this implementation and to the natural standpoint generally. Just as there is always room for new discoveries in science, there is always room for new implementations, new engineering solutions.
The overall project in Word and Object is not pursued from a perspective outside all theories, or even outside all common sense or scientific theories, as is most of «Two Dogmas» and even some important parts of Chapter I. It is written entirely from the natural standpoint. Since «Two Dogmas» there has been in Quine's work both a growing emphasis on the importance of standpoint and a shift to the natural standpoint. Failing to follow this shift of standpoints may account for the chronic misunderstanding of Quine's project by philosophers who, by tradition and commitment, minimize assumptions and maximize critical questions. Word and Object, being written mostly from the natural standpoint, is heavy with presumptions.
3. This flatfooted physicalism may grate on friends of «Two Dogmas « who admire the revolutionary doctrine that all intellectual undertakings from physics and common sense to theology are myth eaten posits and the view of language as an articulated whole.»Foot note 3 Some sentences are closer to the intersection of experience and language and are more quickly given up in the face of contrary experience and less easily defeated by theoretical considerations. And some are further away and less influenced by the vagaries of sense but more easily defeated by theoretical considerations. Science apparently has the pragmatic advantage of better predictions; indeed, one might say that this pragmatic advantage is a defining characteristic of science.Foot note 4 But the point is that the high line on high science is not prominent in «Two Dogmas». Likewise in much of Chapter I the whole of human theories viewed from outside are mere posits. From that lofty view there is not even a unique best physics, many possible conflicting ones being tied for first place. So, much of Chapter I is just a natural development of «Two Dogmas».
But not all. Quine recognizes at the end of that chapter that we have no choice but to judge from some standpoint, from within some theory, and it might as well be the best available. What saves us from a relativism in which criticism is pointless is that «... we continue to take seriously our own particular aggregate science, our own particular world-theory or loose fabric of quasi-theories, whatever it may be.»Foot note 5 Now it is this shift to the natural standpoint which fixes the frame of reference for the translation theory and the ontology of Word and Object and the epistemology and value theory of later work. It is a question of language theory within science, philosophy of language from the natural standpoint. Ontology, epistemology, philosophy of mind, value theory: all are naturalized in turn.
The appearance of the observation sentence does not mark a departure from the holism of «Two Dogmas» after all, a retreat to Protokalsaetze and positivistic ideas about language. The radical holism of that earlier work flows from the standpoint there taken, a minimalist standpoint, as far as possible, outside common sense and scientific theories of the world. From the minimalist standpoint, radical holism remains. On the other hand the role of observation sentences in Word and Object is surely nearly implied by the natural standpoint which figures so prominently. The implementation of physics at least for language theory is a precise sort of behaviorism whose aim is to reconstruct, in terms acceptable to physics or high science, as many of our notions about translation as possible. The observation sentence is the payoff.Foot note 6 Indeterminacy sets in exactly where physics thus implemented leaves off.
What troubles the friends of «Two Dogmas» and no doubt contributes to widespread misunderstanding is that Quine seems in places to treat the natural standpoint not as method or standpoint but fact. Granted he emphasizes that what appears a hopeless relativism from outside of theory building becomes, from within a given theory, a firm ground for judging truth and avoiding relativism. Internal realism, indeed. But in answering critics who may not share the standpoint and in criticizing others, he does not argue the utility of method but the obviousness of scientific result (no change without a change in subatomic states, etc.) The standpoint is transparent even at the level of philosophical exchange.Foot note 7
This position can be usefully contrasted with that of Hume's Newtonian project for human nature. Hume meant to apply the same or similar techniques, look for confirmation in experience, organize things around one or two central ideas, etc. But he surely did not intend to limit himself to concepts which physical science used, accepted and certified or to the use of well-established scientific theories. Perhaps Hume thought there was one and only one reasonable method, a method based on impressions and ideas, and Newton used it for the physical world and he would use it for the human world. The logical positivists, perhaps following Hume, looked to empirical methods. But Quine begins and ends with scientific results.
