‘AD HOC’ HYPOTHESES WHICH BECOME FACTUAL EVIDENCE
An ad hoc amendment to a refuted hypothesis may be a qualifying clause
excluding the refutation with no further ado. This is just too high-handed and so utterly arbitrary and so inexcusable.
It may be an auxiliary hypothesis explaining the irrelevance of the case. Consider the hypothesis, ‘all birds have feathers’ and its alleged refutation, ‘bats have no feathers’. The exclusion of the bats may be the auxiliary hypothesis that bats are not birds. This may be construed as the auxiliary hypothesis that ‘not all birds have feathers but all flying vertebrates are birds’, is refuted by flying bats. Still, it looks rather arbitrary to pin the fault on the latter hypothesis. It really looks as if now the dual theory, ‘all flying vertebrates are birds and have feathers’, degenerates into a verbal device; all feathered vertebrates, whether flying or not, we name birds. This is evidently highly unsatisfactory since it explains nothing and introduces a verbal convention, i.e. an arbitrary element, for no other purpose than to show stubborness.Now suppose we say all and only birds have feathers, all birds lay eggs, have a beak, two legs, and two wings each, etc. Then, first and foremost, birdness or the name bird can drop out, and the hypothesis now says all feathered animals lay eggs and have a beak, two legs and two wings each. Flight now has dropped out of the picture on account of there being flying non-birds. The hypothesis looks decent enough and it need not refer to flight, after all. Yet the bare omission of flight, the very move made, is suspect of arbitrariness. Felicitous as our hypothesis is, we may still feel it would have been more felicitous had it covered flight as well.
Now this uneasy feeling will be easily allayed, as the reader knew all along, the moment bats are discovered to be mammals proper and on the hypothesis that no mammal has a beak and wings or lays eggs.
Also, of course, the existence of chickens and ostriches which do not fly make flight more problematic and so its exclusion pro tern more understandable i.e. ad hoc but not so arbitrary as to be helpless.The picture can get complicated again. The discovery of the duckbilled platypus who has a beak and lays eggs yet is a mammal may raise afresh the problem, is the bat not a bird. Fortunately, the theory behind taxonomy makes it quite satisfactory to omit flight altogether and center on philogenetic characteristics which make the bat definitely a member of a different family altogether.
We can go further. Adolf Grunbaum has contributed to the debate on ad hoc hypotheses a very important observation which runs contrary to almost all that philosophers said about the topic, from Copernicus and Bacon through Poincare and Duhem to Popper: there are instances in history showing that it is not always easy to generate ad hoc hypotheses. Also, as we saw, the ad hoc hypothesis needs some empirical backing to make it preferable to its possible competitors, and this may not be easy to come by. Let us take an example. Combustion, according to phlogis- tonism, always reduces the weight of the fuel: the ashes are lighter than the original fuel. This is not so in the case of metals. An ad hoc explanation came in the form of Archimedes’ hypothesis, and was repeated even by a writer of standing like Peter Shaw although it is immediately clear that the order of magnitude of the Archimedes effect in the air is much too small and so cannot effect the case one way or another. Macquer had the hypothesis that during combustion of metals a secondary process takes place which adds more weight than the weight lost through the primary process. But he could not identify it, let alone empirically back it.
These examples prove Grunbaum’s case quite conclusively, of course, except that historians of science may well exclude these examples ad hoc. Yet, as long as a theory, in our example phlosistonism, is engaging enough, we may put aside the troublesome cases in the hope that they will be covered one day by an ad hoc hypothesis that will be empirically vindicated. After all, we can imagine a universe where Macquer’s hypothesis comes true and is empirically supported.
Examples may help push this point further. There are examples like Macquer’s, and even more extravagant ones which proved correct. Consider the known fact that metals cannot be electrified. William Gilbert’s identification of electricity and magnetism with matter and form respectively is clearly refuted by it, yet he seemed scarcely to be bothered by his own division of things to electric and non-electric, as well as to magnetic and non-magnetic. When, over a century later, Stephan Gray declared metals electric all the same yet whose electricity escapes in a secondary process, no one declared his hypothesis ad hoc, even though it is evidently more extravagant than Macquer’s, simply because he gave it the full empirical backing by electrifying metals while stopping the secondary process by preventing the escape of the electricity, by the use of what was previously known as electric matter and which he presented as insulators.
The difficulty, then, which Grunbaum indicates, is rooted in the fact that a good ad hoc hypothesis, though ad hoc, is at times empirically testable with relative ease, leading to its establishment or refutation. When established it gains a fact-like status.
Another example. To save Kirchhoff’s spectroscopy from refutation a few elements were postulated and their presence in the sun’s atmosphere was blamed for puzzling spectral lines. Now one of these, helium, was later established on earth, all the others were later declared chimerical. Kirchhoff’s law was in part rescued by other means, in part qualified and modified in a quite ad hoc manner and remained so until the advent of the new quantum theory, at least.
It would be quite ad hoc, I confess, to justify those who use a given method successfully and to condemn those who do exactly the same unsuccessfully.
The postulations of the existence of certain elementary particles, particularly the magneton and the neutrino, have the same characteristics. Indeed, the neutrino was more ad hoc than the magneton, since the neutrino came to save a theory from refutation, whereas the other came to clinch a theory.
Yet the one was found and elevated from the status of a very ad hoc hypothesis to the status of an attested fact, whereas the other remains still somewhat ad hoc. And, indeed, Dirac’s theory which encompasses it is somewhat suspect; but other theories with doubtful ad hoc amendments exist all around, and the ad hoc amendments are well tolerated. The most powerful example is the inapplicability of Men- delism to human skin colors which is rescued by the mere say so of geneticists that over twenty genetic factors go into the making of human skin color. We still wait to see this say so put into detailed specifications; they may then easily be established or refuted.To conclude, I observe that we differentiate ad hoc between ad hoc amendments. We have those which we approve of, either because they later earned the status of facts or because they rescue theories we favor. And we condemn others. We forget that the method, purpose, and initial legitimacy are the same, and the differentiation is only post hoc - posterior to further investigation. The procedure is rational in both cases, or else posterior investigations are not necessary. Those, then, who condemn ad hoc amendments as a prioristic because they come - undoubtedly - to rescue theories or hypotheses from empirical refutations, those same people, then, act arbitrarily or are defenders of a much deeper (methodological) a priorism.
Grunbaum’s profound observation, then, has an interesting corollary. A scientist who wants an ad hoc hypothesis but cannot create one on the spur of the moment can leave the task pro tem unattended and pretend or hope that one is coming fairly soon that will be empirically well substantiated. This is a high risk a scientist takes, of course. The question is, is the risk reasonable, and how does it grow in time?
II.