CHAPTER 1
Some surprising phenomena
Paul Feyerabend famously asked, âWhatâs so great about science?â (1976: 310). One possible answer is that it has been surprisingly successful in getting things right about the natural world. And, very plausibly, a reason why science has been of interest to philosophers is because it seems to have been more successful in doing this than non-scientific or pre-scientific systems, or religion, or philosophy itself. We seem, moreover, to be able to point to some general ways in which science has been surprisingly successful in getting things right. Here are three such:
1. Scientists have formulated some theories that have successfully predicted novel observations.
2. Scientists have produced theories about parts of reality that were not observable or accessible at the time those theories were first advanced, but the claims about those inaccessible areas have since turned out to be true.
3. Scientists have, on occasion, advanced on more or less a priori grounds theories that subsequently turned out to be highly empirically successful.
Of course, it may be disputed whether these phenomena really are genuine. But if they are genuine, they are surprising, and therefore require explanation. The main aim of this book is to offer an explanation of these phenomena. The aim of the present chapter is to argue that there are phenomena here that genuinely do require explanation.
THE FIRST PHENOMENON
The first of the phenomena is that scientists have formulated theories that have subsequently turned out to have novel predictive success. observational predictions made by scientific theories are, roughly speaking, of two kinds: predictions of observations that are of the same kind as those on the basis of which the theory was initially formulated and predictions that are of a new or different kind from those initially used. suppose a scientist is studying magnets. The scientist may note that all magnets observed so far have two âpolesâ, a north and a south pole. This leads to the conjecture that all magnets have both a north pole and a south pole. And this conjecture in turn leads to the prediction that the next magnet to be observed will have both a north and a south pole. But this is not a novel prediction. The observation that the next magnet has a north and a south pole is just an observation of the same kind as that which has already been made with previous magnets. We will call it an example of familiar predictive success.
There are some predictions, however, that seem to be novel. One was the prediction that bringing enough uranium-235 into very close proximity will produce an explosion âbrighter than a thousand sunsâ.1 This certainly seems like a novel prediction: no sequence of events of this sort had ever been observed before; it was, moreover, a prediction that turned out to be successful. It is an example of what we would be inclined to call novel predictive success. Another example concerns the behaviour of light. The eighteenth-century scientist Poisson showed that it was a consequence of Fresnelâs wave theory of light that if a perfectly round object is placed in a beam of light, the resulting round shadow will have a small white spot in its centre. This phenomenon had, as far as was known at the time, never been observed before, and so it had not been used as evidence for a theory of light. But when the experiment was performed, the shadow was observed as predicted. Yet another example comes from the special theory of relativity. This theory predicted that if two extremely accurate clocks were first synchronized, one of them moved a sufficient distance while the other remained stationary, and then the two reunited, the one that had remained stationary would be slightly ahead of the other. Again, this sort of phenomenon had not been earlier observed, and so had not been used to first construct the special theory. But when an experiment of this sort was performed, the clock that had remained stationary did turn out to be ahead of the other, in agreement with the amount predicted by the theory (Hafele & Keating 1972).2
Some more examples3 of novel predictive success are:
⢠The prediction of the observable chemical behaviour of the âtransuraniumâ elements. These are artificial elements, produced in particle accelerators. They contain nuclei that are larger than the nucleus of the uranium atom, which is the largest atomic nucleus known to occur naturally. since these elements are artificial, their behaviour had not been observed prior to their creation. Therefore predictions about their chemical behaviour count as novel predictions. But scientists successfully predicted their chemical behaviour.
⢠A related case of novel success is the prediction, from gaps in the periodic table, of hitherto unknown elements and their chemical properties.
⢠The prediction, by Wolfgang Pauli in 1925, of the existence of the neutrino and the subsequent confirmation of its existence in experiments performed by Cowan and Reines in 1951.4
⢠The prediction made by the general theory of relativity, published by Einstein in 1915, that the observed position of a star would be deflected from its actual position by a powerful gravitational field, and the subsequent confirmation of this prediction in 1919 by Eddington during a solar eclipse.5
⢠The prediction by the general theory of relativity that, due to âtime dilationâ caused by the strong gravitational field of the Sun, light emanating from the Sun would appear to be shifted towards the red end of the spectrum, and subsequent confirmations of this effect.6
⢠The prediction of observational results confirming the existence of the W and Z particles as a consequence of the theory that the weak subatomic force and the electromagnetic force were just different manifestations of a single underlying force, and the subsequent confirmation of the existence of these particles in experiments performed at CERN in 1983.7
All of these examples are cases in which a theory successfully predicted phenomena which are, intuitively at least, different from any of those on the basis of which it had initially been formulated. It seems to be beyond serious dispute that science does, at least sometimes, succeed in making novel predictions. so, in what follows, it will be assumed that this phenomenon is genuine.
Of course, one serious problem is: can the notion of a novel prediction be precisely defined and is it, for our purposes, necessary to do so? This question is considered in Chapter 4. For the moment, however, we will simply note that, at an intuitive level, it seems that there are indeed cases of novel predictive success.
