REDUCING THE CONVENTIONS
To this question Paul Feyerabend has a marvelous response: whatever answer you give this question will itself be a theory - a theory of science, that is - which is itself subject to the same questioning.
This looks like infinite regress, but Feyerabend has a surprise in store: rather than a metatheory leading to infinite regress, we can have a competing theory that does not: when a theory is defended ad hoc, then rather than examine its degree of ad hoc defence, we compare it with a competitor and try to prefer the one which is less ad hoc. For a classical example the secular behavior of Mercury was in bad shape from a Newtonian viewpoint and generations of mathematicians and astronomers were trying to see what can be done about it. They chiselled out the difficulties by calculations, and they offered ad hoc hypotheses to explain them away - the body alpha being the best known of these. Yet the ad hoc hypotheses were rejected with ease and nothing remained to be done but to keep chiselling the differences bit by bit by honest calculations of detailed existing perturbations. Then Einstein came to the scene, the difference was almost covered, yet he said the small remainder cannot be handled by a Newtonian theory, only by general relativity! (See above, Chapter 2, Appendix, § 8.)This story is also not in accord with Duhem’s and his followers’ tolerance for ad hoc hypotheses, though it looks so. It is not that ad hoc hypotheses were tolerated till Einstein no longer required their services. Rather, they were not tolerated, and the contradiction was tolerated until resolved or until Einstein said it can’t be resolved and so we must oust Newton.
And who is Einstein to tell us this? It is not his authority but that of the crucial experiments between the two. For clearly, now Newton is so much at a disadvantage no one hopes to remedy all the contradictions his theory holds with factual reports: the hope of reconciling them is gone.
Do we, nonetheless, have instances of ad hoc amendments to resolve contradictions? Indeed we have myriads of them. The writings of both Poincare and Duhem are full of such historical examples. The example of Macquer’s phlogistonist hypothesis, incidentally, is not good enough, because that hypothesis was barely testable at the time. But the classical example in recent discussions, especially of Grunbaum and Popper, is very good. The Lorentz-Fitzgerald contraction does take place and does rescue current theories and is, indeed, taken over by later theories, though no longer ad hoc. Apply this to Macquer and you could see that he could confirm his hypothesis and say, indeed when metal rusts and thus gives out phlogiston it also absorbs air - as indeed we know it does, since rust is metalic oxide! Yet with Macquer, even when confirmed it would remain ad hoc, but in Lavoisier’s theory it was nothing of the sort, as absorbing oxygen was identical with combustion, calcification, etc.
What this will indicate is that rather than resolve the issue we would like to prevent it from rising, or rather push it away. Contrary to the traditional condoning of ad hoc hypotheses, and in line with Popper’s and Grunbaum’s approaches, we see, once an ad hoc hypothesis is introduced we are unhappy about it and try to eliminate it. Moreover, we make it as hard to introduce them as possible. Rather, we say, we note the need for an ad hoc hypothesis and tolerate it pro tem - until the debt either becomes too big and bankrupcy is declared, or else the ad hoc hypotheses come with the factual evidence that supports them.
But this is the way we like things to be, not the way things always are. I do not mean that whenever an ad hoc hypothesis is needed, instead of keeping the token someone makes an ad hoc hypothesis. This is quite in order since some people ignore it and others test it, and thus either offer an empirical backing to it to render it a fact or refute it. Rather, we sometimes take seriously an ad hoc hypothesis which we cannot refute or back by experiment.
The classical example, we remember, is the neutrino. Without it, beta decay refutes the strict conservation of evergy. Yet it was not tested until the days of non-parity, over a generation after Pauli has proposed it. Why?
The answer turns out - not surprisingly if we remember our reluctance to concede ad hoc hypotheses - quite involved. The Bohr-Kramers-Slater hypothesis of statistical conservation of energy was refuted at that time. This is one classic example of a strict Popperian refutation - one theory, one experiment, clash, and the rejection of a theory. This made it hard to accept beta-decay as a refutation of strict conservation: as energy conserves, in fact, its conservation must be strict or statistical. Of course, there is a way out: conservation can be at times strict, at times not. But why? Under what conditions is it strict? No one knew. (This supports Feyerabend’s idea that the decision to permit this or that degree of ad hoc is theory-laden, and undermines his theory as we have here a decision to exclude an ad hoc hypothesis based on pure ignorance.) And, assuming strict conservation, the existence of the neutrino was least ad hoc: it fitted most of the existing framework. The discovery of the neutrino, incidentally, modified our view of it almost entirely: it is still chargeless, of course, as a matter of known fact, and it is still a particle as a matter of general principle; all else has changed!
Let me recapitulate. We saw, with Grunbaum, that ad hoc hypotheses are hard to invent and need empirical testing and confirmation. We saw that we can agree that an ad hoc hypothesis is needed either while we provide none or while we provide an outline of it, until we find a good one which gets empirical backing or until the hope for finding one is given up. The question, how much patience we have before giving up is at times decided by a new theory, at times by circumstances. Yet the circumstances may produce a stalemate, especially when no one cares about the need for improvement. This, to return to my initial point, was Copernicus’ complaint: not that Ptolemy used ad hoc hypotheses, but that the failed to rectify them after so long a period. Was Copernicus in the right? Perhaps not; I think his real complaint lay elsewhere, namely in his heliocentrism; and the real question is, was his complaint launched from a heliocentric viewpoint or was it one logically prior to heliocentrism and one of which the latter relieves us?
IV.