CHAPTER 14:.

Why should the world obey laws?.

Introduction.

In the foregoing chapters we have been concerned with epis temological questions, that is, questions concerning how scientific knowledge is vindicated by appeal to evidence, and the nature of that evidence. In this and the next chapter we turn to ontological questions, questions about the kinds of things there are in the world. What kinds of ent.i.ties are a.s.sumed or shown to exist in the world by modern science? Part of an answer to that question has been taken for granted in this book up until now. It has been taken for granted that there are such things as laws which govern the behaviour of the world and which it is the business of science to discover. This chapter is concerned with what kinds of ent.i.ties these laws are.

The idea that the world is governed by laws that it is the business of science to discover is commonplace. However, the question of what this idea amounts to is far from being unproblematic. A fundamental problem was highlighted by Robert Boyle in the seventeenth century. The notion of a law originates in the social sphere where it makes straightforward sense. Society's laws are obeyed or not obeyed by individuals who can comprehend the laws and the consequences of violating them. But once laws are understood in this natural way, how can it be said that material systems in nature obey laws? For they can hardly be said to be in a position to comprehend the laws they are meant to obey, and, in any case, a fundamental law as it applies in science is supposed to be exceptionless, so there is no correlate to an individual's violating a social law and taking the consequences. What is it that makes matter conform to laws? This is a reasonable and straightforward question, it would appear, and yet it is not one that is easily answered. I take it that Boyle's answer, namely that G.o.d makes matter behave in accordance with the laws He has ordained, leaves a lot to be desired from a modern point of view. Let us see if we can do better.

Laws as regularities.

One common response to the question "What makes matter behave in accordance with laws?" is to deny its legitimacy. The line of reasoning involved here was forcefully expressed by the philosopher David Hume, and has been influential ever since. From the Humean standpoint it is a mistake to a.s.sume that lawlike behaviour is caused by anything. Indeed, the whole idea of causation in nature is brought into question. The reasoning goes like this. When, for example, two billiard b.a.l.l.s collide, we can observe their motions immediately before and immediately after collision, and we may be able to discern a regular way in which the speeds after impact are connected to the speeds before impact, but what we never see is something in addition to this which can be identified with the causal effect of the one ball on the other. From this point of view causation is nothing other than regular connection, and laws take the form "Events of type A are invariably accompanied or followed by events of type B". For instance, Galileo's law of fall would take the form "Whenever a heavy object is released near the earth's surface it falls to the ground with a uniform acceleration". This is the so-called regularity view of laws. Nothing makes matter behave in accordance with laws because laws are nothing other than de facto regularities between events.

A standard, and telling, set of objections to the regularity view of laws involves the claim that it does not distinguish between accidental and lawlike regularities. Popper gives the example "no moa lives beyond fifty years' as an example. It may well be the case that no moa, a species now extinct, ever lived beyond fifty years, but some might well have done so had the environmental conditions been more favourable, and for this reason we are inclined to discount the generalisation as a law of nature. But it qualifies as a law on the ground that it is an exceptionless regularity. It may well be the case that whenever the factory hooter sounds at the end of the working day in Manchester the workers down tools in London, but even if there are no exceptions to this generalisation, it hardly qualifies as a law of nature. Examples of this kind abound, and they suggest that there is something more to a law of nature than mere regularity. Another difficulty with the regularity view is that it faiM to identify the direction of causal dependency. There is a regular connection between instances of smoking and lung cancer, but this is because smoking causes lung cancer, not the reverse. That is why we can hope to decrease the occurrence of cancer by eliminating smoking, but cannot hope to combat smoking by finding a cure for cancer. A regularity exhibited by events is not a sufficient condition for the regularity to const.i.tute a law for there is more to lawlike behaviour than mere regularity.

Apart from difficulties with the idea that regularities are a sufficient condition for a law, straightforward considerations about laws as they figure in science strongly suggest that regularity is not a necessary condition either. If the view that laws describe exceptionless regular connections between events is taken seriously, then none of the claims typically taken to be scientific laws would qualify. Galileo's law of fall, mentioned above, is a case in point. Autumn leaves rarely fall to the ground with a uniform acceleration. On an unqualified regularity view this would make the law false. In a similar fashion Archimedes' principle, which claims in part that objects denser than water sink, is refuted by floating needles. If laws are taken to be exceptionless regularities, then it is very difficult to find a serious candidate for a law for want of the appropriate regularities. More to the point, most if not all of the generalities taken to be laws within science fail to qualify.

From the point of view of scientific practice, and commonsense for that matter, there is a ready response to these observations. After all, it is well understood why Autumn leaves do not fall to the ground in a regular fashion. They are influenced by draughts and air-resistance which act as a disturbing influence, just as the sinking of a needle can be inhibited by surface tension. It is because physical processes are hindered by disturbing influences that physical laws characterising those processes need to be tested in contrived experimental circ.u.mstances in which the hindrances are eliminated or controlled. The regularities of relevance to science, and which are indications of lawlike behaviour, are typically the hard-won results of detailed experimentation. Think, for example, of the lengths to which Henry Cavendish had to go to get attracting spheres to exhibit the inverse square law of attraction and how J. J. Thomson eventually succeeded, where Hertz had failed, to exhibit the regular deflection of moving electrons in an electric field.

