Part 4 (1/2)

Advantages of falsificationism over inductivism.

With a summary of the basic features of falsificationism behind us, it is time to survey some of the advantages that this position can be said to have over the inductivist position according to which scientific knowledge is inductively derived-from given facts, which we discussed in earlier chapters.

We have seen that some facts, and especially experimental results, are in an important sense theory-dependent and fallible. This undermines those inductivists who require science to have an unproblematic and given factual foundation. The falsificationist recognises that facts as well as theories are fallible. Nevertheless, for the falsificationist there is an important set of facts that const.i.tute the testing ground for scientific theories. It consists of those factual claims that have survived severe tests. This does have the consequence that the factual basis for science is fallible, but this does not pose as big a problem for falsificationists as it does for inductivists, since the falsificationist seeks only constant improvement in science rather than demonstrations of truth or probable truth.

The inductivist had trouble specifying the criteria for a good inductive inference, and so had difficulty answering questions concerning the circ.u.mstances under which facts can be said to give significant support to theories. The falsificationist fares better in this respect. Facts give significant support to theories when they const.i.tute severe tests of that theory. The confirmations of novel predictions are important members of this category. This helps to explain why repet.i.tion of experiments does not result in a significant increase in the empirical support for a theory, a fact that the extreme inductivist has difficulty accommodating. The conduct of a particular experiment might well const.i.tute a severe test of a theory. However, if the experiment has been adequately performed and the theory has survived the test, then subsequent repet.i.tions of that same experiment will not be considered as severe a test of the theory, and so will become increasingly less able to offer significant support for it. Again, whereas the inductivist has problems explaining how knowledge of the un.o.bservable can ever be derived from observable facts, the falsificationist has no such problem. Claims about the un.o.bservable can be severely tested, and hence supported, by exploring their novel consequences.

We have seen that inductivists have trouble characterising and justifying the inductive inferences that are meant to show theories to be true or probably true. The falsificationist claims to bypa.s.s these problems by insisting that science does not involve induction. Deduction is used to reveal the consequences of theories so that they can be tested, and perhaps falsified. But no claims are made to the effect that the survival of tests shows a theory to be true or probably true. At best, the results of such tests show a theory to be an improvement on its predecessor. The falsificationist settles for progress rather than truth.

Further reading.

For Popper's mature reflections on his falsificationism see his 1983 text, Realism and the Aim of Science. Schilpp (1974), in the Library of Living Philosophers series, contains Popper's autobiography, a number of articles on his philosophy by critics, and Popper's reply to those critics, as well as a detailed bibliography of Popper's writings. Accessible overviews of Popper's views are Ackermann (1976) and O'He ar (1980). The modification of Popper's views involved in the section ”Confirmation in the falsification account of science” is discussed in more detail in Chalmers (1973).

CHAPTER 7:.

The limitations of falsificationism.

Problems stemming from the logical situation.

The generalisations that const.i.tute scientific laws can never be logically deduced from a finite set of observable facts, whereas the falsity of a law can be logically deduced from a single observable fact with which it clashes. Establis.h.i.+ng by observation that there is just one black swan falsifies ”all swans are white”. This is an unexceptional and undeniable point. However, using it as grounds to support a falsificationist philosophy of science is not as straightforward as it might seem. Problems emerge as soon as we progress beyond extremely simple examples, such as the one concerning the colour of swans, to more complicated cases that are closer to the kind of situation typically met with in science.

If the truth of some observation statement, 0, is given, then the falsity of a theory T which logically entails that 0 is not the case can be deduced. However, it is the falsificationists themselves who insist that the observation statements that const.i.tute the basis of science are theory-dependent and fallible. Consequently, a clash between T and 0 does not have the consequence that T is false. All that logically follows from the fact that T entails a prediction inconsistent with 0 is that either T or 0 is false, but logic alone cannot tell us which. When observation and experiment provide evidence that conflicts with the predictions of some law or theory, it may be the evidence which is at fault rather than the law or theory.

Nothing in the logic of the situation requires that it is always the law or theory that should be rejected on the occasion of a clash with observation or experiment. A fallible observation statement might be rejected and the fallible theory with which it clashes retained. This is precisely what was involved when Copernicus's theory was retained and the naked-eye observations of the sizes of Venus and Mars, which were logically inconsistent with that theory, discarded. It is also what is involved when modern specifications of the moon's trajectory are retained and estimates of its size based on unaided observation rejected. However securely based on observation or experiment a factual claim might be, the falsificationist's position makes it impossible to rule out the possibility that advances in scientific knowledge might reveal inadequacies in that claim. Consequently, straightforward, conclusive falsifications of theories by observation are not achievable.