THE EXPLANATION OF FAMILIAR PREDICTIVE SUCCESS ANDTHE EXPLANATION OF NOVEL PREDICTIVE SUCCESS
Examples of familiar predictive success are common and easy to obtain. I note that whenever I have started my lawnmower, the neighbourâs dog has started to bark; I then predict that the dog will bark next time I start my mower, and this prediction proves to be successful. I note that whenever my car has frost on it, it is hard to start; I predict the next time it has frost on it, it will again be hard to start, and this prediction is successful. And so on. Predictive successes of this sort occur very frequently. How do we explain familiar predictive successes? Presumably, the explanation will be along the following lines:
The world contains certain regularities or uniformities of the form: âWhenever A, then Bâ. If a person has noticed that, in the past, instances of A have been followed by instances of B, and if they consequently come to predict that the next instance of A will be followed by an instance of B, then there is a good chance their prediction will be successful.
Of course, it is only instances of certain types of generalizations that lead to predictions that have a good chance of being successful. For example, perhaps every coin I have taken out of my pocket this morning was minted before 1998, but that does not make it likely that the next coin I take out of my pocket will also have been minted before 1998. It seems that only certain types of past regularities are likely to persist in to the future. so, if we are to explain familiar predictive success, it is important to stipulate that the instances of regularities that have been observed in the past must have been instances of regularities of a certain type. The problem of saying just what types of regularities those are is a difficult one: here we will just refer to them as âappropriateâ regularities. Provided that the notion of an appropriate type of regularity can be explicated, the explanation of familiar success will be relatively straightforward. But the explanation of novel success is not so easy. Consider, for example, the prediction of the white spot in the middle of a circular shadow. Instances of the regularity âWhenever a round object is placed in a beam of light, there will be a white spot in the middle of the shadowâ had never been observed prior to their first derivation from theory. So we cannot use the explanation of familiar predictive success to explain why the prediction of the white spot was successful. It seems that it wasnât just a new instance of a regularity that had been observed before, but a new regularity. An explanation of a different kind is needed if we are to understand how scientists came to make this successful prediction. There is, therefore, a problem with explaining novel predictive success that is not present when we are merely trying to explain familiar predictive success.
On the face of it, cases of novel predictive success are extremely surprising. It is as if someone were to observe that all magnets have north and south poles and to then make the surprisingly unconnected prediction that, for example, the next bird to be observed will have green plumage, and for this prediction to be found to be correct. From a purely empiricist point of view, it certainly seems very odd that a series of observations in one domain should lead to a prediction in another, quite different domain, and for that prediction to subsequently turn out to be right. It is also very puzzling if an instrumentalist interpretation of scientific theories is adopted. It is as surprising as if a tool originally designed to do one job, such as opening tin cans, should also be able to do another, quite different job, such as programming DVD recorders.
There are some philosophers who deny that any special status attaches to novel predictive success. It has, for example, been argued that the novel predictive success of a theory does not support a scientific realist interpretation of that theory, and that neither does it confer any especially high degree of confirmation on the theory. But it is not claimed here that novel predictive support does do either of these things. Here it is only claimed that the various forms novel success described require explanation, and the inductive explanation that can be given of non-novel success is not, or at least is not obviously, satisfactory for those types of predictive success we are inclined to classify as ânovelâ.
Of course, the phenomenon of novel predictive success is often advanced as a reason for accepting scientific realism, or the doctrine that (mature) scientific theories are true, or approximately true, descriptions of (sometimes unobservable) parts of reality. But if scientific realism is true, it merely raises another puzzle. Our original problem was âHow have we managed to hit upon true predictions of types of phenomena not observed at the time of the prediction?â If scientific realism is accepted, we are confronted with the new problem âHow have we managed to hit upon true descriptions of parts of reality not directly observable?â Whether scientific realism is true or false, it offers a solution to our original problem by postulating another phenomenon which, on the face of it, appears to be at least as puzzling.
THE GENUINENESS OF THE SECOND PHENOMENON
The second phenomenon is the ability of science to produce theories that make true statements about parts of the world that were not accessible to those who formulated those theories.
The claim that this second phenomenon is genuine must be distinguished from the thesis of scientific realism, at least as that doctrine is frequently interpreted. A common interpretation of scientific realism is that mature scientific theories are typically (approximately) true. An alternative formulation, which certainly need not be equivalent to the first, is that the terms of mature scientific theories are typically referential, or that the entities postulated by typical mature scientific theories actually exist. But in either of these formulations, scientific realism is a much stronger thesis than the claim that there are some genuine examples of phenomenon 2. scientific realism is a thesis about typical mature scientific theories. It therefore implies, at least, that most mature scientific theories are approximately true, or referential, or both. But phenomenon 2 does not make any claim at all about most theories; it only says that some theories have given us true descriptions of parts of the world inaccessible to those who advanced the theory.
One example of phenomenon 2 concerns the hypothesis of the planet Neptune. It was first suspected that there might be a planet beyond Uranus that was causing perturbations in its orbit. In 1843 John Adams calculated the position of this hypothetical new planet, and three years later Urbain Le Verrier independently did the same. Initially, therefore, Neptune was merely an entity âof theoryâ: not observed, but postulated to explain the behaviour of Uranus. But the existence of Neptune is now beyond serious dispute. Not only is it clearly visible through telescopes, but spacecraft have flown past it, taking photographs of it. A part of reality not accessible to those who first advanced a theory has since become accessible.
Another example of phenomenon 2 is the germ theory of disease. When first advanced, the theory that disease was due to the presence of tiny organisms made reference to inaccessible parts of reality. But now the reality of ...