A famous remark by Immanuel Kant about the complementarity of ‘concepts’ and ‘percepts’ has been paraphrased by Imre Lakatos to contend that ‘philosophy of science without history of science is empty; history of science without philosophy of science is blind’. In the preceding chapters of this book I have tried to follow Kant’s advice that philosophy and intellectual history should be blended by discussing philosophical questions as occasion has offered within the framework of a (more or less) chronological account of the development of the social sciences. This procedure, convenient for the writer, has, I hope, also served the needs of the reader; but we have not yet confronted directly the central issues that are addressed by the philosophy of science in general and the particular philosophical problems that are encountered in attempting to apply ‘scientific methods’ to the study of social phenomena. These matters have received a great deal of attention, especially during the past half-century or so, from professional philosophers and social scientists. This literature, however, has settled few, if any, of the epistemic problems of natural or social science. On the contrary, we live in an era in which, while scientists claim to be making progress at a faster pace than ever before, philosophers have thrown a cloud of doubt upon their enterprise by raising fundamental issues concerning the basic foundations of knowledge which, though largely disregarded by practising scientists, cannot be ignored if one is to avoid the blindness that Kant spoke of. In this chapter I will sketch and appraise the recent developments in the philosophy of science that have raised these doubts, discuss the main suggestions that have been advanced by those who contend that some radical new approach to the understanding of the scientist’s beliefs about the world is required, and discuss the special problems that are encountered when the object of the scientific enterprise is to advance our knowledge of human society.

The reader of the preceding pages will know already that I have a high regard for science and for its contributions to Western civilization. Criticism of the logical foundations of science, and warranted concern about the effects of some of its applications, do not negate the fact that science has furnished us with reliable knowledge about the world we inhabit and has enabled us to

589

conquer many of the ills that, until just yesterday on the time-scale of man’s existence, ubiquitously beset the human condition. In saying this I am referring not only to the progress of pure science in revealing the structure and organization of nature, nor only to technological progress in the form of such things as eyeglasses, electric motors, antibiotics, and hybrid corn. Equally, or more, significant is the role that science has played in emancipating us from certain metaphysical beliefs that made the social lives of our ancestors fearful, servile, and miserable. We no longer throw women, bound hand and foot, into a pond to ascertain whether or not they are witches, not because scientists have devised a better test, but because the scientific way of thinking has undermined belief in occult powers. The four primary forces that physicists tell us are the bases of our universe are incomprehensible to the layman, but they are quite unlike the forces that mystics of old invoked to bully, maim, and murder the powerless members of their communities.

In Chapter 8 above we examined, in the context of political theory and social philosophy, the notions of ‘progress’ and ‘perfection’. We found there that, while some social philosophers have been content with the assurance that man can improve his social life, others will settle for no less than a perfect social order. For the latter, any flaw in the social order is sufficient to condemn it altogether. The literature of the philosophy of science is punctuated by a similar opposition. Some regard the philosophy of science as undertaking to explain how our knowledge of the world has been able to grow more reliable and more extensive; others view it as an exercise in apodictics — the search for principles that guarantee the absolute certainty of knowledge. Just as utopian social philosophers are unable to find any functioning society that meets their demand for perfection, apoclictic philosophers of science find that the practices of working scientists must be denounced, because they cannot guarantee certainty. In section A of this chapter I will begin by examining the historical background of the demand for certainty and its modern embodiment in the philosophy of ‘positivism’. Then I will discuss various philosophies that have sought to occupy the domain that became vacant when it was finally realized that certainty is impossible. Finally, I shall present a brief account of an ‘instrumentalist’ philosophy of science, which takes the stance that objectivity and progress in our search for knowledge are possible, even though certainty is not.

