This essay was written for a collection of papers about rationality and relativism edited by Martin Hollis and Steven Lukes. Both were advocates of a sensible and sensitive rationalism and disliked the increasingly relativist tendencies of the time (1980). Edinburgh had recently become the most threatening well-argued English-language power base for relativism in epistemology. The Edinburgh School, led by Barry Barnes and David Bloor, had a "strong programme in the sociology of knowledge" that cheerfully avowed that it was relativist. Lukes and Hollis arranged their contributors from left to right, the most relativist at the beginning of the book, and the most rationalist at the end. A new paper by Bloor and Barnes came first. I was perhaps disingenuous to be surprised when I found that my own paper was placed as the second most relativist contribution.
This is the first piece in which I took up the idea of a "style of reasoning" that I first encountered in 1978, in Pisa, listening to a paper by the senior historian of science, Alistair Crombie. He himself did not bring out his gigantic three-volume study of "styles" until 1994, but I was able to read a good deal of it many years before. Chapter 12 is a more systematic development of these ideas.
I start from the fact that there have been different styles of scientific reasoning. The wisest of the Greeks admired Euclidean thought. The best minds of the seventeenth century held that the experimental method put knowledge on a new footing. At least part of every modern social science deploys some statistics. Such examples bring to mind different styles of reasoning with different domains. Each has surfaced and attained maturity in its own time, in its own way.
An inane subjectivism may say that whether p is a reason for q depends on whether people have got around to reasoning that way or not. I have the subtler worry that whether or not a proposition is as it were up for grabs, as a candidate for being true-or-false, depends on whether we have ways to reason about it. The style of thinking that befits the sentence helps fix its sense and determines the way in which it has a positive direction pointing to truth or to falsehood. If we continue in this vein, we may come to fear that the rationality of a style of reasoning is all too built-in. The propositions on which the reasoning bears mean what they do just because that way of reasoning can assign them a truth value. Is reason, in short, all too self-authenticating?
My worry is about truth-or-falsehood. Consider Hamlet's maxim, that nothing's either good or bad but thinking makes it so. If we transfer this to truth and falsehood, it is ambiguous between (a) nothing which is true is true, and nothing which is false is false, but thinking makes it so, and (b) nothing's either true-or-false but thinking makes it so. It is (b) that preoccupies me. My relativist worry is, to repeat, that the sense of a proposition p, the way in which it points to truth or falsehood, hinges on the style of reasoning appropriate to p. Hence we cannot criticize that style of reasoning as a way of getting to p or to not-p, because p simply is that proposition whose truth value is determined in this way.
The distinction between (a) and (b) furnishes a distinction between subjectivity and relativity. Let (a) be subjectivism: by thinking, we might make something true or make it false. Let (b) be the kind of relativity that I address in this paper: by thinking, new candidates for truth and falsehood may be brought into being. Many of the recent but already classical philosophical discussions of such topics as incommensurability, indeterminacy of translation, and conceptual schemes seem to me to discuss truth where they ought to be considering truth-or-falsehood. Hence bystanders, hoping to learn from philosophers, have tended to discuss subjectivity rather than relativity. For my part, I have no doubt that our discoveries are "objective," simply because the styles of reasoning that we employ determine what counts as objectivity. My worry is that the very candidates for truth or falsehood have no existence independent of the styles of reasoning that settle what it is to be true or false in their domain.
It is not the case that nothing's either true or false but thinking makes it so. Plenty of things that we say need no reasons. That is the core of the discredited philosophical doctrine of observation sentences, the boring utterances that crop up in almost any language, and which make radical translation relatively easy. Translation is hard when one gets to whole new ranges of possibility that make no sense for the favored styles of reasoning of an-other culture. It is there that ethnographers begin to have problems. Every people has generated its own peculiar styles. We are no different from others, except that we can see more clearly, from our own written record, the historical emergence of new styles of reasoning.