Moreover, the much discussed additional indeterminacy of the indeterminacy of translation is additional because the natural standpoint, engineered with the scheme of stimulus, stimulus meaning, and the like, leaves untouched the question of which translation scheme is the right one. It is the burden of Chapter II is to show this, to show that the natural standpoint cannot be implemented in this way so as to yield a scientific account which settles most questions of translation. The natural standpoint encompasses a doctrine that is the major premise to the argument that this indeterminacy is additional: science suitably extended by philosophical work is the test of what can be, of what can be true and of what can be a matter of fact. Most questions of translation fail the test and therefore are not questions of fact at all.
4. It is hoped that much of this reconstruction will sound right to Quine's readers. But the key idea of horizontal reduction may not. For the careful reader will know that in places Quine seems explicitly to reject reduction to physics. In his review of Goodman's Ways of Worldmaking he objects to Goodman's acceptance of a large patchwork of worlds or world versions, from physics through common sense and on to worlds created by the arts.Foot note 8 He sees Goodman's acceptance of multiple mutually exclusive physical theories, which is consonant with his own view, as leading Goodman unnecessarily down a slippery slope to a multiplicity of strange worlds: Quine would stop relativization at physical theory.
But for the maker of many worlds there appears no reason to give physical theory a special place. There can be multiple equally right conflicting physical theories. There can also be multiple equally right conflicting or merely incommensurable versions of all kinds, from physical theories to common sense or artistic versions. Physical theory is but one among many world versions which can be right. Now the answer to this, Quine says, is «... not that everything worth saying can be translated into the technical vocabulary of physics; not even that all good science be translated into that vocabulary.»Foot note 9 The special standing of physical theory turns, rather, on its central task of universal coverage: nothing happens without «... some redistribution of microphysical states.»1Foot note 10 Any putative counterexample to this merely calls for a revision in physical theory. Thus the job description of physics bestows a special standing, a standing which only those who reject the enterprise of physics altogether can deny.
From outside the natural standpoint one can see many competing, equally good, irreducible, conflicting physical theories; but the relativism of outside falls before the immanent notions of truth and matter of fact of inside, inside the physical theory bequeathed us by history and tradition. This standpoint is privileged since it alone answers the need for universal coverage. No version of any other kind can even compete in this arena since any real competitor is bound to be another physics. This move of Quine's invites the reply that job definitions are neither analytic nor necessary by Quine's own doctrines. No untenable dualism in language theory; but likewise none in theory of science as well.
Philosophers who think there is more in heaven and earth that is dreamt of in a complete physics face no challenge from Quine to produce physical definitions or to produce translations into physical theory. There might be much «worth saying» even in another science which fails in principle to be translatable -- worth saying, to be sure, but factually empty. One is reminded of tamer logical positivists who proclaim ethics cognitively meaningless notwithstanding its importance (emotional or otherwise). Indeed one reason Quine may have for not wanting to call this reductionism is that he does not seem to believe that indeterminacy has consequences for the practice of translation or that arguments against the sensibility of common psychological language imply that such talk should be abandoned.
Still it seems not far wrong to say that the physicalism espoused in the claim that nothing happens without a redistribution of underlying microphysical states speaks of a kind of reductionism even if it is of neither the term-by-term nor the translational sort. For as Quine sees it this implies that from within the natural standpoint matters of fact are determined by distributions of microphysical states and only by such distributions. Thus if conflicting manuals of translation are physically equivalent in the sense of resting on no distinct distributions of microphysical states, then the translation is indeterminate, and there is no fact of the matter which separates them. It is certainly a kind of ontological reduction.
The philosopher's job, from the natural standpoint, is to implement physics or high science for the solution of philosophical problems; and where there is impossibility of implementation there is indeterminacy; and where there is indeterminacy, there is no fact. In seeking implementations one is doing what any scientist may do, which is to use the results science has so far established; and insofar as one wants a factual philosophical outcome, one must use only such results. Because of physic's special job, its results are the ultimate results. They alone determine the facts.
