Empiricism and the Theory of Knowledge

As we mentioned in Chapter 1, the history of modern science and the history of theories of knowledge have been closely bound up with each other. Sciences such as physics and chemistry, which rely a great deal on observation and experiment, have tended to justify their methods and knowledge-claims in terms of the empiricist view of knowledge.

Although empiricist philosophy is concerned with the nature and scope of knowledge in general, our concern is more narrowly with its account of natural science. We will also be working with an ‘ideal-typical’ construct of empiricist philosophy, which does not take much notice of the many different versions of empiricism. Anyone who wants to take these debates further will need to read more widely to get an idea of the more sophisticated variants of empiricism. For our purposes, the empiricist view of science can be characterized in terms of seven basic doctrines:

1.

2. Any genuine knowledge-claim is testable by experience (observation or experiment).

3. This rules out knowledge-claims about beings or entities which cannot be observed.

4. Scientific laws are statements about general, recurring patterns of experience.

5. To explain a phenomenon scientifically is to show that it is an instance of a scientific law. This is sometimes referred to as the ‘covering law' model of scientific explanation.

6. If explaining a phenomenon is a matter of showing that it is an example or ‘instance’ of a general law, then knowing the law should enable us to predict future occurrences of phenomena of that type. The logic of prediction and explanation is the same. This is sometimes known as the thesis of the ‘symmetry of explanation and prediction’.

7. Scientific objectivity rests on a clear separation of (testable) factual statements from (subjective) value judgements.

We can now put some flesh on these bare bones. The first doctrine of empiricism is associated with it historically, but it is not essential. In the seventeenth and eighteenth centuries, empiricists tended to accept some version of the association of ideas as their theory of how the mind works, and how learning takes place. This governed their view of how individuals acquire their knowledge (that is, from experience, and not from the inheritance of innate ideas, or instinct). Today’s empiricists are not bound to accept this, and they generally make an important distinction between the process of gaining or acquiring knowledge (a matter for psychology) and the process of testing whether beliefs or hypotheses (however we acquired them) are true. In the terminology of Karl Popper, this is the distinction between the ‘context of discovery’ and the ‘context of justification’.

The second doctrine of empiricism is at the core of this philosophical approach.

It is important to note that our statement of the second doctrine of empiricism could be misleading. For empiricism, a statement can be accepted in this sense as genuine knowledge, or as scientific, without being true. The important point is that statements must be capable of being shown to be true or false, by referring to actual or possible sources of evidence. On this criterion, ‘The moon is made of green cheese' is acceptable, because it can be made clear what evidence of the senses will count for it, and what evidence will count against it. A statement such as ‘God will reward the faithful' is ruled out, because it cannot be made clear what evidence would count for or against it, or because believers continue to believe in it whatever evidence turns up. This latter possibility is significant, since for some empiricists the testability of a statement is not so much a matter of the properties of the statement as of the way believers in it respond to experiences which appear to count against it.

But once we recognize that there might be a choice about whether to give up our beliefs when we face evidence which seems to count against them, this raises problems about what it is to test a belief, or knowledge-claim.

In this case, a potentially beneficial treatment might have been abandoned if its advocates had been too ready to accept apparent evidence against it. On the other hand, to keep hanging on to a belief against repeated failure of test- expectations starts to look suspicious. However, because tests rarely, if ever, provide conclusive proof or disproof of a knowledge-claim, judgement is generally involved in deciding how to weigh the significance of new evidence. In practice it can be very difficult to see where to draw the line between someone who is being reasonably cautious in not abandoning their beliefs, and someone who is dogmatically hanging on to them come what may.

This is a big problem for the empiricist philosophers of science who want a sharp dividing line between science and pseudo-science, and want to base it on the criterion of ‘testability' by observation or experiment.

This example can also be used to illustrate the second way of tightening up on testability. If we consider the implications of the claim that all swans are white, it is clear that it is about an indefinitely large class of possible observations. Someone interested in testing it could go out and observe large numbers of swans of different species, in different habitats and in different countries. The more swans observed without encountering a non-white one, the more confidence the researcher is likely to have that the universal statement is true: each successive observation will tend to add to this confidence, and be counted as confirmation. This seems to be common sense, but, as we will see, there are serious problems with it. However, for empiricist philosophers of science, the issue is seen as one of finding a set of rules which will enable us to measure the degree of confidence we are entitled to have in the truth of a knowledge-claim (the degree of confirmation it has) on the basis of any given finite set of observations. A great deal of ingenuity has gone into applying mathematical probability theory to this problem.

