Local realism and the crucial experiment

Yoav Ben-Dov

Cohn Institute for the History of Science
Tel-Aviv University

Conference Talk published in: Frontiers of Fundamental Physics, Ed. F. Selleri, Plenum Publications, London 1994

back to my english or main index page

It is well known that the history of quantum mechanics is riddled with conceptual debates. The most famous debate is that which took place between Einstein and Bohr from 1927 onwards, and in which the two protagonists can be taken as representing two possible answers to what is perhaps the most central question concerning quantum mechanics. Both sides in this debate agree that the usual quantum formalism, as developed by Heisenberg, Schroedinger and Dirac, is inconsistent with what was believed, from the 17th century onwards, to be the central prerequisites of a sound physical theory: determinism, locality (which means that things don't act at places where they are not present), clear separation between object and subject (what we shall call here "realism" - the view that physical reality can be defined "as it is" without any reference to an observer), and so on. They differ, however, in their reaction to this situation. Should we look for an alternative theory, which would satisfy these prerequisites while still accounting for the observed quantum phenomena (Einstein's position), or should we rather accept quantum mechanics as it is, and try to adapt our epistemological prerequisites to the new formalism (Bohr's position)?

From a historical point of view, one may see some point in Bohr's claim. Einstein's prerequisites are modelled after the classical physical theories, for example Newtonian mechanics. But the acceptance of Newtonian mechanics was a historical event, and the prerequisites associated with it were introduced along with the same event. In fact, Newton's mechanics was at that time inconsistent with the previously accepted prerequisites of Aristotelian science, for example with the demand that a theory describe the purpose of things' being as they are, and not only their quantitative behavior. Thus, epistemological prerequisites change with the current scientific theories (so that they are actually "post-requisites"). Why should not the same hold for quantum mechanics?

Supporters of Einstein's view may, however, turn this argument around, and claim that Aristotelian science was much less successful than Newtonian science exactly because it tried to satisfy the wrong prerequisites. From the 17th century onwards, modern science arrived at immense achievements by following the classical prerequisites. Doesn't their definitive abandonment involve the risk of undermining the basis of the whole modern scientific enterprise? In other words, isn't a truly non-classical science simply a contradiction in terms?

Of course, Einstein's position must be abandoned if one can prove that no theory which satisfies the classical prerequisites can possibly account for the observed phenomena. The question is, do we have such a proof? As for the prerequisite of determinism (or "classical causality"), we know today that Von Newmann's alleged "proof" to this effect is invalid [1], and with the de Broglie-Bohm "pilot wave" theory [2] we even have a counter-example of a fully deterministic theory which perfectly reproduces all the quantum phenomena. However, many physicists and philosophers today believe that the violations of Bell's inequality [3], such as are manifested in Aspect's [4] and other experiments, are inconsistent with the conjunction of realism and locality, so that at least one of these two classical prerequisites should be abandoned.

Presented thus, Aspect's experiments appear as a "crucial experiment" in the original sense of Francis Bacon: in its progress, science arrives at a crossroad, where two mutually exclusive possibilities are open. An experiment is devised to decide between them, so that one can be discarded and the other accepted as true. Here, the alternatives are "Nature can be described by a local realistic theory" and "Nature cannot be described by a local realistic theory", and Aspect's experiment decides in favor of the second. It is in this sense that John Bell was called, in one obituary, "the man who proved Einstein wrong": any attempt to retain the complete list of classical prerequisites, and in particular, any attempt to devise a successful local realistic theory, is doomed to fail.

However, we should perhaps be more careful. At the beginning of this century, the great physicist and historian of science, Pierre Duhem, tried to introduce the following distinction between physics and metaphysics [5]. In his view, physics concerns the classification and prediction of experimental results, while metaphysics concerns the nature of reality itself. Following Duhem's distinction, we can claim that a physical theory can never be regarded as absolutely definitive, because our belief in it is based only on a finite series of experiments, namely those which have been performed so far. It is always possible that a different theory will still explain the same results, and also account for the results of future experiments for which the present theory proves inadequate. In this sense, a physical theory can be "correct" at some moment and "incorrect" later. In contrast, a metaphysical claim is supposed to be absolute, that is, "true" or "false" for all time, irrespectively of any experiment which we are able to conduct at a given moment.

According to this distinction, Aspect's experiments belong to the domain of physics, that is, they should be accounted for by physical theory. On the other hand, the statement "no local realistic theory is possible" is a metaphysical claim, as it asserts something about reality itself. The question is, whether a physical experiment can decide in a definite manner a metaphysical question - that is, whether a "crucial experiment" whose verdict is absolute is at all possible. Indeed, there are some who, apparently accepting a distinction not very different from Duhem's, refer to Bell's inequality and Aspect's experiments as "experimental metaphysics".

Duhem, however, regarded "experimental metaphysics" as an impossible expression. In his view, Metaphysical statements and experimental results exist in two different worlds. To bring them together, much elaboration and interpretation is required, and this involves many assumptions which can always be eventually revised. even if we work out our metaphysical positions into specific hypotheses cast in the mathematical language of physics, still "an experimentum crucis (to choose between them, Y.B.) is impossible in physics". The reason is twofold.

First, a single hypothesis by itself is not testable by experiment. At most, the experiment can test a complete theory, which always involves many additional hypotheses, some of them (perhaps the most important, and probably the most problematic) implicit. Suppose for example that the experiment has clearly condemned a certain theory. Still, we do not know which of the many hypotheses entering the theory is the one at fault, and we can never be sure that it is the one that we originally wanted to test. Second, whenever we formulate a choice between hypotheses as a true dilemma - either p is true, or q - it is possible to cast a doubt on the nature of the dilemma itself: perhaps an additional hypothesis has implicitly entered our considerations, or our imagination is too weak; how can we be absolutely sure that these are really the only two possibilities - that some intermediate (or otherwise different) possibility cannot be found, or that the very terms in which the question was posed are for ever beyond revision?