The logical problems for falsification do not end here. ”All swans are white” is certainly falsified if an instance of a non-white swan can be established. But simplified ill.u.s.trations of the logic of a falsification such as this disguise a serious difficulty for falsificationism that arises from the complexity of any realistic test situation. A realistic scientific theory will consist of a complex of universal statements rather than a single statement like ”All swans are white”. Further, if a theory is to be experimentally tested, then more will be involved than those statements that const.i.tute the theory under test. The theory will need to be augmented by auxiliary a.s.sumptions, such as laws and theories governing the use of any instruments used, for instance. In addition, in order to deduce some prediction the validity of which is to be experimentally tested, it will be necessary to add initial conditions such as a description of the experimental set-up. For instance, suppose an astronomical theory is to be tested by observing the position of some planet through a telescope. The theory must predict the orientation of the telescope necessary for a sighting of the planet at some specified time. The premises from which the prediction is derived will include the interconnected statements that const.i.tute the theory under test, initial conditions such as previous positions of the planet and sun, auxiliary a.s.sumptions such as those enabling corrections to be made for refraction of light from the planet in the earth's atmosphere, and so on. Now if the prediction that follows from this maze of premises turns out to be false (in our example, if the planet does not appear at the predicted location), then all that the logic of the situation permits us to conclude is that at least one of the premises must be false. It does not enable us to identify the faulty premise. It may be the theory under test that is at fault, but alternatively it may be an auxiliary a.s.sumption or some part of the description of the initial conditions that is responsible for the incorrect prediction. A theory cannot be conclusively falsified, because the possibility cannot be ruled out that some part of the complex test situation, other than the theory under test, is responsible for an erroneous prediction. This difficulty often goes under the name of the Duhern/Quine thesis, after Pierre Duhem (1962, pp. 183-8) who first raised it and William V.O. Quine (1961) who revived it.

Here are some examples from the history of astronomy that ill.u.s.trate the point.

In an example used previously, we discussed how Newton's theory was apparently refuted by the orbit of the planet Ura.n.u.s. In this case, it turned out not to be the theory that was at fault but the description of the initial conditions, which did not include a consideration of the yet-to-be-discovered planet Neptune. A second example involves an argument by means of which the Danish astronomer Tycho Brahe claimed to have refuted the Copernician theory a few decades after the first publication of that theory, If the earth orbits the sun, Brahe argued, then the direction in which a fixed star is observed from earth should vary during the course of the year as the earth moves from one side of the sun to the other. But when Brahe tried to detect t his predicted parallax with his instruments, which were the most accurate and sensitive ones in existence at the time, he failed. This led Brahe to conclude that the Copernican theory was false. With hindsight, it can be appreciated that it was not the Copernican theory that was responsible for the faulty prediction, but one of Brahe's auxiliary a.s.sumptions. Brahe's estimate of the distance of the fixed stars was many times too small. When his estimate is replaced by a more realistic one, the predicted parallax turns out to be too small to be detectable by Brahe's instruments.

A third example is a hypothetical one devised by Imre Lakatos (1970, pp. 100-401). It reads as follows: The story is about an imaginary case of planetary misbehaviour. A physicist of the pre Einsteinian era takes Newton's mechanics and his law of gravitation, N, the accepted initial conditions, I, and calculates, with their help, the path of a newly discovered small planet,p. But the planet deviates from the calculated path. Does our Newtonian physicist consider that the deviation was forbidden by Newton's theory and therefore that, once established, it refutes the theory N? No. He suggests that there must be a hitherto unknown planet p', which perturbs the path of p. He calculates the ma.s.s, orbit, etc. of this hypothetical planet and then asks an experimental astronomer to test his hypothesis. The planet p' is so small that even the biggest available telescopes cannot possibly observe it; the experimental astronomer applies for a research grant to build yet a bigger one. In three years time, the new telescope is ready. Were the unknown planet p' to be discovered, it would be hailed as a new victory of Newtonian science. But it is not. Does our scientist abandon Newton's theory and his idea of the perturbing planet? No. He suggests that a cloud of cosmic dust hides the planet from us. He calculates the location and properties of this cloud and asks for a research grant to send up a satellite to test his calculations. Were the satellite's instruments (possibly new ones, based on a little-tested theory) to record the existence of the conjectural cloud, the result would be hailed as an outstanding victory for Newtonian science. But the cloud is not found. Does our scientist abandon Newton's theory, together with the idea of the perturbing planet and the idea of the cloud which hides it? No. He suggests that there is some magnetic field in that region of the universe which disturbed the instruments of the satellite. A new satellite is sent up. Were the magnetic field to be found, Newtonians would celebrate a sensational victory. But it is not. Is this regarded as a refutation of Newtonian science? No. Either yet another ingenious auxiliary hypothesis is proposed or ... the whole story is buried in the dusty volumes of periodicals and the story never mentioned again.