The third doctrine of empiricism was initially meant to rule out as unscientific appeals to God's intentions, or nature's purposes, as explanatory principles.

Other appeals to unobservable entities and forces have not been accepted. These include the view, widely held among biologists until the middle of the last century, that there were fundamental differences between living and non- living things. Living things displayed ‘spontaneity, in the sense that they did not behave predictably in response to external influences, and they also showed something like ‘purposiveness' in the way individuals develop from single cells to adult organisms. These distinctive features of living things were attributed, by ‘vitalist' biologists, to an additional force, the ‘vital force'. The opponents of this view had several different criticisms of it. Some were philosophical materialists in their ontology, and were committed to finding explanations in terms of the chemistry of living things. But the vitalists were also criticized in empiricist terms for believing in unobservable forces and ‘essences'. More recently, the empiricists have directed their attention to psychoanalysis as a pseudo-science which postulates unobservable entities such as the unconscious, the superego and so on (Cioffi and Borger 1970; Craib 1989).

The fourth doctrine of empiricism is its account of the nature of scientific laws. It is acknowledged that a very large part of the achievement of modern science is its accumulation of general statements about regularities in nature. These are termed ‘scientific laws, or ‘laws of nature’. We have already mentioned Newton's law of gravitation. Put simply, this states that all bodies in the universe attract each other with a force that is proportional to their masses, but also gets weaker the further they are apart. Not all laws are obviously universal in this way. For example, some naturally occurring materials are unstable and give off radiation. The elements concerned (such as uranium, radium and plutonium) exist in more than one form. The unstable form (or ‘isotope’) tends to emit radiation as its atoms ‘decay’ Depending on the isotope concerned, a constant proportion of its atoms will decay over a given time period. The law governing radioactive decay for each isotope is therefore statistical, or probabilistic, like a lot of the generalizations that are familiar in the social sciences. A common way of representing this is to state the time period over which, for each isotope, half of its atoms undergo decay. So, the half-life of uranium-235 is 700 million years, that of radon-220 a mere 52 seconds. Of course, this can also be represented as a universal law in the sense that each and every sample of radon-220 will show the same statistical pattern.

In biology, it is harder to find generalizations which can count as universal in the same way. One of the best-known examples is provided by the work of the nineteenth­century Augustinian monk Gregor Mendel. He was interested in explaining how the characteristics of organisms get passed on from generation to generation. He did breeding experiments on different varieties of pea plants, using pairs of contrasting characteristics, or ‘traits’, such as round- versus wrinkled-seed shapes, and yellow versus green colour. He showed that the offspring of cross-breedings did not, as might be expected, show blending of these characters. On the contrary, the offspring in successive generations showed definite statistical patterns of occurrence of each of the parental traits. These statistical patterns are Mendel’s laws, and Mendel is generally acknowledged as the founder of modern genetics.

However, Mendel did not stop at simply making these statistical generalizations. He reasoned back from them to their implications for the nature of the process of biological inheritance itself. His results showed that some factor in the reproductive cells of the pea plants is responsible for each of the traits, that this factor remains constant through the generations, and that when two different factors are present in the same cell (as must be the case for at least some of the offspring of cross-breeding), only one of them is active in producing the observed trait. Subsequently, it became conventional to refer to these factors as ‘genes’, and to distinguish between ‘dominant’ and ‘recessive’ genes according to which trait was produced when the genes for both were present together. This way of thinking also led to an important distinction between two different ways of describing the nature of an organism: in terms of its observable characteristics or traits (the phenotype), and in terms of its genetic constitution (the genotype).

With these examples of scientific generalizations in mind, we can see how well or badly the empiricist view fits them. As we saw above, empiricists are committed to accepting as scientific only those statements which are testable by observation or experiment. The most straightforward way to meet this requirement, we saw, was to limit scientific generalizations to mere summaries of observations. But it would be hard to represent Newton's law of universal gravitation in this way. For one thing, the rotation of the earth and planets around the sun is affected to some degree by the gravitational forces of bodies outside the solar system. These forces have to be treated as constant, or for practical purposes as irrelevant, if the pattern of motions within the solar system is to be analysed as the outcome of gravitational attractions operating between the sun and the planets, and among the planets themselves. The law of universal gravitation is therefore not a summary of observations, but the outcome of quite complex calculations on the basis of both empirical observations and theoretical assumptions. Moreover, it could be arrived at only by virtue of the fact that the solar system exists as a naturally occurring closed system, in the sense that the gravitational forces operating between the sun and planets are very large compared with external influences.