A. THE PHILOSOPHY OF SCIENCE

1. The rise and fall of positivism

The philosophy we shall be examining here is the theory of the foundations of knowledge promulgated in the 1920’s by the Vienna Circle philosophers as ‘logical positivism’. It was later renamed ‘logical empiricism’ but it is still referred to in the literature of epistemology as ‘positivism’. The term was coined

590 Index

byAuguste Comte but, as we noted above in Chapter 12 D, there is little affinity between the positivism that Comte and Saint-Simon and their disciples espoused and the epistemological doctrine that, following the work of the Vienna Circle, was widely accepted by philosophers of science and by most practising scientists who explicitly considered the epistemic foundations of their craft. Rudolph Carnap, one of the members of the Circle, suggested the term ‘logical empiricism’ in order to avoid the association with the ideas of Comte that ‘positivism’ conveyed. There was, however, one important point on which their views were the same. Comte had adopted the term to signify that science can furnish knowledge of which one can say that one is not the least bit doubtful. The Vienna Circle and their successors had very different ideas as to how such knowledge could be obtained, but they were inspired by the same notion that absolute certainty was possible.

Comte invented the term, but not the idea. As a mathematician he was heir to a tradition that went back to the development, in ancient Greece, of knowledge derived by logical deduction from propositions that were construed to be self-evident ‘axioms’ and, therefore, indubitably true. The corpus of Euclidian geometry, which contained many propositions concerning the properties of space that were not self-evident in themselves, was viewed as beyond dispute because it was derived from axioms. In the era that we call the ‘scientific revolution’, Euclidian geometry was widely regarded as the ideal which all seekers of truth should aspire to emulate. Descartes, in his Discourse on Method (1637), undertook to deduce, from a single indubitable axiom, not only new mathematical propositions, but the orbits of the planets, the existence of God, and the location of the human soul. Newton’s physical mechanics was more empirically constrained, but his great Principia (1687) was laid out in Euclidian form. Spinoza tried to do likewise in his Ethics (1677), and Hobbes wrote his Leviathan (1651) in the conviction that a science of politics as demonstrative as Euclid’s geometry could be constructed. These are just a few of those whose work reflected the sway of Euclidian certainty over the seventeenth-century mind. The influence of this way of thinking was undermined by the steady advance of empirical science, which, especially in fields outside physics, had to deal with materials that did not lend themselves to axiomatization. Euclidian geometry itself was dethroned from its pinnacle, during the first half of the nineteenth century, by Lobachevsky, Bolyai, and Riemann, who demonstrated that if one of Euclid’s axioms (that parallel lines, when extended, cannot cross) were abandoned, other geometries could be devised which represent different spatial worlds. In physics, the bastion of mathematical certainty, the properties of the material universe were similarly rendered more contingent by the development, in the early twentieth century, of Einstein’s relativity theory, Heisenberg’s uncertainty principle, and the replacement of Newtonian physics by quantum mechanics.

None the less, philosophers of science did not abandon the quest for certainty. Biologists, geologists, and even physicists might have had to regard

591

their explanations of natural phenomena as tentative, subject to modification, but the methodology of scientific investigation need not itself be construed as unavoidably contingent. The Vienna Circle positivists and their followers took the view that though we may be unable to arrive at apodictic truths about the world for practical reasons, we can establish, once and for all, the ideal epistemic principles that must be followed by those engaged in the search for truth. Epistemology was not construed by them as an empirical science; it was a metascience that undertook to establish the ‘higher-level’ rules of scientific investigation. At this level, certainty is not only possible, they argued, but essential to the furtherance of proper science and the rejection of ‘pseudo-science’.

Before we embark on an examination of how positivism undertook to realize its epistemic goals, we must note another trend in thought which, during the nineteenth century especially, claimed to have discovered a method of cognitive certainty. This was romanticism, and its method was intuition. According to the romantics, man’s capacity for obtaining knowledge by intuition is not restricted to the propositions about space that provide the foundational axioms of Euclidian geometry. The power of intuition can enable us (or, at least, some of us) to apprehend infallibly the real nature of the world and its fundamental properties, its metaphysics that lies beneath its physics, the transcendental entities and forces that are more fundamental than the immediate appearances of things and events. This line of thought, a revival of Platonism, had more influence in the arts than in the sciences, but, especially through Hegel, it had a considerable impact upon European philosophy. In stating their principles of epistemology, the positivists aimed to destroy the metaphysical pretensions of romanticism. In this they were successful, but they went too far, claiming that science has no need of any metaphysical assumptions about the world and that the presence of such assumptions in a theory is sufficient warrant to reject it as pseudo-science. But we are getting ahead of the story. Let us turn now to examine the principles that the positivists sought to establish as the proper philosophy of science.