I take the word "style" from A. C. Crombie's title, Styles of Scientific Thinking in the European Tradition (1994). He concluded an anticipatory paper with the words:
The active promotion and diversification of the scientific methods of late medieval and early modern Europe reflected the general growth of a re-search mentality in European society, a mentality conditioned and increasingly committed by its circumstances to expect and to look actively for problems to formulate and solve, rather than for an accepted consensus without argument. The varieties of scientific method so brought in to play may be distinguished as,
The first three of these methods concern essentially the science of individual regularities, and the second three the science of the regularities of populations ordered in space and time. (Crombie 1981, 284.)
Coincidentally, at the same conference to which Crombie read these words, Winifred Wisan (1981) presented a paper on "the emergence of a new scientific style." Both Crombie's and Wisan's papers were about Galileo, who has long been a favorite candidate for advancing a new style of thought. Sometimes words more dramatic than "style" are used, as when Althusser (1972, 185) wrote of Thales opening up a new continent, that of mathematics, Galileo opening up the continent of dynamics, and Marx that of history. But often the word "style" is chosen. It is to be found in Collingwood. Stephen Weinberg, the theoretical physicist, recalled Husserl speaking of a Galilean style for "making abstract models of the universe to which at least the physicists give a higher degree of reality than they accord the ordinary world of sensation" (Weinberg 1976, 28). Weinberg found it remarkable that this style should work, "for the universe does not seem to have been prepared with human beings in mind." The linguist Noam Chomsky picked up this remark, urging that "we have no present alter-native to pursuing the `Galilean style' in the natural sciences at least" (Chomsky 1980, 9).
Like T. S. Kuhn's "paradigm", the word "style" serves my four contemporary authors to point to something general in the history of knowledge. There are new modes of reasoning that have specific beginnings and trajectories of development. Even these four thinkers would surely not agree in carving up histories into styles. The historian will find many styles where Chomsky sees only one. Doubtless the very word "style" is suspect. It is cribbed from art critics and historians, who have not evolved a uniform connotation for the word. Nor would all their remarks about style tidily transfer to modes of reasoning. That is a problem that Wisan's paper begins to address. The success of the word "style," as an analytic term for the history of science, may depend on the reception of Crombie's immensely learned historical analysis. Use of a borrowed word needs detailed examples to flesh it out. Despite these reservations, I shall take the fact that these recent writers employ the word in similar ways as an excuse for not attempting my own exegesis here.
The existence of styles of reasoning does not immediately suggest relativism. Before elaborating the relativist worry sketched at the beginning of this chapter, I shall first state a rationalist position informed by a proper respect both for history and for the idiosyncrasies of ourselves and others. I shall call it arch-rationalism. (1, too, am an arch-rationalist most of the time.)
The arch-rationalist believes what right-thinking people have known all along. There are good and bad reasons. It has taken millennia to evolve systems of reasoning. By and large our Western tradition has contributed more to this progress than any other. We have often been narrow, blinkered, and insensitive to foreign insights. We have repressed our own deviant and original thinkers, condemning many to irretrievable oblivion. Some of our own once-favored styles of reasoning have turned out to be dead ends and others are probably on the way. However, new styles of reasoning will continue to evolve. So we shall not only find out more about nature, but we shall also learn new ways to reason about it. Maybe Paul Feyerabend's (1975) advocacy of anarchy, or at least dadaism, is right. To compel people to reason in approved ways is to limit us and our potentialities for novelty. Arch-rationalism is convinced that there are good and bad reasons, but since it does not commit us to any specific regimentation like that of formal logic or that of Karl Popper, it is fairly receptive to Feyeraband's imitation anarchy.
My arch-rationalist thinks that there is a fairly sharp distinction between reasons and the propositions they support. Reasons merely help us find out what is the case. The arch-rationalist wants to know how the world is. There are good and bad reasons for propositions about nature. They are not relative to anything. They do not depend on context. The arch-rationalist is not an imperialist about reason. Maybe there could be people who never reason nor deliberate at all. They tell jokes, make and break promises, feign insults and so forth, but they never reason. Just as statistical reasons had no force for the Greeks, so one imagines a people for whom none of our reasons for belief have force. On the other hand, the arch-rationalist is an optimist about human nature. We who value truth and reason do imagine that a truthless and unreasoning people would, if left alone, evolve truth and reason for themselves. They would in their own way acquire a taste for speculation about the diagonal of a square, for motion on the inclined plane, for the tracks of the planets, for the inner constitution of matter, the evolution of the species, the Oedipus complex, and amino acids.