If this story is regarded as a plausible one, it ill.u.s.trates how a theory can always be protected from falsification by deflecting the falsification to some other part of the compfek web of a.s.sumptions.

Falsificationism inadequate on historical grounds.

An embarra.s.sing historical fact for falsificationists is that if their methodology had been strictly adhered to by scientists then those theories generally regarded as being among the best examples of scientific theories would never have been developed because they would have been rejected in their infancy. Given any example of a cla.s.sic scientific theory whether at the time of its first proposal or at a later date, it is possible to find observational claims that were generally accepted at the time and were considered to be inconsistent with the theory. Nevertheless, those theories were not rejected, and it is fortunate for science that they were not. Some historical examples to support my claim follow.

In the early years of its life, Newton's gravitional theory was falsified by observations of the moon's...o...b..t. It took almost fifty years to deflect this falsification on to causes other than Newton's theory. Later in its life, the same theory was known to be inconsistent with the details of the orbit of the planet Mercury although scientists did not abandon the theory for that reason. It turned out that it was never possible to explain away this falsification in a way that protected Newton's theory.

A second example concerns Bohr's theory of the atom, and is due to Lakatos (1970, pp. 140-54). Early versions of the theory were inconsistent with the observation that some matter is stable for a time that exceeds about 10-8 seconds. According to the theory negatively charged electrons within atoms...o...b..t around positively charged nuclei. But according to the cla.s.sical electromagnetic theory presupposed by Bohr's theory orbiting electrons should radiate. The radiation would result in an orbiting electron losing energy and collapsing into the nucleus. The quant.i.tative details of cla.s.sical electromagnetism yield an estimated time of about 10-8 seconds for this collapse to occur. Fortunately, Bohr persevered with his theory, in spite of this falsification.

A third example concerns the kinetic theory and has the advantage that the falsification of that theory at birth was explicitly acknowledged by its originator. When Maxwell (1965, vol. 1, p. 409) published the first details of the kinetic theory of gases in 1859, in that very same paper he acknowledged the fact that the theory was falsified by measurements on the specific heats of gases. Eighteen years later, commenting on the consequences of the kinetic theory, Maxwell (1877) wrote: Some of these, no doubt, are very satisfactory to us in our present state of opinion about the const.i.tution of bodies, but there are others which are likely to startle us out of our complacency and perhaps ultimately to drive us out of all the hypotheses in which we have hitherto found refuge into that thoroughly conscious ignorance which is a prelude to every real advance in knowledge. All the important developments within the kinetic theory took place after this falsification. Once again, it is fortunate that the theory was not abandoned in the face of falsifications by measurements of the specific heats of gases, as the naive falsificationist would be forced to insist.

A fourth example, the Copernican Revolution, will be outlined in more detail in the following section. This example emphasises the difficulties that arise for the falsificationist when the complexities of major theory changes are taken into account. The example also sets the scene for a discussion of some more recent and more adequate attempts to characterise the essence of science and its methods.

The Copernican Revolution.

It was generally accepted in mediaeval Europe that the earth lies at the centre of a finite universe and that the sun, planets and stars...o...b..t around it. The physics and cosmology that provided the framework in which this astronomy was set was basically that developed by Aristotle in the fourth century BC. In the second century AD, Ptolemy devised a detailed astir nomical system that specified the orbits of the moon, the sun and all the planets. In the early decades of the sixteenth century, Copernicus devised a new astronomy, an astronomy involving a moving earth, which challenged the Aristotelian and Ptolemaic sys tem. According to the Copernican view, the earth is not stationary at the centre of the universe but orbits the sun along with the planets. By the time Copernicus's idea had been substantiated, the Aristotelian world view had been replaced by the Newtonian one. The details of the story of this major theory change, a change that took place over one and a half centuries, do not lend support to the methodologies advocated by the inductivists and falsificationists, and indicate a need for a different, perhaps more complexly structured, account of science and its growth.