In dealing with the ideas of any group of people in general terms, one unavoidably does less than justice to the individual members. The Vienna Circle was a close-knit group of thinkers. They met frequently to discuss their philosophic views and issued a manifesto expressing their common opinion. Nevertheless, even the three or four leading members of the Circle held somewhat different views. I shall not discuss these differences, but concentrate here on the epistemological doctrines that are today usually identified as the central theses of their shared philosophy.

Vienna Circle positivism derived from three traditions in the philosophy of science: one, exemplified by Euclidian geometry, emphasized the power of a priori reasoning in obtaining knowledge about the world; another was the tradition of empiricism as established mainly by English writers such as Francis Bacon, Locke, Hume, Whewell, John Herschel, and J. S. Mill; and a third

592 Index

which stemmed from the efforts of various philosophers, following Kant, to treat his novel notion that there are concepts, such as space and time, that are both a priori and ‘synthetic’ (i.e. empirical) as posing a semantic problem, and the insistence on close examination of the language in which thought is expressed by philosophers such as G. E. Moore. The Vienna Circle undertook to blend these diverse traditions into a unified philosophy that would state the foundations of human knowledge. In calling their manifesto ‘The Scientific Conception of the World’ they did not mean that they intended to delineate the particular world-view of natural scientists, or to restrict their principles to the domain of material phenomena. Though they often seemed to have physics in mind when speaking of ‘science’, and matter and energy in mind when speaking of ‘phenomena’, they felt that they had arrived at foundational epistemic principles that apply to all properly conducted attempts to obtain knowledge, not excluding those that deal with psychological and social phenomena. Indeed, the manifesto of the Circle, issued in 1929, ended with the confident statement that ‘We witness the spirit of the scientific world-conception penetrating in growing measure the forms of personal and public life, in education, upbringing, architecture, and the shaping of economic and social life according to rational principles.’

Social scientists paid little attention to the Vienna Circle philosophers, but we should keep in mind, as we consider their doctrines, that the members of the Circle, and most of their successors, regarded positivist principles as applying, without amendment, to the social sciences. These principles were viewed as mandatory normative rules for the investigation of all phenomena. The Vienna Circle philosophers, despite holding the view that physics is the archetypical science, did not undertake merely to describe the methods that physicists and other successful scientists employ; their aim was canonical, to prescribe methodological maxims for all rational procedures of inquiry.

Euclidian geometry, as we have seen, undertook to establish indubitable propositions about reality by logical analysis, using premises that were considered as factually true by ‘self-evidence’. The positivists had no objection to the use of deductive logic but they were wary of the notion of self-evidently true factual propositions. In their view, the only reliable source of factual information about the real world is the empirical data we obtain by our senses. Euclidian geometry claimed that the world cannot be otherwise, a contention that had been cast down by the construction of non-Eucidian geometries. The positivists took the stance that the task of science is to tell us how the world is and, in this enterprise, a priori axioms, or metaphysical assumptions, or any other notions that do not represent observable entities are not permissible. The positivists were ultra-empiricist in insisting that the concepts of science must refer only to sensory-world things and events and that the language of scientific discourse must be strictly representational. They were greatly influenced in this by Ludwig Wittgenstein’s Tractatus Logico-philosophicus (1921). This advanced the view (which Wittgenstein later abandoned) that a language of

593

communication consists of terms that directly correspond to sensory-world entities. One may, as an individual, have thoughts that do not consist of ‘pictures’ of the real world outside one’s mind, but such thoughts cannot be expressed in language, for language cannot be private; it is a social phenomenon. The positivists took the same view and, going further than Wittgenstein, declared that statements that do not represent observable entities are simply meaningless noises or unintelligible marks on paper, and applied this severe judgement not only to professional scientific discourse but to all domains of human communication.