The arch-rationalist not only grants that our kinds of truth and reason may not play as great a role in the life of other peoples as in our own culture; he may also be a romantic, hankering after a simpler, less reason-impregnated life. He will grant that our values are not inevitable, nor perhaps the noblest to which our species can aspire. But he cannot escape his own past. His admission of the historicity of our own styles of reasoning in no way makes it less objective. Styles of reasoning have histories, and some emerged sooner than others. Humankind has got better at reasoning. What ground for relativism could there be in all that?
Instead of challenging the assumptions of the arch-rationalist, I shall extract a hint of incoherence from his heartland, which is, in the end, positivism.
Positivism is commonly taken to be a hard-headed antagonism to all forms of relativism. I shall create a question for the arch-rationalist from three aspects of positivism itself. I draw them from Auguste Comte, Moritz Schlick, and Michael Dummett, that is, the original positivist of the 1840s, the leader of the Vienna Circle in 1930, and the most gifted present exponent of one among that family of doctrines.
Comte. He was a historicist. His epistemology is a massive and almost unreadable account of human knowledge, a narrative of the human mind in which each intellectual innovation finds its own niche. One of his ideas is that a branch of knowledge acquires a "positivity" by the development of a new, positive, style of reasoning associated with it. He is none too clear what he means by "positive"; he sometimes says he chose the word chiefly because it had overtones of moral uplift in all European languages. A positive proposition is one that is by some means befitting the branch of knowledge to which it belongs. We may pun on his word: a positive proposition is one that has a direction, a truth value. It is no distortion to say that for Comte a class of positive propositions is a class of propositions that are up for grabs as true-or-false.
There are many aspects of Comte's thought from which one hastily withdraws—I refer both to questions of ideology and to issues of interest to analytic philosophers of science (his analysis of causation, for example). I draw attention only to the idea of a historical evolution of different styles of reasoning, each bringing in its train its own body of positive knowledge. Each finds its place in great tabular displays of the sciences that serve as pull-outs from his gigantic epistemological text, the Cours de philosophie positive. Comte did not think that the evolution of styles and of positive knowledges had come to an end. His life goal was the creation of a new positive science, sociology. This would require a new style of reasoning. He ill foresaw what this style would be, but his meta-conception of what he was doing was sound.
Schlick. One of the more memorable statements of logical positivism is Moritz Schlick's, "the meaning of a sentence is its method of verification" (1936, 361). Those words could not stand unmodified, because the Vienna Circle had succumbed to Gottlob Frege's dictum that meanings are definite, objective, and fixed. Schlick's maxim would imply that a change or advance in a method of verification would change the meanings of a sentence. Rather than give up the idea of meanings handed down from generation to generation, tranquil and unmodified, logical positivists revised Schlick's maxim again and again, although with no satisfactory outcome. (See Hacking 1975b, ch. 9, for an account of repeated failures.) But for Comte, or any other of those fortunate writers of 1840 not yet infected by Fregean theories of meaning, Schlick's statement would be just fine. It is precisely, for Comte, the methods of verification—the ways in which the positive truth values are to be established—that determine the content of a body of knowledge.
Dummett. In logic, a proposition that has a definite truth value, true or false, is called bivalent. Dummett's work has made philosophers think closely about bivalence (Dummett 1976). It was first inspired by a philosophical reconstruction of some of the thoughts behind intuitionist mathematics. In what is called a nonconstructive proof, one cannot exhibit the mathematical objects that are proved to exist. (So one might have a step in which one asserts that there is a prime number with a certain property, but be unable to say which prime number it is.) Nonconstructive proofs may also assume of a proposition that it is either true or false, without being able to show which truth value it has. Some philosophical mathematicians, including Dummett, have doubted whether such nonconstructive proofs are admissible.
Dummett is attracted to the following basis of his doubt. Whether or not a proposition is bivalent must depend upon its meaning. He wonders how we can confer meanings on statements in nonconstructive mathematics—meanings in virtue of which the statements are bivalent, although there is no known way to settle the truth values. It is we who, through our linguistic practices, are the sole source of the meanings of what we say.