An obvious response that the defender of the regularity view of laws can give to these observations is to restate that view in a conditional form. Laws can be formulated in the form "events of type A are regularly followed, or accompanied, by events of type B provided disturbing factors are not present". So Galileo's law of fall becomes "heavy objects fall to the ground with a uniform acceleration provided they do not encounter a variable resistance or are not deflected by winds or other disturbing factors". The phrase "other disturbing factors" is indicative of a general problem concerning how a precise statement of the conditions to be satisfied for a law to apply can be formulated. But I will leave that difficulty aside, because I suggest there is a much more fundamental one facing the regularity view here. If we accept the characterisation of laws as regularities stated in conditional form, then we must accept that laws only apply when those conditions are satisfied. Since the satisfaction of the appropriate conditions will normally only obtain in special experimental set-ups, we are forced to conclude that scientific laws generally apply only within experimental situations and not outside of them. Galileo's law of fall will be considered to apply only when heavy objects are dropped in situations where air resistance and the like have been removed. So Autumn leaves are not subject to Galileo's law of fall, according to this revised version of the regularity view. Does this not clash with our intuition? Do we not wish to say that an Autumn leaf is governed by the law of fall, but is also governed by the laws governing air-resistance and aerodynamics as well, so that the resulting fall is the complicated result of the various laws acting in conjunction? Because the regularity view, in its conditional form, restricts the applicability of laws to those experimental situations where the appropriate conditions are met, it is incapable of saying anything about what happens outside of those conditions. On this view, science is incapable of saying why Autumn leaves usually end up on the ground!

The difficulty here echoes a problem which arises if the new experimentalism is taken as exhausting what can be said of scientific knowledge. For, as we saw in the previous chapter, although it may well be the case that the new experimentalism can capture a strong sense in which the progress of science can be understood as a steady acc.u.mulation of experimental knowledge, to leave it at that leaves us with no account of how knowledge arrived at inside experimental situations can be transported outside of those situations and used elsewhere. How are we to explain the engineer's use of physics, the use of radioactive dating in historical geology or the application of Newton's theory to the motion of comets? If scientific laws are a.s.sumed to apply outside, as well as inside, of experimental situations then laws cannot be identified with the regularities that are achievable in experimental situations. The regularity view of laws will not do.

Laws as characterisations of powers or dispositions.

There is a straightforward way out of the problems with the idea of a law that we have so far discussed. It involves taking seriously what is implicit in much commonsense as well as science, namely that the material world is active. Things happen in the wend of their own accord, and they happen because ent.i.ties in the world possess the capacity or power or disposition or tendency to act or behave in the way that they do. b.a.l.l.s bounce because they are elastic. Warnings on containers that declare the contents to be poisonous or inflammable or explosive tell us what the contents are capable of doing or how they are inclined to act. Specifying the ma.s.s and charge of an electron indicates how it will respond to electric and magnetic fields. An important element of what a thing is, is what it is capable of doing or becoming. We need to characterise things in terms of their potential as well as their actual being, as Aristotle correctly observed Just as the ability to grow into an oak tree is an important part of what it is to be an acorn, so the capacity to attract unlike and repel like charges, and to radiate when accelerating, is an important part of what it is to be an electron. We experiment on systems to find out how they are disposed to behave.

Once we admit such things as dispositions, tendencies, powers and capacities into our characterisation of material systems, then laws of nature can be taken as characterising those dispositions, tendencies, powers or capacities. Galileo's law of fall describes the disposition heavy objects possess to fall to the ground with a uniform acceleration and Newton's law of gravitation describes the power of attraction between ma.s.sive bodies. Once we interpret laws in this way, we need, no longer expect laws to describe sequences of happenings in the world because those happenings will typically be the result of several dispositions, tendencies, powers or capacities acting in conjunction in complex ways. The fact that the tendency of a leaf to fall in accordance with Galileo's law is swamped by the effect of the wind is no reason in itself to doubt that that tendency continued to act on the leaf in accordance with the law. From this point of view, we can readily understand why experiment is necessary to glean information relevant for the identification of a law. The tendencies corresponding to the law under investigation need to be separated from other tendencies, and this separation requires the appropriate practical intervention to bring it about. Given the irregularities of ocean beds and the attraction of the sun and planets as well as the moon, we cannot hope to arrive at a precise account of the tides from Newton's theory plus initial conditions. Nevertheless, gravity is the major cause of the tides and there are appropriate experiments for identifying the law of gravity.

From the point of view I am advocating, causes and laws are intimately linked. Events are caused through the action of particulars that possess the power to act as causes. The gravitational attraction of the moon is the main cause of the tides, charged particles cause the ionisation responsible for the tracks in a cloud chamber and oscillating charges cause the radio waves emitted from a transmitter. Descriptions of the mode of acting of the active powers involved in such cases const.i.tute the laws of nature. The inverse square law of gravitation describes quant.i.tatively the power to attract possessed by ma.s.sive bodies, and the laws of cla.s.sical electromagnetic theory describe, among other things, the capacity of charged bodies to attract and radiate. It is the active powers at work in nature that makes laws true when they are true. We thus have a ready answer to Boyle's question. It is the powers and capacities possessed by particulars and operative when particulars interact that compel those particulars to behave in accordance with laws. Lawlike behaviour is brought about by efficient causation. Boyle faced the problem he did with laws, and needed to invoke G.o.d, just because he declined to ascribe dispositional properties to matter.

The majority of philosophers seem reluctant to accept an ontology which includes dispositions or powers as primitive. I do not understand their reluctance. Perhaps the reasons are in part historical. Powers were given a bad name by the mystical and obscure way they were employed in the magical tradition in the Renaissance, and they are alleged to have been exploited by the Aristotelians in a cavalier way under the guise of forms. Boyle's rejection of active properties in his mechanical philosophy can be seen as a reaction, and perhaps an overreaction, to the excesses of those traditions, as well as being motivated by theological concerns. However, there need be nothing mysterious or epistemologically suspect about invoking powers, tendencies and the like. Claims concerning them can be subject to stringent empirical tests to as great an extent as any other kind of claim. What is more, however much philosophers may be averse to dispositional properties, scientists systematically invoke them and their work would be incapacitated without them. It is significant to note in this respect that Boyle, in his experimental science as opposed to his mechanical philosophy freely employed dispositional properties such as acidity and the spring of the air. Elasticity in various forms was an embarra.s.sment to the seventeenth-century mechanical philosophers. Hobbes complained that Boyle's attribution of elasticity to air was equivalent to the admission that air could move itself. Boyle and other seventeenth century scientists continued to employ the concept of elasticity, and never succeeded in explaining it away by reference to non-dispositional properties. Nor has anyone succeeded since. I do not understand what grounds philosophers have for questioning, or feeling the need to explain away, this common, indeed ubiquitous, usage by scientists of dispositional properties.