According to the initial positivist view, the task of the scientist is to describe the world, not to explain it. Any purported explanation of a phenomenon, the why of its occurrence, is an effort to delineate its causes, and causation is not a legitimate concept. In this the positivists followed David Hume’s view that causation is not an observable property. We may observe that one event regularly precedes another, for example, but we are not justified in calling one the cause and the other the effect. Our senses inform us only that they are empirically associated; causal connection is a theoretical inference that neither factual observation nor deductive logic can support. The later ‘logical empiricists’ did not take such an abstemious stance. The ‘covering law’ model of science advanced by Carl Hempel, as we have seen in our examination of ‘The Methodology of History’ (Chapter 14), advanced the view that the central task of human inquiry is to explain phenomena, and indeed, that non-observable entities - causal connections - play an essential role in explanation. Hempel and other ‘logical empiricists’ viewed science as proceeding by making theoretical ‘hypotheses’ which need not necessarily refer to observable entities as long as inferences can logically be deduced from them that are verifiable by direct observation. This revision of positivism, though more defensible than the epistemological stance of the Vienna Circle, was subjected to strong criticism, beginning in the 1950’s. What historians now refer to as the ‘downfall of positivism’ resulted more from inadequacies that were discovered in logical empiricism than from the doctrines of the Vienna Circle and their hard-line disciples. Before we examine these criticisms a few more remarks on the original positivist stance are necessary in order to prepare the ground for consideration of the ‘downfall’ and its effects on the philosophy of science.

We might note first that the rules prescribed by the Vienna Circle philosophers for the conduct of human inquiry were not applied by them to their own investigation of the philosophy of science. They did not base their epistemological propositions on empirical evidence and did not eschew the use of concepts that refer to non-observable entities in advancing their prescriptive doctrines. Nevertheless, they obviously did not regard their own statements as meaningless; in effect, they claimed that the philosophy of science is exempt from the rules of inquiry that must govern all other disciplines. Many philosophers, including critics of the Vienna Circle doctrines, have contended that epistemology, being a ‘meta-science’, is not required to adhere to the rules

594 Index

it prescribes for scientific inquiry. This contention is defensible but, nevertheless, the test of ‘self-reference’ (that no epistemic proposition may demand criteria of validity that it itself cannot meet) would seem to be legitimate, if not crucial. Recently a number of writers have argued that the philosophy of science must itself be an empirical science, using as its primary data the history of science and the practices of contemporary scientists. This extension of positivist descriptivism is prominent in the work of Thomas Kuhn, Imre Lakatos, and a number of writers on the ‘sociology of science’. These approaches are of special interest to the social scientist because they emphasize the point that knowledge is a social fact and that scientific investigation is a social phenomenon. We will examine these views anon.

One of the main objectives of the Vienna Circle was to banish what they called ‘metaphysics’ from the domain of rational discourse. The opening paragraph of their manifesto refers to ‘metaphysical and theologising thought’ and ‘speculation’ as being ‘on the increase’ but expresses confidence that ‘the opposite spirit of enlightenment and anti-metaphysical factual research is growing stronger’ (italics in original). The word ‘metaphysics’ was used as an omnibus term for all forms of discourse that employed non-observational concepts. A. J. Ayer, whose Language, Truth and Logic (1936) represented the high-water mark of positivist semantics, in an essay entitled ‘Demonstration of the Impossibility of Metaphysics’ (Mind, 1934) declared that

any attempt to describe the nature or even to assert the existence of something lying beyond the reach of empirical observation must consist in the enunciation of pseudo-propositions, a pseudo-proposition being a series of words that may seem to have the structure of a sentence but is in fact meaningless.