How then can we confer a meaning on a statement, such that it is bivalent, when nothing we know how to do bears on the truth or the falsehood of the statement? Maybe statements of nonconstructive mathematics acquire bivalence only as we perfect means of determining their truth values or exhibiting the mathematical objects of which they speak?
Although this subtle question arose in sharp form in the intuitionist critique of classical mathematics, Dummett extends it to other forms of discourse. Many statements about the past cannot now be settled by any practicable means. Are they bivalent? Might bivalence recede into the past as historical data become irrevocably erased? Dummett does not claim that his worries are conclusive, nor does he expect parallel answers for every kind of discourse. One might, on reflection, come out for bivalence in the case of history, but reject it for nonconstructive mathematics.
Positive and bivalence. I have spoken of being true-or-false, and have used Comte's word "positive." Is this the same idea as bivalence? Not as I shall use the words. Being positive is a less strong characteristic than bivalence. Outside mathematics, I suspect that whether a statement is bivalent or not is an abstraction imposed by logicians to facilitate their analysis of deductive argument forms. It is a noble abstraction, but it is a consequence of art, not nature. In the speculative sciences that concern me in this paper, the interesting sentences are the ones that are up for grabs as true or false—ones for which we believe we have methods that will determine the truth values. The applications of these methods may require as yet unimagined technological innovation. Moreover, as we find out more about the world, we find out that many of our questions no longer make sense. Bivalence is not the right concept for science. Allow me a couple of examples to point to the distinction required.
At the time of Pierre-Simon de Laplace it was very sensible to think that there are particles of caloric, the substance of heat, that have repulsive forces that decay rapidly with distance. Relying on this hypothesis, Laplace solved many of the outstanding problems about sound. Propositions about the rate of extinction of the repulsive force of caloric were up for grabs as true or false, and one knew how to obtain information bearing on the question. Laplace had an excellent estimate of the rate of extinction of the repulsive force, yet it turns out that the whole idea is wrongheaded. I would say that Laplace's sentences once were "positive." They were never bivalent. Conversely, James Clerk Maxwell once said that some propositions about the relative velocity of light were intrinsically incapable of de-termination, yet a few years after he said that A. A. Michelson had invented the technology to give precise answers to Maxwell's questions. I would say that the sentences of interest to Maxwell had positivity when he uttered them, but were bivalent only after a transformation in technology—a transformation whose success depends on delicate experimental details about how the world works.
In short, Comte's "positive" is drawing attention to a less demanding concept than Dummett's "bivalent" Yet the two are connected, and so are the thoughts of both writers. Dummett says: not bivalent unless we have a proof of the truth value, or a known sure-fire method for generating the proof. Comte says: not positive, not in the running for being true-or-false, until there is some style of reasoning that will bear on the question.
Comte, Schlick, and Dummett are no more relativist than Crombie or Chomsky. Yet a positivist train of thought, combined with an emphasis on styles of reasoning, has the germ of relativism. If positivity is consequent upon a style of reasoning, then a range of possibilities depends upon that style. They would not be possibilities, candidates for truth or falsehood, unless that style were in existence. The existence of the style arises from historical events. Hence, although whichever propositions are true may depend on the data, the fact that they are candidates for being true is a consequence of a historical event. Conversely, the rationality of a style of reasoning as a way of bearing on the truth of a class of propositions does not seem open for independent criticism, because the very sense of what can be established by that style depends upon the style itself.
Is that a nasty circle?
I shall proceed as follows. First, I observe that by reasoning I don't mean logic. I mean the very opposite, for logic is the preservation of truth, while a style of reasoning is what brings in the possibility of truth or falsehood. Then I separate my idea of a style of reasoning from the incommensurability of Kuhn and Feyerabend, and from the indeterminacy of translation urged by Quine. Then I examine Davidson's fundamental objection to the supposition that there are alternative ways of thinking. He may refute subjectivity, as I understand it, but not relativity. The key distinction through-out the following discussion is the difference between truth-and-falsehood as opposed to truth. A second important idea is the looseness of fit between those propositions that have a sense for almost all human beings regardless of reasoning, and those that get a sense only within a style of reasoning.