When Copernicus first published the details of his new astronomy, in 1543, there were many arguments that could be, and were, levelled against it. Relative to the scientific knowledge of the time, these arguments were sound ones and Copernicus could not satisfactorily defend his theory against them. In order to appreciate this situation, it is necessary to be familiar with some aspects of the Aristotelian world view on which the arguments against Copernicus were based. A very brief sketch of some of the relevant points follows.

The Aristotelian universe was divided into two distinct regions. The sub-lunar region was the inner region, extending from the central earth to just inside the moon's...o...b..t. The super-lunar region was the remainder of the finite universe, extending from the moon's...o...b..t to the sphere of the stars, which marked the outer boundary of the universe. Nothing existed beyond the outer sphere, not even s.p.a.ce. Unfilled s.p.a.ce is an impossibility in the Aristotelian system. All celestial objects in the super-lunar region were made of an incorruptible element called Ether. Ether possessed a natural propensity to move around the centre of the universe in perfect circles. This basic idea became modified and extended in Ptolemy's astronomy. Since observations of planetary positions at various times could not be reconciled with circular, earth-centred orbits, Ptolemy introduced further circles, called epicycles, into the system. Planets moved in circles, or epicycles, the centres of which moved in circles around the earth. The orbits could be further refined by adding epicycles to epicycles etc. in such a way that the resulting system was compatible with observations of planetary positions and capable of predicting future planetary positions.

In contrast to the orderly, regular, incorruptible character of the super-lunar region, the sub-lunar region was marked by change, growth and decay, generation and corruption. All substances in the sub-lunar region were mixtures of four elements, air, earth, fire and water, and the relative proportions of elements in a mixture determined the properties of the substance so const.i.tuted. Each element had a natural place in the universe. The natural place for earth was at the centre of the universe; for water, on the surface of the earth; for air, in the region immediately above the surface of the earth; and for fire, at the top of the atmosphere, close to the moon's...o...b..t. Consequently, each earthly object would have a natural place in the sub-lunar region depending on the relative proportion of the four elements that it contained. Stones, being mostly earth, have a natural place near the centre of the earth, whereas flames, being mostly fire, have a natural place near to the moon's...o...b..t, and so on. All objects have a propensity to move in straight lines, upwards or downwards, towards their natural place. Thus stones have a natural motion straight downwards, towards the centre of the earth, and flames have a natural motion straight upwards, away from the centre of the earth. All motions other than natural motions require a cause. For instance, arrows need to be propelled by a bow and chariots need to be drawn by horses.

These, then, are the bare bones of the Aristotelian mechanics and cosmology that were presupposed by contemporaries of Copernicus, and which were utilised in arguments against a moving earth. Let us look at some of the forceful arguments against the Copernican system. Perhaps the argument that const.i.tuted the mdst serious threat to Copernicus was the so-called tower argument. It runs as follows. If the earth spins on its axis, as Copernicus had it, then any point on the earth's surface will move a considerable distance in a second. If a stone is dropped from the top of a tower erected on the moving earth, it will execute its natural motion and fall towards the centre of the earth. While it is doing so the tower will be sharing the motion of the earth, due to its spinning. Consequently, by the time the stone reaches the surface of the earth the tower will have moved around from the position it occupied at the beginning of the stone's downward journey. The stone should therefore strike the ground some distance from the foot of the tower. But this does not happen in practice. The stone strikes the ground at the base of the tower. It follows that the earth cannot be spinning and that Copernicus's theory is false.

Another mechanical argument against Copernicus concerns loose objects such as stones and philosophers resting on the surface of the earth. If the earth spins, why are such objects not flung from the earth's surface, as stones would be flung from the rim of a rotating wheel? And if the earth, as well as spinning, moves bodily around the sun, why doesn't it leave the moon behind?

Some arguments against Copernicus based on astronomical considerations have been mentioned earlier in this book. They involved the absence of parallax in the observed positions of the stars and the fact that Mars and Venus, as viewed by the naked eye, do not change size appreciably during the course of the year.

Because of the arguments I have mentioned, and others like them, the supporters of the Copernican theory were faced with serious difficulties. Copernicus himself was very much immersed in Aristotelian metaphysics and had no adequate response to them.