If this account is correct, then perhaps Quine's philosophy can be more directly understood. The natural standpoint is a unifying principle. Not that it all comes crashing down when the natural standpoint is rejected. Much would remain, of course. Perhaps sweeping away the metaphilosophy would leave the philosophy largely in place. In any case this naturalism is not easily disposed of. What are the alternatives?
5. The role of the natural standpoint and the consequent method of horizontal reduction are central in Quine's philosophy. That is the burden of these observations so far. Are there any problems arising out of these doctrines? Indeed there are. What follows is an application of the doctrine, namely the famous thesis of the indeterminancy of translation, and a discussion of a serious problem it reveals.
Radical translation is translation from an independent language, the language of the Other. Thus for all the translator knows the language of the Other contains predicates like «grue». What bearing does this possibility have on the doctrine of indeterminacy of translation?
The Other speaks Gruese. We would say, characterizing the situtatuon in our language, that the Other's occasion sentence «Grue» prompts the Other to affirm when presented with a green stimulus before t and prompts the Other to affirm when presented with a blue stimulus at some time later than t.Foot note 11 The affirmative stimulus meaning of «Grue» for all speakers of Gruese is at each time the same: for all times before some time t, it includes all green stimulations but no blue ones; and for all times after t, it includes all blue stimulations but no green ones. Thus, the affirmative stimulus meaning of «Grue» contains grue stimuluations and the negative stimulus meaning of «Grue» contains only stimulations other than grue. The observationality of «Grue» may be high, as high as «Green» in English.
The translator who translates «Grue» as «Green» makes an error if the time is not right. The Other who translates «Green» as «Grue» makes an error if the time is not right. To say this defends the determinacy of the translation of «Grue», a defense which is necessary for Quine since in his translation theory all observational sentences are subject to determinate translation. If there is a language like Gruese, like all languages, all of its observation sentences have a determinate translation.
By Quine's account, then, the Other's sentence «Grue» is subject to determinate translation. It is a solid fact that «Grue» translates into some observation sentence of English. This is not to say that translators can make no errors with these sentences. Determinacy is not incorrigibility. A translator may collect his evidence on matters of stimulus meaning with great care, generalize from the evidence with exquisite subtlety, and in the end be mistaken. The point, Quine reminds us, is that in the case of observation sentences there is something to be mistaken about. Since «Grue» is observational in Gruese it is a scientific fact that any observational sentence of English whose stimulus meaning is the same as «Grue» is a correct translation of «Grue».
6. The thesis of indeterminacy of translation is naturally of much philosophical interest. Perhaps it would not be far wrong to compare its importance with that of other famous negative theses such as the theorems of Goedel and Church and the indeterminacy principle of Heisenberg. And just as the fascination of Goedel's theorem depends on there being completeable systems, indeterminacy gains its interest by contrast with determinacy. If there were no significant class of computable functions, there would be less interest in the theorem that establishes the existence of noncomputable ones. The careful constructions of Chapter II of Word and Object -- stimulus, stimulation, modulus of stimulation, stimulus meaning, observation sentence, etc. -- are all in service of finding a precise, scientifically certified sense of determinant translation. How far does such a precisely defined determinacy take us? The thesis of indeterminancy of translation is that this determinancy cannot take us all the way, that in the end a large area of indeterminacy must remain. But observational «Grue» presents a difficulty for that determinacy, a difficulty which subsequent paragraphs explore. If this difficulty cannot be met, the importance of the indeterminacy of translation thesis may be compromised.
