METAPHYSICS AND ‘AD HOC’ HYPOTHESES
In my view judging Ptolemy from a geocentric viewpoint and judging him from a heliocentric viewpoint we obviously come to quite different conclusions. And, no doubt, Copernicus’ harshness to Ptolemy is in part at least rooted in the latter’s heliocentrism.
Now this is quite alright, except that it is question-begging in the sense that a geocentrist need not be overly impressed with it. That is to say, everyone was always unhappy about the epicycles and/or eccentricities. The question was, what else could be done? And the geocentrists, we know, did all they thought they could within geocentrism. And so within their framework they could not find themselves half as culpable as Copernicus wanted to present them; and why should they have endorsed his heliocentrism in any kind of self-incrimination ?The answer is, to my surprise, at least, seemingly Feyerabendian: it was from the viewpoint of the new theory that the trouble they were working on turned out to be an unsurmountable scandal. Hence there is no difference between the Ptolemaic and the Newtonian efforts to array theory and fact so as to eliminate all contradiction and all ad hoc theory.
I do not wish to ascribe this Feyerabendian view to Feyerabend, both because he never expressed it and because it is not satisfactory. For, whereas the Newtonians were working in part on ad hoc hypotheses, in part on non-ad hoc elaborate calculations, the Ptolemains were mixing the ad hoc and the elaborate calculations. The initial theory, the geocentric theory, was thus doubly defended against facts, both by ad hoc hypotheses and by promises of better calculations; there was no chance of ever collecting such a doubly protected promise, and so Copernicus had the right to lose his patience.
More congenially, we can say, then, what Copernicus offered is not so much a complaint from the heliocentric viewpoint, as a complaint from a neutral viewpoint which may make it more desirable even for a geocentrist to look for an alternative, and then, by accident, the man who threw stones at your glass windows just happens to stand round the corner selling glass.
You may resent him and refuse his services; this will not, however, restore your glass.With this I can now try and reexamine Feyerabend’s story of the secular motions of Mercury. Contrary to Feyerabend I will say this. Admittedly, astronomers and mathematicians existed who tried to prove consistency between Newton and Mercury with no ad hoc hypotheses. Admittedly they went a long way towards achieving that goal and admittedly they had all reasonable hope to reduce the difference between observed and calculated paths to within the limits of observation error. But once Einstein came in, then without admitting the truth of his theory one could see the hope to be very questionable! Not just because his theory agreed with facts within the limits of observation without further ado and thus was superior. No. But the logical fact that certain differences conform to one mathematical regularity made it desperately hopeless that it should also conform to another, so different.
The similarity between Newton and Einstein, between any old and new theory, is nowadays quite commonplace: the old is, of course, an approximation and a special case of the new. But the difference between old and facts looks most different from the old viewpoint than from the new! This is what makes us abandon the old.
Am I giving the same facts as Feyerabend in a different wrapping or am I giving a new theory? I do not know but let me show the important difference between our views if I can.
In Feyerabend’s case the secular motion of Mercury was problematic for Newton but not for Einstein. In Copernicus’ case epicycles were problematic for both the geocentrists and the heliocentrists. In both cases, I feel, the very introduction of a new theory altered the situation from a neutral viewpoint, whereas according to Feyerabend the new Einsteinian theory offered a new viewpoint from which to judge the old situation. To clinch matters, let us try a case where the problem for the old view radically differs from the problem for the new view.
My choice is Ampere’s hypothesis of minute currents which gives matter its magnetic properties. The old theory had the problem, why only the few magnetic materials showed magnetism? When diamagnetism was discovered, the question became more pressing. The problem got most pressing when Faraday discovered that the opposite of diamagnetism is paramagnetism, to be distinguished from the very strong ordinary magnetism which is known as ferro-magnetism. If molecular currents are universal, why should there be such a division? With the Rutherford-Bohr theory of the electron in the atom’s shell, it became clear that all matter should be of the same kind of magnetism. The Bohr magneton theory then, indeed, made all matter diamagnetic. Ampere’s hypothesis designed to explain ferro-magnetism now turned out to explain diamagnetism alone. As to paramagnetism, it remained to be later explained in part by the spin of the electron.
Nobody, I think, protested. The new viewpoint of quantum theory reversed the direction of Ampere’s currents, shifted the problematic from diamagnetism to paramagnetism, and everyone accepted the switch as a matter of course. The reason, of course, is that the case is highly localized and is treated within the context as we find the context with no intention of turning the context to fit our local problem.
Is this fair? Is this proper? First let me say it accords very well with the Poincare-Duhem approach: the general situation prescribes our attitude much more than a mere single hypothesis, much less a single fact. And so, here, perhaps, we can look for the strongest support for the view of Duhem et al. that therefore a fact cannot by itself overthrow a hypothesis.
Let me also stress, so as to avoid confusion, that Duhem would certainly not endorse the Cassirer-Kuhn-Lakatos view, since he both declared facts as imposing themselves on us, and logic to be always similarly or more obligatory. What Duhem said was, once theory conflicts with facts we must modify some of our ideas or others, but it matters little which, and the smaller the modification usually the better.
Duhem’s followers, however, sometimes say this: since we can make this that or the other modification, the very acceptance of a fact is not the same as the overthrow of the theory it conflicts with. At least Cassirer said so outright and perhaps also Lakatos.Duhem would not agree. Conflict he saw as logical, and thus as immutable. But the theory conflicting with the fact he saw as a huge apparatus modifiable in innumerably many ways.
Can we see here, then, a conflict which nobody bothered about? Are there such conflicts? Polanyi offered such examples, but his examples are ones which men of science never admitted, such as Miller’s 1926 repetition of Michelson’s experiment with different results (which are deemed erroneous by the world).
But there is little difficulty in illustrating many facts which clashed with theory and yet were left to rest for a while: the theories were not rejected, the facts were not rejected, there were not even good ad hoc hypotheses to reconcile them, and fields were declared to be “in a mess”. Is this what Cassirer, or perhaps Kuhn or Lakatos, could have in mind? I do not think so. On the contrary, Cassirer says, in the name of simplicity we usually, though not always, do oust a refuted hypothesis. And Kuhn says, after the ad hoc modification becomes too much, we make a revolution: he is not in favor of too much mess either. What Lakatos would have said I do not know. He died too young.
I do not know then what is meant by the claim that we do not always reject a theory on the basis of conflicting evidence. I observe that this view is now gaining much popularity in the field of the philosophy of science. I do feel that the field is now in a mess.
V.