Abstract
The main aims of this book are to provide a new way of looking at the Realism/anti-Realism issue and to develop a method for deciding whether it is Realism or anti-Realism that is correct in a given domain. In the previous chapter it was suggested that we can have empirical evidence that a Realist attitude is appropriate towards a particular scientific theory if that theory enjoys novel predictive success under a range of conditions of the Intentional. The idea was that if a theory enjoys novel predictive success then we have evidence that it is true tout court, and that if it enjoys novel predictive success in a wide range of conditions of the Intentional, then we have evidence that it is true in all conditions of the Intentional, that is, that it is true independently of the Intentional.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Notes
In Leplin, J. (ed), Scientific Realism ( University of California Press, Berkeley, 1984 ).
The core of the phlogiston theory was that the combustion of a substance was the expulsion of ‘phlogiston’ from that substance. The first version of this theory was formulated about 1750, and the theory persisted in one form or another until the mid to late 1790s. The version of the theory with which we will be concerned was formulated about 1766 as a result of the work of Henry Cavendish and Joseph Priestley. A central feature of this version of the theory is that it identified phlogiston with ‘inflammable air’, or what we would now call ’hydrogen’. The reason for this identification was an observed difference between the behaviour of metals and the calxes of those metals when placed in acid. It was thought that the formation of the calx of a metal (what we would now call an oxide of the metal) upon heating was due to the expulsion of phlogiston from the metal. On this view, a calx could be regarded as a metal that had been deprived of phlogiston. It was also known that when a metal was placed in an acid a salt was formed and that when a calx of the same type of metal was placed in acid, the same salt was formed. But there was also a difference: when the metal was placed in the acid, an inflammable gas (’inflammable air’) was given off, but when the calx was placed in the acid, no gas was given off. Otherwise the products of immersing a metal and its calx in an acid were the same. Since a calx was thought to be a metal deprived of phlogiston, it was natural to suppose that the gas given off (inflammable air) was none other than phlogiston. This suggested a novel prediction: if inflammable air was phlogiston, and a calx was metal deprived of phlogiston, it ought to be possible to produce a metal by bringing together its calx and inflammable air. Joseph Priestly succeeded in doing this in an experiment performed in 1783. He heated lead calx in inflammable air, by focussing sunlight onto it with a lens. The inflammable air was confined by water. The result was the lead calx was transformed to metallic lead, and there was a reduction in the volume of inflammable air, since the confining water was perceived to rise. This is very plausibly an example of strong predictive success, yet it is evident that the account given by the phlogiston theory of what is going on at the theoretical level is completely off the track How great a difficulty does this create for the claim that the explanation we ought to accept of the success of currently accepted (unfalsified) theories is that certain key sentences in those theories are true, or approximately true? First, we should note that it is only claimed that the inference from the novel predictive success of unfalsified theories to their partial truth is a good inference, or that it is better than the inference to any other explanation of success. But is not claimed that the inference is deductively valid. Of course, a single counter-example to a mode of inference is deadly to the claim that it is deductively valid. But if it merely claimed that a mode of inference is powerful, although short of deductively valid, then an occasional counter-example is only what is to be expected. So, the case of phlogiston theory is, I think, not damaging to the claim that it is reasonable to infer that a key sentence of a successful theory is true. There is, I think, another reason why phlogiston theory does not cast into doubt the reasonableness of the inference. This can be brought by examining, in detail just why phlogiston theory was strongly successful. The phlogiston theory identified phlogiston with `inflammable air’ because inflammable air was expelled from a metal placed in acid but not from a calx placed in an acid. But, of course, by our lights the reason why inflammable air was expelled from the metal in acid was not because it was present in the metal and not the calx, rather the inflammable air originated from the acid,and was not expelled from the calx in acid because it combined with the oxygen in the calx to form water. The water thus formed was not noticed because the reaction took place in water. That is, inflammable air was not produced from the calx in acid, not because it was not present, but rather because it combined with something (oxygen) present in the calx but not the metal. The prediction that a calx and inflammable air would form a metal was confirmed not because a metal is a combination of a calx and inflammable air, but because, again, the inflammable air and a component of the calx combine to yield a product (water) leaving the metal. The water was not detected because the experiment was confined by water and the water formed by the reaction simply mixed in, un-noticed, with the confining water. In summary we can say (a) The phlogiston theory of combustion was a success in its prediction that heating a calx and inflammable air together would produce a metal, because of a fortunate coincidence. The gas that was expelled from metal in acid, but not from calx in acid, was not expelled from the calx in acid, not because it was not present, but rather because in that case it combined with other reagents to yield the (undetectable) product water; moreover, by a lucky coincidence, the propensity of the gas to form water was the very same propensity as the one also responsible for the creation of a metal when a calx is heated in inflammable air. The propensity is, according to our theory, the propensity of hydrogen to combine with oxygen to form water. That is, we can say that phlogiston theory was successful because the feature of `inflammable air’ which prevented it from appearing when calx is immersed in acid, also just happened to be also a feature that would cause metal to be formed when calx and inflammable air are heated together. (b) In both experiments (calx versus metal in acid and calx and inflammable air heated together) the phlogiston theorists’ interpretation was observationally tenable because, it just so happened, the water produced from the reactions of the calx and the inflammable air was undetectable. In the first case it is undetectable because the reaction takes place in water, in the second because the atmosphere in which the reaction takes place (inflammable air) is confined by water. Let us now consider the question of to what extent the case of phlogiston creates a difficulty for the claim that novel success is best explained by truth. It will be argued that it creates very little difficulty. The case would create a difficulty for this claim if it revealed a nothighly-improbable mechanism whereby a theory that is totally off the track could come to have strong predictive success. If there was such a mechanism then the inference from `Unfalsified theory T has novel predictive success’ to `Sentences crucial in the derivation of T are true or close to the truth’ would be cast into doubt, since there would be some credibility to the claim that the success of T was due to a mechanism very different from the one T asserts to be the case. But the case of phlogiston theory does not give us reason to believe this. Phlogiston theory was strongly predictively successful and empirically adequate only because of a series of improbable circumstances. As noted above, its prediction that a calx heated in inflammable air was correct only because that tendency of inflammable air which prevented it from being expelled when a calx was placed in acid was the very same tendency that would cause metal to be formed when a calx was heated in it. The identity of these tendencies is not something predicted by the phlogiston theory; from the standpoint of that theory it is just `pure luck’ that they are the same. Moreover, phlogiston theory was able to provide observationally adequate accounts of what took place both with a calx in acid and a calx heated in inflammable air over water only because in both cases the water produced, but not predicted by phlogiston theory, could not easily be observed and so was not observed. The novel predictive success of phlogiston was due, therefore, to a series of lucky chances. So this case of a false theory having novel predictive success does not cast into doubt the claim that the inference from the novel predictive success of a theory to its probable truth is a reasonable, though fallible, inference.
One familiar argument for the Strict Empiricist version of UTD is that if T is an empirically adequate hypothesis, and pl, p2, p„…, neither add to nor detract from T’s empirical adequacy, then Tandp„ T and p2, Tandp, etc. will all be empirically adequate, and e mpirically equivalent to T. This argument is discussed in the next paragraph.
See, e.g., W. H. Newton-Smith ‘The Underdetermination of Theory by Data’ in R. Hilpinen (ed.) Rationality in Science pp.91–110, esp. p.96. I have also argued that there is good reason to believe that a theory that has simplicity, or lack of ad hotness,is more likely to be true than an empirically equivalent theory that lacks these properties. (Science and the Theory of Rationality,ch.3.)
It would appear that there could be two types of reason for accepting the Scientific Methodology UT: a general conceptual argument, or evidence from the history of science that there have in fact been rival, empirically adequate theories judged to be equally simple, elegant, etc. But I do not think either of these approaches succeeds in making the Scientific Method UT plausible. The only philosopher that I know of who has produced even a sketch of a general argument is Quine. In his Word and Object (p.23) he rejects the idea that scientific method would yield a single best theory, claiming that ‘It seems likelier, if only on account of symmetries or dualities, that countless alternative theories would be tied for first place’. But in this passage Quine gives no reason for supposing that these countless, alternative theories will indeed be genuinely different theories, rather than merely different formulations of the same theory. His reference to symmetries or dualities suggests that these may be in some sense ‘reflections’ of one another or they may exploit particle/field duality. But if that is so, it is rather unclear whether they should be regarded as genuinely different theories at all. It might be noted here that in a later paper ’On Empirically Equivalent Systems of the World’, Quine adopts a very liberal criterion of theory identity, according to which empirically equivalent theories are to be counted as identical if they are mutually relatively interpretable. It is doubtful whether theories that were in some sense mutual reflections, or which were particle and field formulations of a single set of laws, would count as genuinely different on such a criterion of theory identity. One difficulty with attempting to establish the Scientific Methodology UT from an investigation into the history of science is that there is little agreement on how the historical evidence is to be interpreted. Philosophers of science disagree on what is to count as factors that increase the extent to which it is rational to accept a theory, and other factors which, although they might have a causal role in the acceptance of a theory, do not make it any more rational. For example, do appeals to ’common-sense’, or to metaphysical beliefs, increase the degree of rational support for a theory? Advocates of an austere epistemology in science, such as for example, Clark Glymour, would, I venture to suggest, say that they do not. But advocates of a more liberal epistemology (L. Laudan, A. E. Burtt) may say that they do. The lack of consensus amongst philosophers of science on what constitutes a reason in favour of the acceptance of a theory makes the Scientific Method UT thesis difficult to assess. It should also be noted that, although in some cases two theories may seem on an intuitive assessment of simplicity to be more or less as simple as each other, the use of a more refined concept of simplicity may show one to be more simple. I have attempted to show this in Chapter VI of my Science and the Theory of Rationality.