But Newton's law cannot be treated as a mere summary of observations for another reason, namely that it applies to the relationship between any bodies in the universe. The scope of the law, and so the range of possible observations required to conclusively establish its truth, is indefinitely large. No matter how many observations have been made, it is always possible that the next one will show that the law is false. It is, of course, also the case that we cannot go back in time to carry out the necessary measurements to find out if the law held throughout the past history of the universe. Nor will we ever know whether it holds in parts of the universe beyond the reach of measuring instruments. In fact, subsequent scientific developments have modified the status of Newton's law to an approximation with restricted scope. However, it is arguable that if the law had not made a claim to universality, then the subsequent progress of science in testing its limitations and so revising it could not have taken place.

This suggests that it is in the nature of scientific laws that they make claims which go beyond the necessarily limited set of observations or experimental results upon which they are based. Having established that the half-life of radon is 52 seconds from a small number of samples, scientists simply assume that this will be true of any other sample. As we will see, this has been regarded as a fundamental flaw in scientific reasoning. It simply does not follow logically, from the fact that some regularity has been observed repeatedly and without exception so far, that it will continue into the future. The leap that scientific laws make from the observation of a finite number of examples to a universal claim that ‘always' this will happen cannot be justified by logic. This problem was made famous by the eighteenth-century Scottish philosopher David Hume, and it is known as the problem of ‘induction'. A common illustration (not unconnected with Newton's law) is that we all expect the sun to rise tomorrow because it has always been observed to do so in the past, but we have no logical justification for expecting the future to be like the past. In fact, our past observations are simply a limited series, and so the logic is the same as if we were to say ‘It has been sunny every day this week, so it will be sunny tomorrow,' or ‘Stock markets have risen constantly for the last ten years, so they will carry on doing so.'

As we saw above, a possible response to this problem for empiricists is to resort to a relatively weak criterion of testability, such that statements can be accepted as testable if they can be confirmed to a greater or lesser degree by accumulated observations. Intuitively, it seems that the more observations we have which support a universal law, without encountering any disconfirming instances, the more likely it is that the law is true. Unfortunately, this does not affect the logic of the problem of induction. No matter how many confirming instances we have, they remain an infinitesimally small proportion of the indefinitely large set of possible observations implied by a universal claim. So, in the terms allowed by empiricism, it seems that we are faced with a dilemma: either scientific laws must be excluded as unscientific, or it has to be accepted that science rests on an untestable and metaphysical faith in the uniformity and regularity of nature.

This brings us to the empiricist account of what it is to explain something scientifically. Let us take a biological example. Some species of dragonfly emerge early in the spring. Unlike later-emerging species, they generally exhibit what is called ‘synchronized emergence’. The immature stages or ‘nymphs' live underwater, but when they are ready to emerge they climb out of the water and shed their outer ‘skin’ to become air-breathing, flying, adult dragonflies. In these species, local populations will emerge together over a few days, even in some cases in one night. How can this be explained? The current view is that larval development ceases over the winter (a phenomenon known as ‘diapause’), leaving only the final stage of metamorphosis to be completed in the spring. A combination of increasing day length and reaching a certain temperature threshold switches on metamorphosis so that each individual emerges at more or less the same time. To explain why a particular population of a particular species emerged on a particular night would involve a pattern of reasoning somewhat like this:

Emergence is determined by day length d plus, combined with temperature t.

On 17 April, population p was exposed to temperature t, and day length d had already been passed.

Therefore: population p emerged on 17 April.

This could fairly easily be stated more formally as a logically valid argument, in which the premisses include the statement of a general law linking temperature and day length with emergence and particular statements specifying actual day lengths and temperatures. The conclusion is the statement describing the emergence of the dragonflies - the event we are trying to explain. The ‘covering law, combined with the particular conditions, shows that the event to be explained was to be expected.