Neither deductive logic nor induction occur on Crombie's list. How strange, for are they not said to be the basis of science? It is instructive that no list like Crombie's would include them. The absence reminds us that styles of reasoning create the possibility for truth and falsehood. Deduction and induction merely preserve it.
We now understand deduction as that mode of inference that preserves truth. It cannot pass from true premises to a false conclusion. The nature of induction is more controversial. The word has been used in many ways. There is an important tradition represented alike by the philosopher C. S. Peirce and the statistician Jerzy Neyman: induction is that mode of argument that preserves truth most of the time. (Hacking 1980a shows how Neyman's theory of testing hypothesis connects with Peirce's theory of probable inference.)
Deduction and induction were important human discoveries. But they play little role in the scientific method, no more than the once revered syllogism. They are devices for jumping from truth to truth or from truth to probable truth. Not only will they give us no original contingent truth from which to jump, but also they take for granted the class of sentences that assert possibilities of truth or falsehood. That is why they do not occur in Crombie's list. In deduction and induction alike, truth plays the purely formal role of a counter on an abacus. It matters not what truth is, when we employ the mechanics of the model theory of modern logicians. Their machine works well so long as we suppose that the class of sentences that have truth values is already given. (Or, in the case of intuitionist logic, one supposes that the class of sentences that may, through proof, acquire truth values is already given.) Induction equally assumes that the class of possible truths is predetermined. Styles of reasoning of the sort described by Crombie do something different. When they come into being they generate new classes of possibilities.
Philosophers have recently given us two doctrines that pull in opposite directions. Both seem to use the idea of a conceptual scheme, a notion that goes back at least to Kant, but whose modern nominalist version is due to W. V. Quine. He says that a conceptual scheme is a set of sentences held to be true. He uses the metaphor of core and periphery. Sentences at the core have a kind of permanence and are seldom relinquished, while those on the periphery are more empirical and more readily given up in the light of "recalcitrant experience."
My talk of styles of reasoning does not mesh well with Quine's idea of a conceptual scheme (Quine 1960, ch. 2). In his opinion, two schemes differ when some substantial number of core sentences of one scheme are not held to be true in another scheme. A style of reasoning, in contrast, is concerned with truth-or-falsehood. Two parties, agreeing to the same styles of reasoning, may well totally disagree on the upshot, one party holding for true what the other party rejects. Styles of reasoning may determine possible truth values, but, unlike Quine's schemes, are not characterized by assignments of truth values. It is to be expected, then, that Quine's application of the idea of a conceptual scheme will not coincide with my idea of styles of reasoning.
Quine's most memorable thesis is the indeterminacy of translation. Let L and M be languages spoken by two truly disparate communities. Quine holds that there are indefinitely many possible but incompatible translations between L and M. No matter how much speakers of L and M might converse, there is in principle no way of settling on a definitely right translation. This is not a matter of settling on nuances; Quine means that you could take a sentence s of L and translate it by one system of translation into p of M, and translate it by another system into q of M, and p and q would, in M, be held to be incompatible.
As we shall see in the next section, Donald Davidson has noticed that the notion of conceptual scheme does not ride well with the indeterminacy of translation. For how are we to say that speakers of L have a scheme different from we who speak M? We must first pick out the true sentences from the core of the scheme of L, and show that many of these translate into sentences of M that we who speak M hold to be false. But what is to assure that this is the right translation? When translating, there is a strong instinct to render central doctrines of L as main truths of M. Once you focus on truth rather than truth-or-falsehood, you begin a chain of considerations that call in question the very idea of a conceptual scheme.