To illustrate his point, Duhem gives a historical example, which the years have rendered even more striking. At the beginning of the 19th century, two rival hypotheses concerning the nature of light existed in physics: the corpuscular theory, which is usually ascribed to Newton, and the ondular theory, which originated with Descartes and Huygens, and was revived by Young and Fresnel. According to the first theory, light is a stream of tiny particles; according to the second, it is a disturbance propagating in a medium, the luminiferous Ether. These two theories could account, each in its own terms, for many experimental phenomena, including the refraction of light as it passes from one transparent medium to another. But there was an important point of difference between them: the corpuscular theory assumed for its explanation of refraction (ironically, first formulated by Descartes) that the speed of light in a transparent medium (such as glass or water) is greater than the speed of light in empty space. The ondular theory, on the other hand, assumed exactly the opposite. On the basis of this difference, François Arago proposed a "crucial experiment" which would compare the two speeds, thus deciding once and for all whether light is a particle or a wave. Only some decades later, experimental technique was refined enough so that Léon Foucault could devise an actual experimental set-up along these lines. The experiment was carried out in 1850, with clear results: light travels more slowly in water that in air (which, for this purpose, is approximately equivalent to empty space). Therefore, as the physicists then concluded, light is a wave, and the corpuscular hypothesis can be discarded, not to be bothered with any more.

But for Duhem, this conclusion is very doubtful. First, he points out that Foucault's experiment did not decide between two isolated hypotheses, but between two complete theoretical systems. True, the corpuscular theory as held by early 19th-century Newtonians is condemned. But it is not inconceivable that a future theory might be built upon the corpuscular hypothesis, with the aid of some new auxiliary hypotheses which would be different from those entering the Newtonian system. In such a theory, the refraction of light might be explained in a different manner, so that it would be possible to account for the results of Foucault's experiment while still maintaining that light is a particle. Second, it is not at all certain that the current concepts "wave" and "particle" are the only possible ones; perhaps a new concept might be formulated, which would go beyond this dichotomy, possibly by combining some aspects of both concepts.

When Duhem formulated his arguments, this was only an abstract possibility. However, shortly afterwards, Einstein has re-introduced into physics the idea of a corpuscular nature of light, starting with his 1905 paper on light quanta. This paper has paved the road to present-day quantum mechanics. But as we know, the debate on the nature of quantum objects (including light quanta) and on the wave- particle duality still goes on. Thus, we can say that more than a century after Foucault's "crucial" experiment was supposed to decide it once and for all, the question of the nature of light is still open. Moreover, we do not even expect any more a simple answer such as "only wave" or "only particle" in the classical sense, a situation which provides a striking example of the acuteness of Duhem's vision. We should also note that nowhere in the present debate on the nature of quantum objects is Foucault's result even mentioned as evidence. Of course, no one doubts the result itself, that is, we are still convinced that light travels more slowly in water than in air, and that any future physical theory should account for this fact. But the experimental result itself has lost all relevance to the question of the nature of light. In Duhem's terms, today we interpret Foucault's set-up and his results as an experiment in physics, not in metaphysics.

Perhaps similar considerations can also be applied to Aspect's experiments. Between the metaphysical statement "a local realistic theory is impossible" and the actual experimental set-up there is a huge gap, which can only be bridged with the aid of many auxiliary hypotheses. Any one of these could be wrong. Proceeding from the experimental side, we can, for example, point out that there are "experimental loopholes" [6] in Aspect's experiments, which, if investigated further, might turn out to be responsible for the result. We can also suspect the existence of some "selection effect" which influences the detection probabilities, so that Aspect's experiments do not actually test Bell's inequality [7]. We can accept the possibility that some additional implicit assumption has entered into the mathematical derivation of Bell's inequality, or doubt that the mathematical criteria used in this derivation are accurate and complete translations of the metaphysical concepts "realism" and "locality". And surely, there are many more possibilities, which we cannot see from within the network of present-day physical concepts.

All this, of course, is not meant to undermine the value of Aspect's experiments. These are surely beautiful physical experiments. But those who interpret them as "experimental metaphysics" seem to ignore the force of Duhem's arguments. It is quite possible that exactly like Foucault's experiments, physicists in the next centuries would interpret Aspect's experiments as physically valid, but metaphysically irrelevant. Thus, notwithstanding Aspect's experiments, it cannot be maintained that a local realistic theory which will account for the observed phenomena is definitely ruled out, once and for all. At most, as Duhem puts it, it may be against the "good sense" of our time to put too much effort in looking for such a theory. But the context may change, and a future "good sense" may indicate otherwise.


[1]. J.S. Bell, Rev. Mod. Phys. 38:447 (1966).

[2]. D. Bohm, Phys. Rev. 85:166 and 180 (1952).

[3]. J.S. Bell, Found. of Phys. 12:989 (1982).

[4]. A. Aspect, J. Dalibard and G. Roger, Phys. Rev. Lett. 49:1804 (1982).

[5]. P. Duhem, "La théorie physique, son objet - sa structure", Vrin, Paris (1989). English translation: "The Aim and Structure of Physical Theory", Atheneum, NY (1974).

[6]. A. Shimony, An exposition of Bell's theorem, in: "62 Years of Uncertainty", A.I. Miller, ed., Plenum Press, NY (1990).

[7]. F. Selleri, Einstein-de Broglie waves and two-photon detection, in: "Bell's Theorem and the Foundations of Modern Physics", A. van der Merwe, F. Selleri and G. Tarozzi, eds., Plenum Press, NY (1992).