7. Though thoroughly observational, the determinacy of «Grue» falls far short of the determinacy of Quine's standard examples. Occasion sentences with little variation in stimulus meaning within the linguistic community are sentences whose stimulus meaning is little effected by extraneous clues, collateral information. «Gavagi»'s stimulus meaning may vary from expert hunter, to stay-at-home novice who knows nothing of gavagis. The Other's word for bachelor, if there is one, varies widely in stimulus meaning across the linghistic community. Technical and scientific terms may vary widely and have thin stimulus meanings. Color words, but not only color words, vary little and thus are counted observational sentences when used as such. The translator-engineer, applying physics to translation problems through stimulus response theory, may for various reasons adjust the boundaries of the observational; observationality is a matter of degree. Still sameness of stimulus meaning across the population is a safe guide to observationality. From there it is just a problem of finding a word or phrase in the home language which has the same, or roughly the same, stimulus meaning.
Nevertheless, evidence of observationality does not increase the likelihood that «Green» is the right translation of «Grue». For unlike other observational sentences, «Grue» is itself underdetermined by scientific results on stimulus meaning. Nothing in the stimulus meaning of any observational predicate, whether grue-like or not, differentiates between mistaken translation and change of meaning. On rechecking the stimulus meaning tables for «Grue» the translator after t finds that green stimuli no longer prompt native affirmations. Nothing in the stimulus meanings will decide whether the original translation was in error, there was a change in meaning, or a live grue predicate has been discovered. Whether Gruese is rife with the likes of «Grue» or has only a few such sentences does not matter. Each such sentence is prima facie a counterexample to the claim that all observational sentences are determinately translatable.
If only there were some universal mark of the grue-like pathological predicate or sentence, we could solve this problem by simply setting them aside. It seems doubtful that any such mark can be found. And without an independent characterization of gruelike sentences for the business of radical translation generally, we cannot be sure that any observational sentence is translatable. And this is a difficulty apart from the humdrum problem of always being somewhat uncertain about the conclusion of an induction. Given that «Grue» is observational and indeterminate, and given that there is no independent way for the radical translator to spot gruelike sentences in the speech of the Other, no amount of empirical evidence about the stimulus meaning of any given sentence can increase the probability that it is not one of the gruelike and that its stimulus meaning is a guide to its translation. Unless we have an independent way of spotting them, some new way must be found to show that «Grue» and its ilk are, despite appearance, determinately translatable.
Again, the thesis of indeterminacy of translation takes its philosophical import from the contrast with that limited class of cases for which determinacy of translation is well-defined and well-established. Without a clear sense of determinacy of translation, there is no interesting sense of indeterminacy of translation. Indeterminacy would be everywhere and thus nowhere.
Such a situation would also be bad for the feasibility of a behavioristic account in general. For if such an account does not cover observational sentences, it is not likely to cover any other class of sentences. Without determinacy, the stimulus-meaning account, though intended to be a test of how far a scientific theory of meaning could take us, would lose its credibility and the famous thesis of indeterminacy of translation its point.
It might be pressed that these examples confuse indeterminacy of translation with indeterminacy of induction. We may never find out the actual stimulus meaning of «Grue» and thus be defeated in our inductions from cases. Still the translation of «Grue» into English is just Goodman's famous formula, whether we can ever discover that fact or not. The evidence that «Grue» is to be translated «Green» can accumulate, but evidence that «Grue» is to be translated by a version of Goodman's formula will not accumulate. Quine's scientific recipe for radical translation, at the first significant series of blue stimulations' prompting of «Grue» from the native, calls for the conclusion that a change of meaning has occurred or that a batch of linguistically deviant natives had been used or that some strange error was made in recording data. Thus Gruese, even in its observational base, completely escapes Quine's method.
8. This suggests the following argument. There can be two equally good translation manuals according to one of which «Grue» is always translated as «Green» and according to the other as «Green» up to t and «Blue» otherwise. Both accord equally well with the hard factual evidence. Therefore «Grue» has no determinate translation. This seems to undermine the whole notion of determinacy as it is set out in Word and Object.