The view that laws characterise the dispositions, powers, capacities or tendencies of things has the merit that it acknowledges at the outset what is implicit in all scientific practice, namely that nature is active. It makes it clear what makes systems behave in accordance with laws, and it links laws with causation in a natural way. It also offers a ready solution to the problem, encountered in the previous chapter, concerning the transportability of knowledge acquired in experimental situations beyond those situations. Oncee the a.s.sumption is made that ent.i.ties in the world are what they ire by virtue of the powers and capacities that they possess, .and I claim that that a.s.sumption is implicit in scientific practice as well as everyday life, then the laws describing those powers and capacities, identified in experimental situations, can be presumed to apply outside of those situations too. Nevertheless, I cannot leave things here with a good conscience, because there are important laws of science that are difficult to fit into this scheme.

Thermodynamic and conservation laws.

Let us refer to the view I have outlined and defended in the previous paragraph, which understands laws as characterising causal powers, as the causal view of laws. There are important laws in physics that do not fit well into this scheme. The first and second laws of thermodynamics do not and nor do a range of conservation laws in fundamental particle physics. The first law of thermodynamics a.s.serts that the energy of an isolated system is constant. The second law, which a.s.serts that the entropy of an isolated system cannot decrease, has consequences such as ensuring that heat flows from hot to cold bodies and not the other way round and ruling out the possibility of extracting heat energy from the sea and putting it to useful work, where the only price paid for the work is a decrease in temperature of the sea. A machine that succeeded in doing this would be a perpetual motion machine of the second kind, distinct from a machine that results in a net increase in energy which is a perpetual motion machine of the first kind. The first law of thermodynamics rules out perpetual motion machines of the first kind and the second law rules out perpetual motion machines of the second kind. These quite general laws have consequences for the behaviour of physical systems, and can be used to predict their behaviour, quite independently of the details of the causal processes at work. That is why it is not possible to construe these laws as causal laws.

Let me give an example that ill.u.s.trates my point. If ice is subjected to pressures higher than normal atmospheric pressure its melting point is lowered. This is why a wire from which weights are suspended will cut its way through a block ofice. The explanation of this at the molecular level is far from straightforward and a precise, detailed account is probably not available. Since pressure tends to push molecules closer together, one might expect the forces of attraction between them to increase under such circ.u.mstances, leading to an increase in the thermal energy necessary to drag them apart and thus to an elevation in melting point. This is precisely what happens in a typical solid near melting point. But ice is not a typical solid. The water molecules in ice are rather loosely packed, more so than they are in the liquid state, which is why ice is less dense than water. (This is just as well, otherwise lakes and rivers would freeze from the bottom up, and would freeze in their entirety in periods of prolonged cold, thus eliminating fish and anything evolved from fish as a viable life form.) If the molecules in ice are forced closer together than normal, the force between them decreases, so less thermal energy is needed to separate them, and the melting point falls. The precise way in which the forces depend on molecular positions is complicated, depending on fine quantum mechanical detail involving exchange as well as Coulomb forces, and is not known with precision. Given the above complications, it may come as a surprise that James Thomson was able to predict the depression of the freezing point of water with pressure in 1849 thereby antic.i.p.ating the empirical discovery of the phenomenon. All he needed for his derivation were the laws of thermodynamics plus the empirically known fact that water is denser than ice. Thomson devised, in thought, a cyclic process that involved extracting heat from water at 0C and converting it into ice at 0C. It seemed as if this engine provided a means of extracting heat from water and converting all of it into the work done by the expansion involved, thus comprising a perpetual motion machine of the second kind, ruled out by the second law of thermodynamics. Thomson realised that this unacceptable conclusion could be blocked by a.s.suming the freezing point to be lowered by an increase in pressure. The feature of this case that I wish to highlight is that Thomson's prediction was made in ignorance of the details of the causal process at the molecular level. A characteristic feature, and a major strength, of thermodynamics is that it applies at the macroscopic level whatever the details of the underlying causal process. It is precisely this feature of the laws of thermodynamics that prevents them being construed as causal laws.

The difficulties for the causal view do not stop here. The behaviour of a mechanical system can be understood and predicted by specifying the forces on each component of the system and using Newton's laws to trace the development of the system. Within this approach Newton's laws can readily be interpreted as causal laws describing the disposition of objects to exert and respond to specified forces. However, this is not the only way of dealing with mechanical systems. The laws of mechanics can also be written in a form that takes energy, rather than forces, as the starting point. In the Hamiltonian and Lagrangian formulations of mechanics, where this approach is adopted, what is required is expressions for the potential and kinetic energy of a system as a function of whatever coordinates are necessary to fix them. The evolution of a system can then be completely specified by feeding these expressions into the Hamiltonian or Lagrangian equations of motion. This can be done without a detailed knowledge of the causal processes at work.

James Clerk Maxwell (1965, vol. 2, pp. 783 4), who at- tempted to cast his electromagnetic theory in Lagrangian form, ill.u.s.trated this point in a characteristically vivid way. We imagine a belfry in which a complicated piece of machinery is driven by bell ropes that drop to the bell ringers room below. We a.s.sume the number of ropes to be equal to the number of degrees of freedom of the system. The potential and kinetic energy of the system as a function of the position and velocity of the ropes can be determined by experiments done with the ropes. Once we have these functions we can write down Lagrange's equations for the system. It is then possible, given the positions and velocities of the ropes at any one instant, to derive their positions and velocities at any other instant. We can do this without needing to know the details of the causal story of what is happening in the belfry. La-grange's equations do not state causal laws.

It might be objected that these observations about the Lagrangian formulation of mechanics do not const.i.tute a serious counter-example to the causal view of laws. It might be pointed out, for example, that, although a Lagrangian treatment of the mechanism in the belfry can work as well as it does by ignoring the detailed causal story of the mechanism in the belfry there is such a story to be had that can be formulated in Newtonian, and hence causal, terms once appropriate empirical access to the belfry is gained. After all, it might be observed, Lagrange's equations can be derived from Newton's.