The thesis of indeterminacy of translation pulls in one direction and the idea of incommensurability pulls in another. We owe incommensurability to Kuhn and Feyerabend. For one slightly unusual version of this famous notion, see Feyerabend (1978, 65-70). The idea is that disparate systems of thought are not mutually expressible. Kuhn has tended to make the idea fit commonplace situations, while Feyerabend emphasizes the extreme. Thus Feyerabend's favorite example of incommensurability is the break between the cosmologies of archaic and classical Greece. Kuhn, in contrast, comes back to the idea of "no common measure" in the original meaning of the word, and applies it to more everyday "advances" in knowledge. When there has been a scientific revolution, the new science may address new problems and employ new concepts. There is no way of settling whether the new science does its job better than the old one, because they do different jobs. Kuhn finds this sort of incommensurability in all sorts of revolutions that strike the outsider as minor, while Feyerabend focuses on big shifts in human thought. Both writers had at one time suggested that incommensurability should be understood in terms of schemes and translation. Incommensurability meant that there would simply be no way of translating from one scheme to another. Thus this idea pulls in a direction exactly opposite to Quine's. Indeterminacy says there are too many translations between schemes, while incommensurability says there are none at all.
Would either the Kuhnian or the Feyerabendian idea of incommensurability apply if styles of reasoning were to supersede each other? The Kuhnian "no common measure" does not apply in any straightforward way, be-cause when we reason differently, there is no expectation of common measure of the sort that successive Kuhnian paradigms invite. Hence it is to the more extreme, Feyerabendian, use of the term that we must look. That is surely the popular conception of incommensurability: the inability of one body of thought to understand another.
I do admit that there is a real phenomenon of disparate ways of thinking. Some styles of reasoning have been so firmly displaced that we cannot even recognize their objects. The renaissance medical, alchemical, and astrological doctrines of resemblance and similitude are well-nigh incomprehensible. One does not find our modern notions of evidence deployed in those arcane pursuits. There is very little truth in all that hermetic writing, and to understand it one cannot search out the core of truth that meshes with our beliefs. Yet that stuff may not be best described as incommensurable with our modern chemistry, medicine, and astronomy. It is not that the propositions match ill with our modern sciences, so much as that the way propositions are proposed and defended is entirely alien to us.
You can perfectly well learn hermetic lore, and when you do so, you end up talking the language of Paracelsus, possibly in translation. What you learn is not systems of translation, but chains of reasoning which would have little sense if one were not re-creating the thought of one of those magi. What we have to learn is not what they took for true, but what they took for true-or-false. (For example, that mercury salve might be good for syphilis because mercury is signed by the planet Mercury, which signs the marketplace where syphilis is contracted.)
Understanding the sufficiently strange is a matter of recognizing new possibilities for truth-or-falsehood, and of learning how to conduct other styles of reasoning that bear on those new possibilities. The achievement of understanding is not exactly a difficulty of translation, although foreign styles will make translation difficult. It is certainly not a matter of designing translations which preserve as much truth as possible, because what is true-or-false in one way of talking may not make much sense in another until one has learned how to reason in a new way. One kind of understanding is learning how to reason. When we encounter old or alien texts we have to translate them, but it is wrong to focus on that aspect of translation that merely produces sentences of English for sentences of the other language. With such a limited focus, one thinks of charitably trying to get the old text to say as much truth as possible. But, even after Paracelsus is translated into modern German, one still has to learn how he reasoned in order to understand him. Since the idea of incommensurability has been so closely tied to translation rather than reasoning, I do not use it here.
The indeterminacy of translation is an equally wrong idea. It is empirically empty, because we know that unequivocal translation evolves between any two communities in contact. As observed in the previous chapter, anecdotal counterexamples to this assertion do not stand up to scrutiny. Indeterminacy is the wrong theoretical notion, because it starts from an idea of truth-preserving matching of sentences. In fact, the possibilities available in one language are not there in the other. To get them into the second language one has to learn a way of reasoning and, when that has been done, there is no problem of translation at all, let alone indeterminacy.
There is perfect commensurability, and no indeterminacy of translation, in those boring domains of "observations" that we share with all people as people. Where we as people have branched off from others as a people, we find new interests, and a looseness of fit between their and our common-places. Translation of truths is irrelevant. Communication of ways to think is what matters.
In his famous paper "On the Very Idea of a Conceptual Scheme," Donald Davidson (1974) argues more against incommensurability than indeterminacy, but he is chiefly against the idea of a conceptual scheme that gives sense to either. He provides "an underlying methodology of interpretation" such that "we could not be in a position" to judge "that others had concepts or beliefs different from our own." He makes plain that he does not reach this result by postulating "a neutral ground, or a common coordinate system" between schemes. It is the notion of a scheme itself to which he is opposed. He rejects a "dogma of dualism between scheme and reality" from which we derive the bogey of "conceptual relativity, and of truth relative to a scheme."