It might be said that this argument fails to appreciate that in Quine's book translation is naturalized. Radical translation takes place within the assumption that science, specifically physics, is true. It is this natural standpoint that gives point to the enterprise. The question the theory of translation tries to answer is that given this, how can we formulate a scientific account of translation between radically different languages? That is, how much of our naive theory of meaning can be scientifically reconstructed? Now given the the theory of evolution is fundamentally correct Gruese is impossible, for projection of «Grue» and the like would promote catastrophe. Speakers of Gruese, or those with a gruese-like prelinguistic quality space, would have fallen by the wayside. Thus within this overall assumption, the grue problem cannot arise.
Perhaps, one might argue, the grue problem does not arise within the natural standpoint. But Quine himself in «Indeterminacy Again»Foot note 12 shows that the native's physics may be quite different from ours and that our translation of his physics up to observationality is radically indeterminate even on the assumption that our physics is the right one. The grue argument here presses the point that the qualification «up to observationality» be dropped. And with it the whole idea of determinate translation even within the natural standpoint. Moreover, the evolutionary argument may fail to be convincing for a reason cited above, that there is no known means of giving an independent characterization of grue-like predicates.
9. This argument from «Grue», if it succeeds, prompts a new perspective on translation. Evaluation has a hidden link to habit, custom, tradition. A valid deductive argument remains defective if a premise is false or it begs the question. A inductive argument with all true premises remains useless if it is based on unprojectible (unentrenched) predicates. Samples are representative only if taken according to accepted practice. These are all instance or rightness, and if Goodman is right, rightness arises from habit or tradition. Analytical hypotheses, like first-level hypotheses about emeralds in a given language, must both fit the evidence and be right. Analytical hypothese must meet Quine's conditions (1)-(4) and pass some further test of rightness. Perhaps the principle of charity is best regarded as a principle of rightness. The argument here shows that the rightness requirement for translation extends all the way down, all the way down to all sentences. What such a test might be is an intriguing question. But however that question is answered, uniqueness and thus determinacy of translation are not to be expected at any level however theoretical or observational. Rightness in translation theory, as rightness in metaphysics, bestows objectivity but not uniqueness. What goes for translation goes for theory generally. We have no firm foundation; we are at sea with only our traditions and habits to carry us along.
M. G. Yoes
Professor of Philosophy, Emeritus
Department of Philosophy
University of Houston
mgyoes [at] comcast.net
[Foot Note 1]
W. V. Quine, From a Logical Point of View (Cambridge: Harvard University Press, l981), Essay 4.
[Foot Note 2]
W. V. Quine, Ways of Paradox (New York: Knopf, l966), p.230f; and W. V. Quine, Word and Object (Cambridge: MIT Press,l960), p. 22f.
[Foot Note 3]
An example might be Putnam in his review of Quine's Quiddities in London Review of Books, 21 April 1988, p.12f.
[Foot Note 4]
W. V. Quine, The Pursuit of Truth (Cambridge: Harvard Univerisity Press, 1990), p. 20: «...when I cite predictions as the checkpoints of science, I do not see it as normative. I see it as defining a particular language game, in Wittgenstein's phrase: the game of science, in contrast to other good language games such as fiction and poetry.»
[Foot Note 5]
Word and Object, p.24.
[Foot Note 6]
In Word and Object (p.44) Quine confesses that the observation sentence does seem to be the Protokalsatz, but that it differs in that observationality is a matter of degree. And of course the particular account of language he gives even under the natural standpoint is contextualist and holistic notwithstanding the observation sentence.
[Foot Note 7]
Perhaps this is evident in Quine's reply to Chomsky in Words and Objections (p. 303) and in his review of Nelson Goodman's Ways of Worldmaking in the New York Review of Books, November 23, 1978.
[Foot Note 8]
See Theories and Things (Harvard University Press, 1981), pp. 97f.
[Foot Note 9]
W. V. Quine, Theories and Things (Harvard, 1981), p. 98.
[Foot Note 10]
ibid., p. 98.
[Foot Note 11]
Of course Goodman's definition is formulated in terms of things examined before t and things not so examined. He is after confirmation. Here, since the subject is stimulus meaning, being prompted by a stimulus replaces examinining an object for color.
[Foot Note 12]
Journal of Philosophy (1973).