This last claim is no longer true (if it ever was). In modern physics Lagrange's equations are interpreted in a more general way than the version of those equations that can be derived from Newton's laws. The energies involved are interpreted in a general way that includes all kinds of energy not just energy arising from the motion of ma.s.sive bodies under the influence of forces. For instance, the Lagrangian formulation can accommodate electromagnetic energy, which includes velocity-dependent potential energies and necessitates such things as the electromagnetic momentum of a field, which is a momentum different from that corresponding to a ma.s.s times velocity. When pushed to the limit in modern physics, these Lagrangian (or related Hamiltonian) formulations are not such that they can be replaced by the causal accounts that underlie them. For instance, the various conservation principles, such as conservation of charge and parity, intimately connected with symmetries in the Lagrangian function of the energies, are not explicable by reference to some underlying process.

The outcome of all this can be summarised as follows. A wide range of laws within physics can be understood as causal laws. When this is possible, there is a ready answer to Boyle's question concerning what it is that compels physical systems to behave in accordance with laws. It is the operation of the causal powers and capacities characterised by laws that make systems obey them. However, we have seen that there are fundamental laws in physics that cannot be construed as causal laws. In these cases there is no ready answer to Boyle's question. What makes systems behave in accordance with the law of conservation of energy? I don't know. They just do. I am not entirely comfortable with this situation, but I don't see how it can be avoided.

Further reading.

For a different view of laws than the one characterised here, and for a detailed critique of the regularity view, see Armstrong (1983). The way in which experiment points towards the causal view of laws is shown in Bhaskar (1978). Cartwright (1983) casts doubt on the idea that there can be fundamental laws that are true of the world, but modifies her views to defend something more like the causal view in her 1989 text. The clash between how many philosophers charac terise laws and the notion of laws employed by scientists is described with interesting examples in Christie (1994). The material of this chapter is largely derived from, and is dealt with in a little more detail in, Chalmers (1999). Another recent discussion of the nature of laws is van Fraa.s.sen (1989).

CHAPTER 15:.

Realism and anti-realism.

Introduction.

A natural a.s.sumption to make about scientific knowledge is that it tells us much about the nature of the world that goes >well beyond what it appears to be like on the surface. It tells us about electrons and DNA molecules, the bending of light in gravitational fields, and even about the conditions that prevailed in the world long before there were humans to observe it. Not only does science aim to give us knowledge of such things, but it has, in the main, succeeded in doing so. Science describes not just the observable world but also the world that lies behind the appearances. This is a rough statement of realism with respect to science.

Why would anyone wish to deny realism? There are certainly many contemporary philosophers of science that do. One source of doubts about realism is the extent to which claims about the un.o.bservable world must be hypothetical to the extent that they do transcend what can be firmly established on the basis of observation. Realism with respect to science is too rash, it would seem, insofar as it claims more than can reasonably be defended. These doubts can be reinforced by a historical reflection. Many theories of the past which did make claims about un.o.bservable ent.i.ties did indeed turn out to be rash in this respect because they have been rejected. Newton's particle theory of light, the caloric theory of heat, and also Maxwell's electromagnetic theory insofar as it a.s.sumed electric and magnetic fields to be states of a material ether, provide examples. Although the theoretical parts of those theories have been rejected, the anti-realist can note, those parts of them that were based on observation have been retained. Newton's observations concerning chromatic aberration and interference, Coulomb's law of attraction and repulsion of charged bodies and Faraday's laws of electromagnetic induction have been incorporated into modern science. The enduring part of science is that part which is based on observation and experiment. The theories are mere scaffolding which can be dispensed with once they have outlived their usefulness. This is the typical anti-realist position.

So the realist position reflects the unthinking att.i.tude of most scientists and non-scientists, and realists will ask "how could scientific theories involving un.o.bservable ent.i.ties such as electrons and gravitational fields be as successful as they are if they did not correctly describe the un.o.bservable realm, at least approximately?" The anti-realist, in response, stresses the inconclusiveness of the evidence for the theoretical part of science and points out that, just as theories in the past proved successful in spite of the fact that they were not correct descriptions of reality, so it is reasonable to a.s.sume the same about contemporary ones. This is the debate that we explore in this chapter.

Global anti-realism: language, truth and reality.

There is a form that the realism-anti-realism debate frequently takes in contemporary literature that I do not think is helpful, and which, in any case, is a different debate from the one I, and many others, wish to address. Readers who are unimpressed by the general and abstract terms of this discussion can safely skip this section. Global anti-realism, as I will call it, raises the question of how language of any kind, including scientific language, can engage with, or hook onto, the world. Its defenders observe that we have no way of coming face to face with reality to read off facts about it, by way of perception or in any other way. We can view the world only from our humanly generated perspectives and describe it in the language of our theories. We are forever trapped within language and cannot break out of it to describe reality "directly" in a way that is independent of our theories. Global anti-realism denies we have access to reality in any way, and not just within science.

I doubt if any serious contemporary philosopher holds that we can come face to face with reality and directly read off facts about it. I remind the reader that in this book we left any such idea behind round about chapter 2. So in that sense we are all global anti-realists, but that is not saying much because it is such a weak thesis. It becomes a stronger thesis when this lack of direct access to reality is taken to have consequences justifying a skeptical att.i.tude towards science and to knowledge generally. The idea seems to be that no knowledge can have any kind of privileged position as a characterisation of the world because we lack the kind of access to the world that would serve to justify this. This move is unwarranted. Although it is true that we cannot describe the world without using some conceptual framework, we can nevertheless test the adequacy of those descriptions by interacting with the world. We find out about the world not just by observing and describing it but by interacting with it. As discussed in chapter 1, the construction of, necessarily linguistically formulated, claims about the world is one thing. Their truth or falsity is another. The notion of truth is often seen as having an important bearing on the debates about realism, so a discussion of the notion is called for.