Davidson distinguishes two claims. Total translatability between schemes may be impossible, or there may be only partial untranslatability. Even if we do not follow the intricacies of his argument, nor even accept its premises, we can, like Davidson, dismiss the idea of total untranslatability. As a matter of brute fact, all human languages are fairly easily partially translatable. The fact is closely connected with what I said earlier, that there is a common human core of verbal performances connected with what people tend to notice around them. But I said that there is a looseness of fit between that broad base of shared humanity and the interesting things that people like to talk about. That looseness leaves some space for incommensurability. It is not only the topics of discussion that may vary from group to group, but what counts as a point of saying something. Yet Davidson counters there too, and mounts a magnificent attack against even the notion of partial untranslatability between groups of people. Since in fact even partial untranslatability is chiefly a matter of coming to share the interest of another, and since lots of travelers are pretty sympathetic people, interests do get shared, so we should welcome an argument against partial untranslatability too. Yet since Davidson's argument may seem founded upon a lack of concern for alternative interests, we may fear his premises while we accept his conclusions. My diagnosis is that, like
Quine, he assumes that a conceptual scheme is defined in terms of what counts as true, rather than of what counts as true-or-false.
Davidson concluded his argument against relativity with the words, "Of course the truth of sentences remains relative to a language, but that is as objective as can be." Earlier, he rightly stated what is wrong with the idea of making a sentence true:
Nothing makes sentences and theories true: not experience, not surface irritations [he there alluded to Quine], not the world ... That experience takes a certain course, that our skin is warmed or punctured ... these facts, if we like to talk that way, make sentences and theories true. But this point is better made without mention of facts. The sentence "my skin is warm" is true if and only if my skin is warm. Hence there is no reference to a fact, a world, an experience, or a piece of evidence. (Davidson 1974, 16)
Davidson's example, "my skin is warm," serves me well. I urge a distinction between statements that may be made in any language, and which require no style of reasoning, and statements whose sense depends upon a style of reasoning. Davidson writes as if all sentences were of the former class. I agree that "my skin is warm" is of that class. When I once looked for the best example of a sense-datum sentence to be actually published in the annals of real science, I hit upon precisely this sentence, or rather, "my skin is warmed." It begins Sir William Herschel's investigations of 1800, which are said to commence the theory of radiant heat. He noticed that by using filters of some colors his skin was warmed, while in using other colors he had much light but little heat (Hacking 1983a, 171).
Herschel went on to pose a theory of invisible rays of heat, a theory that we now call correct, although his own experiments made him give it up. In the course of this reasoning he abandoned the following sentence: "The heat which has the refrangibility of the red rays is occasioned by the light of those rays." We can certainly write out a truth condition of the form "s is true if and only if p" for this sentence. But there arises a problem for the sufficiently foreign translator. It is not that words like "ray" and "refrangible" are mildly theoretical and the translator may have no such notions in his vocabulary. If another culture has acquired the styles of reasoning enumerated by Crombie, it can perfectly well learn Herschel's physics from the ground up—that is just what I do in making sense of Herschel's text. The problem is that the sufficiently foreign person will not have Herschel's kind of sentence as the sort of thing that can be true-or-false, because the ways of reasoning that bear on it are unknown. To exaggerate the case, say the translator is Archimedes. I do not choose him at random, for he wrote a great tract on burning mirrors and was a greater scientist than Herschel. Yet I say he would not be able to effect a translation until he had caught up on some scientific method.
I should repeat my opposition to usual versions of incommensurability. It is not that Herschel's science had some Newtonian principles about rays and refrangibility that determined the meaning of sentences in which those words occur, and so those sentences could not have the same meaning in another theory. On the contrary, Herschel's sentences were fairly immune to change in theory. They were up for grabs as true or false in 1800; Herschel thought first that a crucial sentence is true and later held it to be false; many years later the world agreed on the truth of the sentence. Herschel, then, first grabbed the right end of the stick and then grabbed the wrong one. My claim about a translator less well placed than Archimedes is that until he learns how to reason more like Herschel, there are no ends of a stick to grab.