In view of the strength of the case against Copernicus, it might well be asked just what there was to be said in favour of the Copernican theory in 1543. The answer is, ”not very much”. The main attraction of the Copernican theory lay in the neat way it explained a number of features of planetary motion, which could be explained in the rival Ptolemaic theory only in an unattractive, artificial way. The features are the retrograde motion of the planets and the fact that, unlike the other planets, Mercury and Venus always remain in the proximity of the sun. A planet at regular intervals regresses, that is, stops its westward motion among the stars (as viewed from earth) and for a short time retraces its path eastward before continuing its journey westward once again. In the Ptolemaic system, retrograde motion was explained by the somewhat ad hoc manoeuvre of adding epicycles especially designed for the purpose. In the Copernican system, no such artificial move is necessary. Retrograde motion is a natural consequence of the fact that the earth and the planets together orbit the sun against the background of the fixed stars. Similar remarks apply to the problem of the constant proximity of the sun, Mercury and Venus. This is a natural consequence of the Copernican system once it is established that the orbits of Mercury and Venus are inside that of the earth. In the Ptolemaic system, the orbits of the sun, Mercury and Venus have to be artificially linked together to achieve the required result.

Thus there were some mathematical features of the Copernican theory that were in its favour. Apart from these, the two rival systems were more or less on a par as far as simplicity and accord with observations of planetary positions are concerned. Circular sun-centred orbits cannot be reconciled with observation, so that Copernicus, like Ptolemy, needed to add epicycles, and the total number of epicycles needed to produce orbits in accord with known observations was about the same for the two systems. In 1543 the arguments from mathematical simplicity that worked in favour of Copernicus could not be regarded as an adequate counter to the mechanical and astronomical arguments that worked against him. Nevertheless, a number of mathematically capable natural philosophers were to be attracted to the Copernican system, and their efforts to defend it became increasingly successful over the next hundred years or so.

The person who contributed most significantly to the defence of the Copernican system was Galileo. He did so in two ways. First, he used a telescope to observe the heavens, and in so doing he transformed the observational data that the Copernican theory was required to explain. Second, he devised the beginnings of a new mechanics that was to replace Aristotelian mechanics and with reference to which the mechanical arguments against Copernicus were defused.

When, in 1609, Galileo constructed his first telescopes and trained them on the heavens, he made dramatic discoveries. He saw that there were many stars invisible to the naked eye. He saw that Jupiter has moons and he saw that the surface of the earth's moon is covered with mountains and craters. He also observed that the apparent size of Mars and Venus, as viewed through the telescope, changed in the way predicted by the Copernican system. Later, Galileo was to confirm that Venus has phases like the moon, a fact that could be straightforwardly accommodated into the Copernican, but not the Ptolemaic, system. The moons of Jupiter defused the Aristotelian argument against Copernicus based on the fact that the moon stays with an allegedly moving earth. For now Aristotelians were faced with the same problem with respect to Jupiter and its moons. The earthlike surface of the moon undermined the Aristotelian distinction between the perfect, incorruptible heavens and the changing, corruptible earth. The discovery of the phases of Venus marked a success for the Copernicans and a new problem for the Ptolemaics. It is undeniable that once the observations made by Galileo through his telescope are accepted, the difficulties facing the Copernican theory are diminished.

The foregoing remarks on Galileo and the telescope raise a serious epistemological problem. Why should observations through a telescope be preferred to naked-eye observations? One answer to this question might utilise an optical theory of the telescope that explains its magnifying properties and that also gives an account of the various aberrations to which we can expect telescopic images to be subject. But Galileo himself did not utilise an optical theory for that purpose. The first optical theory capable of giving support in this direction was devised by Galileo's contemporary, Kepler, early in the sixteenth century, and this theory was improved and augmented in later decades. A second way of facing our question concerning the superiority of telescopic to naked-eye observations is to demonstrate the effectiveness of the telescope in a practical way, by focusing it on distant towers, s.h.i.+ps, etc. and demonstrating how the instrument magnifies and renders objects more distinctly visible. However, there is a difficulty with this kind of justification of the use of the telescope in astronomy. When terrestrial objects are viewed through a telescope, it is possible to separate the viewed object from aberrations contributed by the telescope because of the observer's familiarity with what a tower, a s.h.i.+p, etc. look like. This does not apply when an observer searches the heavens for he knows not what. It is significant in this respect that Galileo's drawing of the moon's surface as he saw it through a telescope contains some craters that do not in fact exist there. Presumably those ”craters” were aberrations arising from the functioning of Galileo's far-from-perfect telescopes. Enough has been said in this paragraph to indicate that the justification of telescopic observations was no simple, straightforward matter. Those adversaries of Galileo who queried his findings were not all stupid, stubborn reactionaries. Justifications were forthcoming, and became more and more adequate as better and better telescopes were constructed and as optical theories of their functioning were developed. But all this took time.