This analysis of the logic of scientific explanation also enables us to see why there is a close connection between scientific explanation and prediction. If we know an event has happened (for example, the dragonflies emerged on 17 April), then the law plus the statement of the particular circumstances (day length and temperature in this case) explains it. If, on the other hand, the emergence has not yet happened, we can use our knowledge of the law to predict that it will happen when the appropriate ‘initial conditions’ are satisfied. Knowledge of a scientific law can also be used to justify what are called ‘counterfactual’ statements. For example, we can say that the dragonflies would not have emerged if the temperature had not reached the threshold, or if they had been kept under artificial conditions with day length kept constant below d. And these counterfactuals can then be used in experimental tests of the law.

Again, what is clear from these examples is that a scientific law makes claims which go beyond the mere summary of past observations. If the event to be explained was already part of the observational evidence upon which the law was based, then the ‘explanation’ of the event would add nothing to what was already known. Similarly, if the law were treated simply as a summary of past observations, it would not provide us with any grounds for prediction. This point can be made clear by distinguishing between scientific laws, on the one hand, and mere ‘contingent’ or ‘accidental’ generalizations, on the other. The standard example, ‘All swans are white,’ is just such a ‘contingent generalization’. It just so happened that until Western observers encountered Australian swans they had only seen white ones. There was no scientific reason - only habit or prejudice - for expecting swans in another part of the world to be white. To call a generalization a law is to say that it encapsulates a regularity which is more than just coincidence: exceptions are ruled out as impossible, events ‘must’ obey the law and so on.

As we have seen, this presents problems for a thoroughgoing empiricist, since claims as strong and as wide in scope as those made by scientific laws cannot be conclusively tested by observation and experiment. One way out of this was recognized by the philosopher Karl Popper, and it formed the basis for a quite different approach to the nature of science (see Popper 1963, 1968). Popper pointed out the fundamental difference between confirming, or proving, the truth of a scientific law, on the one hand, and disproving or ‘falsifying’ it, on the other. Any number of observations of dragonfly emergence which were consistent with the law would still not prove it to be true, but a single case of dragonflies emerging at lower temperatures, or during shorter day lengths, would be enough to conclusively disprove the law. On this basis, Popper argued that we should not see science as an attempt to establish the truth of laws, since this can never be done. Instead, we should see science as a process whereby researchers use their creative imaginations to suggest explanations - the more implausible the better - and then set out systematically to prove them false. The best that can be said of current scientific beliefs is that they have so far not been falsified. So, for Popper, the testability of a statement is a matter of whether it is open to falsification.

Unfortunately, as Popper himself acknowledged, this doesn’t solve all the problems. As we saw above, evidence which appears to count against a belief or even to disprove it may itself be open to question. Countless experiments conducted in school science labs ‘disprove’ basic laws of electricity, magnetism, chemistry and so on, but scientists don’t see this as a reason for abandoning them. The assumption is that there were technical defects in the way the experiments were set up, instruments were misread or results were wrongly interpreted. Whether we view testability as a matter of verification or falsification, it cannot be avoided that judgements have to be made about whether any particular piece of evidence justifies abandonment or retention of existing beliefs. For this reason, Popper argued that in the end the distinguishing feature of science was not so much a matter of the logical relation between hypotheses and evidence as one of the normative commitment of researchers to the fallibility of their own knowledge-claims.

The empiricist aim of establishing the distinctive character and status of science implies separating out types of statements which can be scientific from those which cannot. We already saw that this means excluding statements which look like factual statements, but in the empiricist view are not, because they are not testable by experience (for example, statements of religious belief, utopian political programmes and so on). Moral or ethical judgements pose special problems for empiricists. They are not obviously factual, but when someone says that torture is evil, for example, they do seem to be making a substantive statement about something in the world.

Empiricists have tended to adopt one or another of two alternative approaches to moral judgements. One is to accept them as a special kind of factual judgement, by defining moral concepts in terms of observable properties. Utilitarian moral theory is the best-known example. In its classical form, utilitarianism defines ‘good’ in terms of ‘happiness’, which is defined, in turn, in terms of the favourable balance of pleasure over pain. So, an action (or rule) is morally right if it (tends to) optimize the balance of pleasure over pain across all sentient beings.

However, in more recent empiricist philosophy of science it has been much more common to adopt the alternative approach to moral judgements. This is to say that they get their rhetorical or persuasive force from having a grammatical form which makes us think they are saying something factual. However, this is misleading, as all we are really doing when we make a moral judgement is expressing our subjective attitude to it, or feelings about it. This, interestingly, implies that there are no generally obligatory moral principles, and so leads to the position referred to in Chapter 1 as moral relativism.