REVOLUTIONS OCCUR WHEN BOOTSTRAP OPERATIONS FAIL

To further clinch matters, let me draw attention to the uncalibrated part of our instruments. There are assumptions which no one questions and so no one tests. Some calibrations, obviously, depend on the geographic location of the instruments; some only on altitude or on lattitude; some not on location at all.

When we say it seems mad, we usually mean to say, you are unlikely to succeed in pursuing this line of thought. Philosophers of the common­sense persuasion from Locke to Duhem have suggested - implicitly, to be sure - that what is seemingly mad will never succeed. As Bohr has hinted, when the seemingly mad succeeds it looks more ingenious than when the plausible succeeds. The seemingly mad may conflict with deep-seated metaphysical theories which are shared by all or most well-corroborated physical theories. But the seemingly maddest may succeed: the odds are smaller and the stakes are higher: the game is fair.

It was Mario Bunge who, in his Scientific Research, recommended to aim always at the smallest upheaval with the hope that it fail and so lead us to a bigger one. I find this a magnificent modification of Popper’s proposal: Everyone says that corroboration is a Good Thing and Popper says that a refutation is a Good Thing as it leads to upheavals. Now Bunge says, when you rebel against a minor part of the Received Opinion, offer a minor correction and hope your correction will not do, thus leading to the search for a more fundamental correction. This, also, is a magnificent modification of Duhem’s proposal (though Bunge strongly rejects Duhem’s instrumentalistics views): try the easiest and smallest modification, but hope to fail at once, not hope to succeed and look for the next small modification.

But now, it is clear, when the new modification, or rather correction, goes far enough from the old view, then we may need to go back and find intermediary steps in order to correlate the old and the new views. We will then need approximate theories between the old and the new, and each approximation will have to be corroborated before it can be used. For example, between Newton’s and Kepler’s theory, we have the theory of interaction, in accord with Newton’s law, only between the sun and each planet alone, and we have to juggle even that theory since the sun interacting with Jupiter is the same as the one interacting with Saturn. But we check the result empirically and in corroborating it we calibrate our system of constants. There are many other examples, from the history of the determination of atomic weights, from the history of genetics, and others. There is no doubt that in the history of measuring atomic weights bootstrap operations were made all along the way, and that many re­futations for every atomic measurement were abundant - and correctly ignored - at each stage of development prior to the development of mass spectrography which overruled many auxiliary hypotheses as redundant. No sooner was one mass determined more accurately than the others, and it was used as the new anchor both to increase the precision of the mea­sures of other atoms and to attack new compounds. The story of the determination of chief constants of quantum theory, to wit, h, k, e, m, and c, bears a great similarity.

In every bootstrap operation the assumption used need not be accepted as true - they may be good approximations - but each of their empirical applications needs corroboration in order to make it harder to ignore, by those who wish to ignore, the results of the future experiments. We may say, then, that corroboration is what gives a claim its factual status; but not that the claim for factual status cannot be overthrown.

One final and striking instance - not for bootstrap operations but for what counts as factual - is this. When levels of celestial long wave radia­tion were on occasion observed, depending on geographical locations and times of day, there was no difficulty in assuming that the variations were due to terrestial transmissions of radio messages rather than to cosmic variations. Of course, in some cases these were corroborated when pre­cision was so increased as to enable one to dicipher the messages rather than some cosmic variations. Of course, clearly, there are celestial sources of radiowaves. There was no problem here to decide which is which since every move was highly corroborated - things were fairly much on the surface - and the high degree of actual corroboration is what gives more facticity to a theory; what is more fact-like and what is more theory-like varies according to experimentation.

And so we can also offer a slightly new variant of a very widely held theory -1 think due to David Hume (in his Dialogues) - of what we con­sider fact-like. One must take some precaution here, since the word ‘fact’ may be stretched without severly violating common usage. For example, Laplace drove home his opinion that Newton’s mechanics is a verified generalization by calling gravitation’s (alleged) absolute obedience to the inverse square of the distance a fact. Clearly, in some sense Newton’s gravitational theory, even if true, is not fact-like. In this essay the word ‘fact’ is used in the narrow sense established by William Gilbert (1600) and Robert Boyle (1662), namely that which is may be reported by an eye-witness.

In particular, one must notice, we take here an a priori stand about what causes are excluded - e.g. witchcraft - what causes may be operative under what conditions - e.g. lies - and what truthfully may be reported as observed yet be an illusion, or the result of defects or of inexact in­struments of observations. Here let me draw attention to one classical difficulty, which has been noted by many writers, including Bacon and Popper: if we use theory to judge the truth of a testimony, then we shall end up dogmatists, by simply having to deny the truth of all testimony which refutes our theories. This led Bacon to the taboo on all theory. It led Boyle, (and many others including Popper) to the dictum: when in conflict, prefer observations to theory.

This rule, Boyle’s rule, we saw, is empirically refuted. Also, it is easy to replace Boyle’s rule: we can assume the theory to be true, and we can assume the theory to be false, and replace it by diverse alternatives. We can judge the situation from each viewpoint in a different manner, but choose the viewpoint from which we judge on different grounds al­together, for example from the viewpoint of high testability, or of in­terest. In particular, we may deliberately pursue more than one line of research.

Science, it is well-known, is in a constant mess. The reason is, I feel, that while we check the details of both theory and fact we stumble upon more interesting and significant problems, and often leave in the middle our attending to older problems and tasks. To see, in particular, what kind of problem is most likely to be left behind, we have to have a closer look at corroboration, since what is left behind, is the job of fitting corrob­orated fact-like hypotheses with corroborated theory-like hypotheses which do not fit together comfortably by themselves. The bad fitting is only possible because the way the two are corroborated may be widely different.

VII.