The theory of truth most conducive to the needs of a realist is the so-called correspondence theory of truth. The general idea is straightforward enough and can be ill.u.s.trated in commonsense terms in a way that makes it appear almost trivial. According to the correspondence theory a sentence is true if and only if it corresponds to the facts. The sentence "the cat is on the mat" is true if the cat is on the mat and is false if it isn't. A sentence is true if things are as the sentence says they are and false otherwise.

One difficulty with the notion of truth is the ease with which it can lead to paradoxes. The so-called liar paradox provides an example. If I say "I never tell the truth" then if what I have said is true then what I have said is false! Another example goes as follows. We imagine a card, on one side of which is written "the sentence on the other side of this card is true", and on the other side is written "the sentence written on the other side of this card is false". A little thought will reveal the paradoxical conclusion that either of the sentences are both true and false.

The logician Alfred Tarski demonstrated how, for a reasonably simple language system, paradoxes can be avoided. The crucial step was his insistence that, when one is talking of the truth or falsity of the sentences in some language, one must carefully distinguish sentences in the language system that is being talked about, the "object language", from sentences in the language system in which talk about the object language is carried out, the "metalanguage". Refei ling to the paradox involving the card, if we adopt Tarski's recommendation then we must decide whether each sentence on the card is in the language being talked about or in the language in which the talking is being done. If one follows the rule that each of the sentences must be in either the object or the metalanguage but not in both, then neither sentence can both refer to the other and be referred to by the other, and no paradoxes arise.

A key idea of Tarski's correspondence theory, then, is that if we are to talk about truth for the sentences of a particular language, then we need a more general language, the metalanguage, in which we can refer both to the sentences of the object language and to the facts to which those object language sentences are intended to correspond. Tarski needed to be able to show how the correspondence notion of truth can be systematically developed for all sentences within the object language in a way that avoids paradoxes. The reason that this was a technically difficult task is that for any interesting language there is an infinite number of sentences. Tarski achieved his task for languages involving a finite number of single placed predicates, that is, predicates such as "is white" or "is a table". His technique involved taking as given what it means for a predicate to be satisfied by an object.

Examples from everyday language sound trivial. For instance, the predicate "is white" is satisfied by x if and only ifx is white. Given this notion of satisfaction for all the predicates of a language, Tarski showed how the notion of truth can be built up from this starting point for all the sentences of the language. (To use technical terminology, taking the notion of primitive satisfaction as given, Tarski defined truth recursively.) Tarski's result was certainly of major technical importance for mathematical logic. It had a fundamental bearing o model theory and also had ramifications for proof theory But hese are matters far beyond the scope of this book. Tarski also showed how it is that contradictions can arise when truth is discussed in natural languages, and showed how those contradictions can be avoided. But I do not think he did more than that, and Tarski himself seemed not to have thought so either. For our purposes I suggest there is nothing more to Tarski's correspondence theory than is encapsulated in the trivial sounding prescription "snow is white" is true if and only if snow is white. That is, Tarski has shown that a commonsense idea of truth can be utilised in a way that is free from the paradoxes that were thought to threaten it.

From this point of view, a scientific theory is true of the world if the world is the way the theory says it is, and false otherwise. Insofar as our discussion of realism involves a notion of truth, this is the notion of truth that I will employ.

Those keen on defending global anti-realism maintain that the correspondence theory of truth does not escape from language to describe a relationship between sentences and the world in the way it is claimed to. If I am asked what a statement such as "the cat is on the mat" corresponds to, then unless I refuse to answer I must offer a statement in reply. I will reply "the cat is on the mat" corresponds to the cat's being on the mat. Those who support the objection I have in mind would respond to this by saying that in giving my reply I have not characterised a relationship between a statement and the world but between a statement and another statement. That this is a misguided objection can be brought out with an a.n.a.logy. If I have a map of Australia and I am asked to what the map refers, then the answer is "Australia". In giving this answer I am not saying that the map refers to the word "Australia". If I am asked what the map refers to, I have no alternative but to give a verbal reply. The map is a map of a large land ma.s.s that is named Australia. Neither in the case of the cat nor the map can it be sensibly said that the verbal reply involves me in the claim that, in the first case, the sentence "the cat is on the mat" and, in the second case, the map refer to something verbal. (It seems to me that, for example, Steve Woolgar's (1988) global anti-realism with respect to science involves the confusion I have tried to unravel here.) To me at least, the claim that "the cat is on the mat" refers to a state of affairs in the world and is true if the cat is on the mat and false if it isn't is perfectly intelligible and trivially correct.

A realist will typically claim that science aims at theories that are true of the world, both observable and un.o.bservable, where truth is interpreted as the commonsense notion of correspondence to the facts. A theory is true if the world is as the theory says it is and false otherwise. In the case of cats on mats, the truth of statements can be fairly straightforwardly established. In the case of scientific theories this is far from being the case. I repeat, the brand of realism I wish to explore does not involve the claim that we can come face to face with reality and read off which facts are true and which are false.

The traditional debate between realists and anti-realists with respect to science concerns the issue of whether scientific theories should be taken as candidates for the truth in an unrestricted sense, or whether they should be taken as making claims about the observable world only. So both sides see science aiming at truth in some sense (a sense which I will interpret as correspondence of the kind discussed above). So neither side of the debate supports global anti-realism. So let us leave global anti-realism behind and get down to the serious business.

Anti-realism.

The anti-realist maintains that the content of a scientific theory involves nothing more than the set of claims that can be substantiated by observation and experiment. Many anti-realists can usefully be called, and often are called, instru mentalists. For them theories are nothing more than useful instruments for helping us to correlate and predict the results of observation and experiment. Theories are not appropriately interpreted as being true or false. Henri Poincare (1952, p. 211) exemplified this position when he compared theories to a library catalogue. Catalogues can be appraised for their usefulness, but it would be wrong-headed to think of them as true or false. So it is with theories for the instrumentalist. The latter will demand of theories that they be general (bringing under their umbrella a wide range of kinds of observation) and simple, as well as the main requirement, that they be compatible with observation and experiment. Bas van Fraa.s.sen (1980) is a contemporary anti-realist who is not an instrumentalist insofar as he thinks that theories are indeed true or false. However, he regards their truth or falsity as beside the point as far as science is concerned. For him the merit of a theory is to be judged in terms of its generality and simplicity and the extent to which it is borne out by observation and leads to new kinds of observation. Van Fraa.s.sen calls his position "constructive empiricism". An advocate of the new experimentalism who sees the growth of science in terms of the growth of controllable scientific effects and nothing more would qualify as an anti-realist in the sense I am discussing it.

