THE NEED FOR CONSTRAINTS IS QUITE REAL
Strangely, this idea - it is hard to refute so much that goes on in science - is precisely the claim presented by Poincare and Duhem. The spectroscopic empirical support of quantum theory which is made available by the success of the calculation (the data are much easier to obtain than the calculation) is invalidated by the fact that there is no empirical undermining made available by the failure of the calculation.
Hence there is no corroboration in science; its theory is accepted a priori, not empirically; as a mathematical framework, not as the truth or putative truth about Nature. It is in the frontiers of science where there are so many continuing auxiliary hypotheses, both empirical and computational, that Duhem’s thesis begins to look real.Yet we are still uneasy; we may be uneasy because we want science to fix our beliefs, because we have inductivist predilections, of which Duhem may wish to chastise us for our own good. But we may be uneasy, feeling that there is more liberty prescribed by Poincare and by Duhem then a scientist notices in his daily researches. We may feel not the need for more empirical support, but for more stringent rules of the game. (This, of course, will make us more sympathetic with Popper.)
Even this need, the need for more constraints, is not new; it has been felt already, and by Duhem of all people. He did so when responding to Poincare’s famous claim that it will always be more advisable to reform our auxiliary hypotheses in order to take account of new experiences rather than to reform our central hypotheses, the hard core of our science, such as Euclidean geometry of Newtonian dynamics. In his criticism of Poincare, he said this. Tinkering with all the working hypotheses and auxiliary hypotheses, necessary for the saving of our central axioms, is usually advisable; but a moment may arrive when the tinkering becomes so cumbersome that we may prefer to upset our fundamental axioms which we had guarded against refutation for so long.
Even Euclidean geometry and Newtonian mechanics may have to go. And yet, Duhem was bitterly opposed to Einstein until his dying day. He had the intuition that all the minor alterations would pave the way to the major alteration so as to prevent any revolution - that the major alteration is decomposable to myriads of minor alterations each made independently until the last one of these would effect the revolution in a coup de grace. This, of course, is Duhem’s continuity theory. It is neither historically true, nor intellectually clear enough. Already Michael Polanyi has suggested in a manner which seems to conflict with that of Duhem, that there are changes in the history of science of varying magnitudes. Polanyi also claimed that the leadership of the scientific community decress these changes; I do not think Duhem would have endorsed this view. Thomas S. Kuhn has repeated Polanyi’s ideas in a more popular presentation. He suggested that there is both continuity and revolution, in a presentation which is much more detailed then Duhem’s; and he also expressed Polanyi’s authoritarian view which ascribes to the leadership the right and duty to declare a revolution. Kuhn’s system, then, is at least overdetermined since it accepts Polanyi’s major points and much of Duhem’s continuity theory. At times I think Kuhn endorses all of Duhem’s theory, considering a major revolution to be just a minor step after many accumulated small changes; at times when I read Kuhn about leading scientists’ sleepless nights I see him more of a Polanyiate and less of a Duhemian. Polanyi himself says that a scientist is constrained by his feelings, by his ineffable personal knowledge; and perhaps Kuhn shares this irrational theory. It is hard to say. It much depends on a detailed exegesis of Kuhn’s view on the mechanism of normal scientific research: thus far he has said less about it then one might think after a cursory examination of his work. And his later expositions are often explicit changes from his earlier ones, often implicit ones, and his view seems now more clouded than when he first explained it. I wish to state, then, merely as a general impressionistic view, that reading Duhem, Polanyi, or Kuhn, one has the impression that science is much more constrained than they describe it to be: they do claim that it is rather constrained, but they do not say how and they seem to mistify the reader. Indeed, it seems that Polanyi explicitly enough criticizes his opponents (chiefly the inductivists and the Popperians) on these lines: I agree with you that science needs constraints, and I understand that your ideas of verifiability and/or of refutability in science is constraining enough; yet there is so much in science which is unverifiable and irrefutable. It is still subject to constraint, but not to yours; rather it is subject to the scientist’s own good sense and his own feeling of when he gives himself too much leeway.The only writer who has explicitly argued, however, against conventionalism and against Polanyi, charging them with being much too lax, is Adolf Grunbaum. What Grunbaum has said, in a detailed example, is that it is not as easy to blame a working hypothesis or an auxiliary hypothesis for a refuted prediction - that ad hoc amendation of a refuted system, though in (logical) principle is admittedly always possible, is in practice much less easy to achieve than one may think when one accepts Duhem’s argument. Nor does Grunbaum agree with Polanyi’s alternative. When a refutation occurs, says Polanyi, a scientist may ignore it and take the responsibility without knowing why; we may trust a scientist to find his way sooner or later, even if he does not quite know as yet why he is so unperturbed by a given seeming refutation, says Polanyi. Polanyi even offers a historical example or two - but I will not explain here why I consider them invalid. Enough it is to note that, as Grunbaum notices, this theory is dangerous as both authoritarian and irrationalistic but that all this does not invalidate it.
To return to Duhem’s argument, Grunbaum’s observation that it is not easy to apply it is correct, but it may be not damaging in the least, nor does it tell us enough about the technical constraints which science does undertake. For, saying that any move in the advancement of science requires imagination and is therefore difficult, does not refute Duhem; indeed it is much in accord with Duhem, perhaps it even accords with Kuhn’s view (it is too vague for me to decide this matter). But I do not wish to register any disagreement with Grunbaum, of whose views I do not claim expert knowledge (he is notoriously taxing on his reader). No doubt, what I am going to say he has at least in part anticipated, in part he may wish to differ from. But I cannot say.
V.