For our purposes we need not consider whether the questionable step is from 2 to 3 or from 3 to 4. If the superior degree of evidential support of a strongly predictively successful theory is regarded as a form of empirical support, then it is the step from 2 to 3 that becomes invalid. But if two theories that explain the same regularities are regarded as having the same empirical support, with the additional support enjoyed by a strongly predictively successful theory being regarded as a form of non-empirical support, then it is the step from 3 to 4 that is blocked.
See my Science and the Theory of Rationality,pp.40–42.
See van Fraassen’s The Scientific Image,especially pp.39–40.
See Truth and Other Enigmas,p.152. Dummett says the phenomenon of what we are here calling ‘novel’ predictive success does not conflict essentially with an anti-Realist view of scientific theories. Dummett’s use of the word ‘essentially’ here suggests that in order to have good reason to abandon anti-Realism in some domain it is necessary to show it to be actually logically incoherent in that domain. It is, of course, far beyond the scope of the present work to address itself to the issues raised by this matter. It will simply be remarked that the present author favours a broadly Quinean view of the relation between philosophy and other disciplines. On such a view, philosophy is not an a priori activity, and neither is it the case that the only grounds we can have for rejecting a philosophical theory is by showing it to be inconsistent.
See Arthur Fine, ‘Unnatural Attitudes: Realist and Instrumentalist Attachments to Science’ in Mind vol XCV (1986), pp.149–179, esp. p.154.
See B. Barnes, Scientific Knowledge and Sociological Theory (Routledge and Kegan Paul, 1974 )
See Laudan, op. cit. p.101 in Science and Reality, edited by J. T. Cushing, C. F. Delaney and Gary M. Gutting (University of Notre Dame Press, 1984 ).
See S. Leeds ‘Theories of Reference and Truth’ in Erkenntnis 13, pp.111–129.
As reported by R. Boyd in his paper ‘On the Current Status of the Issue of Scientific Realism’ in Erkenntnis 19 (1983), pp.45–90 especially p.73.
Consider, for example, Fresnel’s prediction of a white spot in the middle of a circular shadow. Presumably, it is not the case that no one had before observed a white spot in a circular shadow; what was novel about Fresnel’s prediction is that such a spot should be produced by a particular experimental set-up.
See my Science and the Theory of Rationality,especially pp.60–65.
Watkins asks us to imagine a theory T from which we have derived an experimental generalisation g. This experimental generalisation is, we can suppose, of the form a = f(ß, y). We then consult the records of all the tests that have been performed on T (which Watkins imagines to be stored in a computer) and ask ‘Has any test that has been performed a test of g?’ A test would be a test of g iff in it values of a, ß and y had been ascertained. This would count as a test of g since the value of f(ß, y) determined by the values of ß and y runs the risk of being incompatible with the measured value of a, and hence runs the risk of falsifying a = f(ß, y); that is, it runs the risk of falsifying g. If no such test has been carried out, then g is an untested prediction of T, and hence, according to Watkins, the experimental generalisation g is an empirically novel consequence of T. (See Watkins, Science and Scepticism,p.295.) Watkins is here concerned to explicate a notion quite different from the one with which we are here concerned.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 1997 Springer Science+Business Media Dordrecht
About this chapter
Cite this chapter
Wright, J. (1997). Does Novel Success Need to be Explained by Truth?. In: Realism and Explanatory Priority. Philosophical Studies Series, vol 71. Springer, Dordrecht. https://doi.org/10.1007/978-94-017-2844-7_7
Download citation
DOI: https://doi.org/10.1007/978-94-017-2844-7_7
Publisher Name: Springer, Dordrecht
Print ISBN: 978-90-481-4823-3
Online ISBN: 978-94-017-2844-7
eBook Packages: Springer Book Archive