A motivation underlying anti-realism seems to be the desire to restrict science to those claims that can be justified by scientific means, and so avoid unjustifiable speculation. Anti-realists can point to the history of science to substantiate their claim that the theoretical part of science does not qualify as securely established. Not only have theories of the past been rejected as false, but many of the ent.i.ties postulated by them are no longer believed to exist. Newton's corpuscular theory of light served science well for over a hundred years. Not only is it now regarded as false, but there are no such things as the corpuscles that Newton's optics implied. The ether that was centrally involved in nineteenth-century wave optics and electromagnetic theory has been similarly discarded, and a key idea in Maxwell's theory, that electric charge is nothing other than a discontinuity in a strain in the ether, is now regarded as plain wrong. However, the anti-realist will insist that, although these theories proved to be untrue, there is no denying the positive role they played in helping to order, and indeed to discover, observable phenomena. After all, it was Maxwell's speculations about electromagnetism as representing states of an ether that led him to an electromagnetic theory of light and was eventually to lead to the discovery of radio waves. In the light of this, it seems plausible to evaluate theories solely in terms of their ability to order and predict observable phenomena. As such, the theories themselves can be discarded when they have outlived their usefulness, and the observational and experimental discoveries to which they have led can be retained. Just as past theories and the un.o.bservable ent.i.ties employed by them have been discarded, so we can expect our present ones to be. They are simply scaffolding to help erect the structure of observational and experimental knowledge, and they can be rejected once they have done their job.

Some standard objections and the anti-realist response.

The anti-realist presupposes a distinction between knowledge at the observational level, which is regarded as securely established, and theoretical knowledge, which cannot be securely established and is best seen as an heuristic aid. The discussion of the theory-dependence and fallibility of observation and experiment in the early chapters of the book poses problems for this view, at least on the surface. If observation statements and experimental results are regarded as acceptable to the extent that they can survive tests, but are liable to be replaced in the future in the light of new, more discerning tests, then this opens the way for the realist to treat theories in exactly the same way, and to deny that there is a fundamental or sharp distinction between observational and theoretical knowledge of the kind that the anti-realist bases his or her position on.

Let us engage with this issue at the level of experiment rather than mere observation. Here the anti-realist need not deny that theory plays a role in the discovery of new experimental effects. He or she can stress, however, as I did in the chapter on the new experimentalism, that new experimental effects can be appreciated and manipulated in a way that is independent of theory, and this experimental knowledge does not get lost when there is a radical theory change. I gave Faraday's discovery of the electric motor and Hertz's production of radio waves as examples. Cases such as these can be deployed in a way that gives credence to the anti-realist's position. Whether all experimental results as they figure in science can be construed as theory-independent in this kind of way is disputable, however. Let me crystallise the problem by invoking again my story about the use of the electron microscope to investigate dislocations in crystals. Some aspects of the early work can aid the anti-realist. The validity of the observations of dislocations was established by various manipulations and cross-checks that did not rely on an appeal to a detailed theory of the electron microscope and the interaction of electron beams with crystals. However, as the work got more sophisticated, interpretations of the observable images could only be achieved and supported by the agreement between fine detail and the predictions of theory There is no denying that knowledge of dislocations has been of immense practical importance for understanding the strengths of materials and many other properties of solids. What an anti-realist needs to be able to do is show how the experimentally useful part of that knowledge can be formulated and vindicated in a way that is independent of theory. I will not attempt to resolve that issue here, but I do think that knowledge about dislocations in crystals would const.i.tute a very interesting and informative test case.

Another standard objection to anti-realism concerns the predictive success of theories. How can it be, so the objection goes, that theories are so predictively successful if they are not at least approximately true. The argument seems to have particular force in those cases where a theory leads to the discovery of a new kind of phenomenon. How can Einstein's theory of general relativity be considered as a mere calculating device given that it successfully predicted the bending of light rays by the sun? How can it be seriously maintained that the structures attributed to organic molecules were mere instruments when those structures can now be witnessed "directly" with electron microscopes?

The anti-realists can respond as follows. They can certainly agree that theories can lead to the discovery of new phenomena. Indeed, this is one of the desiderata they themselves place on a good theory. (Remember, it is not part of the anti-realist's position that there is no place for theory in science. It is the status of theory that is in question.) However, the fact that a theory is productive in this respect need be no indication that it is true. This is evident from the fact that theories of the past have proved successful in this respect even though, from a modern point of view, they cannot be regarded as true. Fresnel's theory of light as waves in an elastic ether successfully predicted the bright spot discovered by Arago and Maxwell's speculations about the displacement of the ether led to the prediction of radio waves. The realist regards Newton's theory as false in the light of Einstein's theory and quantum mechanics. And yet Newton's theory had over two centuries of predictive success to its credit before it was eventually refuted. So doesn't history force the realist to admit that predictive success is not a necessary indication of truth?