"Truth of sentences," writes Davidson, "remains relative to a language, but that is as objective as can be." I claim that for part of our language, and perhaps as part of any language, being true-or-false is a property of sentences only because we reason about those sentences in certain ways. Subjectivists put their worries in the form of saying that with different customs we could "rightly" take some propositions for true while at present we take them for false. Davidson has dealt sharply with all such formulations. But he has left a space for a relativist fear. The relativist ought to say that there might be whole other categories of truth-or-falsehood than ours.
Perhaps I am proposing a version of the conceptual scheme idea. Quine's conceptual schemes are sets of sentences held for true. Mine would be sets of sentences that are candidates for truth or falsehood. Does such a notion fall into the "dogma of scheme and reality" that Davidson resents? I do not think so. The idea of a style of reasoning is as internal to what we think and say as the Davidsonian form, "s is true if and only if p," is internal to a language. A style is not a scheme that confronts reality. I did speak earlier of styles of reasoning being applied to data and to the formation of data. But data are uttered and are subject to Davidsonian treatment. There is much to be said about the neglected field of study of experimental science, but it has nothing much to do with scheme/reality. My own work on the subject (Hacking 1983a) tries to show how experiment has a life of its own, unrelated to theories or schemes.
This chapter makes two assertions and draws some inferences from them. Each assertion and every inference is in need of clarification. To list them is to show how much more must be done.
This chain of reflections does not lead to subjectivity. It does not imply that some proposition, with a content independent of reasoning, could be held to be true, or to be false, according to the mode of reasoning we adopt. Yet this defeat of subjectivity seems hollow, because the propositions that are objectively found to be true are determined as true by styles of reasoning for which in principle there can be no external justification. A justification would be an independent way of showing that the style gets at the truth, but there is no characterization of the truth over and above what is reached by the styles of reason itself.
Can there not be a meta-reason justifying a style of reason? Can one not, for example, appeal to success? It need not be success in generating technology, although that does matter. Nor is it to be success in getting at the truth, for that would be circular. There can, however, be noncircular successes in truth-related matters. For example, following Imre Lakatos (1978, chs. 1, 2), one might revamp Popper's method of conjecture and refutation, urging that a methodology of research programs constantly opens up new things to think about. I have quoted Chomsky giving a similar meta-reason. On his analysis of the Galilean style, it has not only worked remarkably well, but also, in the natural sciences at least, we have no alternative but to go on using that style, although, of course, in the future it may not work. Although Chomsky does not make the distinction, his meta-reason is less that Galileo's style continues to find out the truth about the universe than that it poses new kinds of probing and answering. It has produced an open-ended dialogue. That might terminate in the face of a nature that ceased to participate in ways that the Galilean can make sense of. We know it might cease to cater to our interests, but at present (says Chomsky) we have no alternative.
Chomsky is saying that if we want to engage in certain pursuits (call them the natural sciences or even the pursuit of truth in our tradition), we must reason with our reasons. Other styles of reasoning may occur; some are current. Other people may have other interests. We ought at least to be cautious, in the social sciences, in looking for other styles of reasoning. Such considerations may lead the arch-rationalist to be a stick-in-the-mud, but since relativity does not imply subjectivity, he can carry on doing what we do with few qualms.
Some arch-rationalists may even find themselves agreeing that an anarcho -rationalism I have learned from Feyerabend is appealing. Our overall interests in truth and reason may well be served by letting other styles of reason evolve in their own ways, unfettered by a more imperial kind of rationalism. But that does not mean to say that I, as an anarcho-rationalist, will take up something so recently killed off in our own tradition as homoeopathic medicine and its appeal to similitudes. That is for others (though if they look healthier than me, I might join up). Anarcho-rationalism is tolerance for other people combined with the discipline of one's own standards of truth and reason. The anarcho -rationalist is at home with the sentiment expressed by Sartre (1980, March 10, 93) in his last interview:
C'est ça ma tradition, je n'en ai pas d'autre. Ni la tradition orientale, ni la tradition juive. Elles me manquent par mon historicite.