There are two important historical episodes in the history of science that have been used in attempts to discredit anti-realism. The first involves the Copernican revolution. As we have seen, Copernicus and his followers faced problems defending their claim that the earth moves. One response to those problems was to take an anti-realist stance with respect to that theory, deny that it be taken literally as describing true motions, and demand merely that it be compatible with astronomical observations.A clear expression of this view was formulated by Osiander in the Preface that he wrote for Copernicus's main work, The Revolutions of the Heavenly Spheres. He wrote, ... it is the duty of an astronomer to compose the history of the celestial motions through careful and skillful observation. Then turning to the causes of these motions or hypotheses about them, he must conceive and devise, since he cannot in any way attain to the true causes, such hypotheses as, being a.s.sumed, enable the motions to be calculated correctly from the principles of geometry, for the future as well as the past. The present author [Copernicus] has performed both these duties excellently. For the hypotheses need not be true nor even probable; if they provide a calculus consistent with observation that alone is sufficient. (Rosen, 1962, p. 125) By taking this stance, Osiander and like-minded astronomers were removed from the need to face up to the difficulties posed by the Copernican theory especially those stemming from the claim that the earth moves. Realists such as Copernicus and Galileo, however, were forced to try to face up to those difficulties and .attempt to remove them. In Galileo's case this led to major advances in mechanics. The moral that the realist wishes to draw from this is that anti-realism is unproductive because difficult questions, which demand a solution from a realist perspective, are swept under the carpet by anti-realists.

The anti-realist can respond that this example is a caricature of the anti-realist position. Among the demands that an anti-realist makes of theories is the insistence that theories be general and unified - that they embrace a wide range of phenomena. From this perspective, the anti-realist must seek to embrace astronomy and mechanics under one theoretical framework, and so would be just as motivated to tackle the mechanical problems a.s.sociated with the Copernican theory as the realist. It is ironic in this connection that a prominent anti-realist, Pierre Duhem (1969), in his book To Save the Phenomena, chose the example of the Copernican revolution to support his case!

The second historical episode frequently invoked involves the vindication of the atomic theory early in the twentieth century. In the closing decades of the nineteenth century Duhem, along with other notable anti-realists such as Ernst Mach and Wilhelm Ostwald, refused to take the atomic theory literally. It was their view that un.o.bservable atoms either have no place in science or, if they do, should be treated merely as useful fictions. The vindication of the atomic theory to the satisfaction of the vast majority of scientists (including Mach and Ostwald, but not Duhem) by 1910 is taken by realists to have demonstrated the falsity, and the sterility, of anti-realism. Once again, the anti-realists have a response. They demand that only that part of science that is subject to confirmation by observation and experiment should be treated as candidates for truth or falsity However, they can acknowledge that as science progresses, and as more probing instruments and experimental techniques are devised, the range of claims that can be subject to experimental confirmation is extended. So the anti-realist has no problem recognising that the atomic theory was not substantiated in the nineteenth century but was in the twentieth. This latter att.i.tude was made quite explicit by Ostwald, for example.

Having given anti-realism an airing, and having showed how it might be defended against some of the standard objections to it, let us now take a look at the situation from the other side of the fence.

Scientific realism and conjectural realism.

I begin by stating realism in a very strong form, to which some have given the name "scientific realism". According to scientific realism, science aims at true statements about what there is in the world and how it behaves, at all levels, not just at the level of observation. What is more, it is claimed that science has made progress towards this aim, insofar as it has arrived at theories that are at least approximately true and discovered at least some of what there is. So, for example, science has discovered that there are such things as electrons and black holes, and, although earlier theories about such ent.i.ties have been improved upon, those earlier theories were approximately true, as can be shown by deriving them as approximations to current theory. We cannot know that our current theories are true, but they are truer than earlier theories, and will retain at least approximate truth when they are replaced by something more accurate in the future. These claims are regarded by the scientific realist as on a par with scientific claims themselves. It is claimed that scientific realism is the best explanation of the success of science and can be tested against the history of science and contemporary science in much the same way as scientific theories are tested against the world. It is the claim about the testability of realism against the history of science that is seen as warranting the naming of this brand of realism "scientific". Richard Boyd (1984) has given a clear statement of scientific realism of the kind I have summarised here.

A key problem for this strong version of realism stems from the history of science and the extent to which that history reveals science to be fallible and revisable. The history of optics provides the strongest example. Optics has undergone fundamental changes in its progress from Newton's corpuscular theory through to modern times. According to Newton, light consists of beams of material corpuscles. Fresnel's theory, which replaced it, construed light as a transverse wave in an all-pervasive elastic ether. Maxwell's electromagnetic theory of light reinterpreted these waves as involving fluctuating electric and magnetic fields, although the idea that those fields were states of an ether was retained. By early in the twentieth century the ether had been eliminated leaving the fields as ent.i.ties in their own right. It soon became necessary to supplement the wave character of light with a particle aspect by introducing photons. I take it that realists and anti-realists alike consider this series of theories to have been progressive from beginning to end. But how can this progress be reconciled with the scientific realist's strictures? How can this series of theories be construed as moving towards better and better approximations to a characterisation of what there is in the world, when what is in evidence is a drastic fluctuation? First light is characterised in terms of particles, then waves in an elastic medium, then as fluctuating fields-in-themselves and then as photons.

Admittedly, there are other examples that seem to fit the realist picture better. The history of the electron is a case in point. When it was first discovered in the form of cathode rays towards the end of the nineteenth century, it was construed as simply a tiny particle with a small ma.s.s and an electric charge. Bohr needed to qualify this picture in his early version of a quantum theory of the atom, in which electrons...o...b..ted a central positive nucleus but without radiating, as circling charged particles would be expected to do. They are now regarded as quantum mechanical ent.i.ties that have a half integral spin, can behave like waves in appropriate circ.u.mstances and obey Fermi-Dirac rather than cla.s.sical statistics. It is reasonable to suppose that throughout this history it is the same electrons that are being referred to and experimented on, but that we have steadily improved and corrected our knowledge of them, so that it is reasonable to see the sequence of theories about electrons as approaching truth. Ian Hacking (1983) has indicated a way in which the realist position can be strengthened from this kind of perspective. He argues that the anti-realists place an inappropriately strong emphasis on what can and what cannot be observed and pay insufficient attention to what can be practically manipulated in science. He argues that ent.i.ties in science can be shown to be real once they can be practically manipulated in a controlled way and used to bring about effects in something else. Beams of positrons can be produced and trained on targets to bring about effects in a controlled way, so how can they not be real, in spite of the fact that they cannot be directly observed? If you can spray them, says Hacking, then they are real (p. 23). If this criterion for judging what is real is adopted, then perhaps my example concerning particles of light and the ether need not tell against realism, because those ent.i.ties were never established as real by practically manipulating them.

There are realists who regard scientific realism as too strong and attempt to weaken it in various ways. The brand of realism advocated by Popper and his followers is of that kind and can be referred to as conjectural realism. The conjectural realist stresses the fallibility of our knowledge, and is well aware that theories of the past, together with their claims about the kinds of ent.i.ties there are in the world, have been falsified and replaced by superior theories that construe the world quite differently. There is no knowing which of our current theories might suffer a similar fate. So the conjectural realist will not claim that our current theories have been shown to be approximately true, nor that they have conclusively identified some of the kinds of things there are in the world. The conjectural realist will not rule out the possibility that the electron might suffer the same fate as the ether. Nevertheless, it is still maintained that it is the aim of science to discover the truth about what really exists and theories are to be appraised for the extent to which they can be said to fulfil that aim. The conjectural realist will say that the very fact that we can declare past theories to be false indicates that we have a clear idea of the ideal that those past theories have fallen short of.

Although conjectural realists will insist that their position is the most fruitful one to adopt in science, they will stop short of describing their position as scientific. Scientific realists claim that their position can be tested against the history of science and can explain the success of science. The conjectural realist regards this as too ambitious. Before a theory in science can be accepted as an explanation for a range of phenomenon it can reasonably be demanded that there be some independent evidence for the theory, independent, that is, of the phenomena to be explained. As John Worrall (1989b, p. 102) has pointed out, there is no question of scientific realism living up to this demand since there is no question of there being evidence independent of the history of science which scientific realism is meant to explain. The general point is that it is difficult to see how scientific realism can be confirmed by the historical evidence once one takes seriously the kind of stringent demands made within science itself concerning what counts as a significant confirmation. Conjectural realism is seen as a philosophical, rather than scientific, position by the conjectural realist, to be defended in terms of the philosophical problems it can solve.

A major problem with conjectural realism is the weakness of its claims. It does not claim that current theories can be known to be true or approximately true nor does it claim that science has conclusively discovered some of the things that there are in the world. It simply claims that science aims to achieve such things, and that there are ways of recognising when science falls short of this aim. The conjectural realist has to admit that even if true theories and true characterisations of what there is were arrived at in science there would be no way of knowing it. It might well be asked what differences there are between this view and that of the most sophisticated anti-realist when it comes to an understanding and appraisal of current or past science.

Idealisation.

A standard objection to realism, raised by Duhem (1962, p. 175) for example, is that theories cannot be taken as literal descriptions of reality because theoretical descriptions are idealised in a way that the world is not. We will all recall that the science we learnt at school involved such things as frictionless planes, point ma.s.ses and inextensible strings and we all know that there are no items in the world that match these descriptions. Nor should it be thought that these are simplifications introduced only in elementary texts, with more complicated descriptions characterising the real state of affairs introduced later in more advanced science. Newtonian science inevitably makes approximations in astronomy, for example, treating the planets as point ma.s.ses or h.o.m.ogeneous spheres and the like. When quantum mechanics is used to derive the properties of the hydrogen atom, such as its characteristic spectra, it is treated as a negatively charged electron moving in the vicinity of a positively charged proton, isolated from its surroundings No real hydrogen atom is ever isolated from its surroundings. Carnot cycles and ideal gases are other idealisations that play a crucial role in science without there being counterparts to them in the real world. Finally, we note that from a realist perspective, the parameters that are taken to characterise systems in the world, such as the position and velocity of a planet or the charge on the electron, are treated as indefinitely precise when manipulated by exact mathematical equations, whereas experimental measurements are always accompanied by some margin of error, so that a measured quant.i.ty will be denoted as x dx, where dx represents the margin of error. The general idea, then, is that in various ways, theoretical descriptions are idealisations that cannot correspond to real-world situations.

My own view is that idealisations in science do not pose the difficulties for realism that they are often thought to do. As far as the undoubted inaccuracy of all experimental measurements is concerned, it does not follow from this that the quant.i.ties measured do not possess precise values. I would argue, for example, that in physics we have strong evidence for the claim that the charge on every electron is absolutely identical, in spite of the inaccuracy of measurements of that charge. Many macroscopic properties, such as the conductivity of metals and the spectra of gases, depend on the way electrons, because of the strong sense in which they are identical, obey Fermi-Dirac rather than cla.s.sical, Boltzmann statistics. This example is not likely to impress the anti-realist who regards the electron as a theoretical fiction, but, like Hacking, it seems to me that the experimental manipulation of electrons that is now commonplace makes an anti-realist att.i.tude with respect to them extremely implausible.

Idealisation can be viewed in an instructive way in the light of the discussion of the nature of laws in the previous chapter. There it was suggested that a common cla.s.s of laws describes the powers, tendencies etc. of particulars to behave or act in certain ways. It was stressed that observable sequences of events should not be expected to reflect the orderly action of these powers and tendencies because the systems in which they operate will typically be complex and involve the simultaneous operation of other powers and tendencies. So, for instance, however accurate we attempt to make an experiment designed to measure the deflection of cathode rays in a discharge tube, we will never be able to completely eliminate the effect of the gravitational attraction on the electrons due to nearby ma.s.ses, the effect the earth's magnetic field and so on. To the extent that it is accepted that the causal account of laws is able to make sense of laws in science where the regularity view fails, then this requires us to view laws as describing causal powers that act behind the appearances, combining with other powers to yield resulting events or sequences of events that may be observable. That is, the causal account of laws is a realist account. The anti-realist seems obliged to capture the functioning of laws in science with some version of the regularity view. We discussed the difficulties they face in the previous chapter.

Unrepresentative realism or structural realism.

If we take the most sophisticated versions of realism and anti-realism, then each seems to have a major point in its favour. The realist can point to the predictive success of scientific theories, and can ask, how can this success be explained if theories are mere calculating devices? The anti-realist can counter by pointing out that past scientific theories were predictively successful even though the realist is forced to characterise them as f