The positivists may have intended to attack the notion, still prevalent in the modern world, that there are invisible spirits, occult forces, and divine powers, beyond the reach of human cognition, that exercise influence on worldly events. In doing so, however, they denied not only scientific status but even unsophisticated intelligibility, or ‘meaning’, to a large domain of human thought: poetry and the other fine arts, ethics and other disciplines engaged in the study of values, and all forms of religious belief. It is one thing to point out that there is a difference between beliefs that are supported by empirical science and those that are not; it is quite another to claim that the latter are necessarily nonsensical. According to the canonical demands of positivism, the social science disciplines that employ non-observational concepts such as ‘motives’, ‘preferences’, and other states of mind, even though they make use of empirical data, would have to be reconstructed so as to eliminate such concepts if they were not to be dismissed as worthless.

Quite apart from their failure to apply the canons of positivism to their philosophy of science, the early positivists did not rigorously adhere to them in their own scientific work. The most striking example was Otto Neurath.

595

Neurath and Rudolf Carnap were the members of the Vienna Circle who insisted on the most extreme interpretation of the view that science deals only with observable entities. They embraced Ernst Mach’s doctrine of ‘physicalism’ - that only physical entities have a real and observable existence and that the language of science must, therefore, consist of quantitatively precise descriptions of space-time points. If we examine Neurath’s writings in his own field of sociology, however, a very different stance appears. (The following remarks are based upon a collection of his papers published in English under the title of Empiricism and Sociology, 1973). Neurath asserts that sociology has attained the full status of a science in that ‘one can today formulate purely scientific sociological statements... in the sense of natural science’ (p. 329), but he does not explicitly identify any such statements, much less show how they can be construed as ‘physicalist’. On the contrary, he makes innumerable statements about social phenomena that could not meet even the least demanding request for supporting empirical evidence and are, in fact, clearly derived from his personal political ideology. Neurath was a Marxist and most of his sociological views were extensions of what he construed (not very accurately) to be Marxian theory. He did no empirical research in sociology and was evidently not well read in the social science literature of his day. He interpreted Marxian ‘materialism’ as epistemically equivalent to his own ‘physicalism’, and lauded Marx and Engels as having provided the foundations for a truly scientific study of society. In the future, he confidently declared, the proletariat will become ‘the bearer of science without metaphysics’ (p. 297). In the communist society that will inevitable come to pass, says Neurath, the economy will dispense with markets and the use of money, and will be administered perfectly without employing any numeraire for measuring and comparing the economic values of commodities. Foundationless declarations of this sort comprise the bulk of his contributions to ‘empiricism and sociology’.

Neurath’s views have some special interest for us because he was a professional sociologist, and they may perhaps be taken as providing some information of historic interest concerning the Vienna Circle view of the social sciences or, at least, help to explain the occasional remarks about them in the manifesto. But Neurath’s role as chief publicity agent for the Circle raises an issue that goes well beyond its attempt to establish the canons of scientific inquiry. Words like ‘scientific’ and ‘meaningful’ are not merely descriptive terms; they carry evaluative connotations, designating something as worthy, deserving admiration and emulation. But people do not have to obtain a licence from some transcendent authority to use such words, so they can use them, if they have a mind to, for ideological propaganda, seeking to persuade by means of declarative labelling, without the use of reasoned argument or empirical evidence. We cannot avoid making evaluations, in scientific work as in other facets of life, but the cognitive enterprise is not furthered when words are used as flags to afford immediate identification of the contending parties in a dispute that is construed as a Manichaean struggle between the forces of good and evil

596 Index

for the governance of the world. It is noteworthy that the early positivists, while insisting that a meaningful language must not employ valuational and emotive terms, did not forgo the use of such terms in advancing the hegemonic claims of their philosophy.

The linguistic orientation of the Vienna Circle led to a dead end, not because of failure to abide by its own canons of meaningful language, but because the positivist programme shifted the focus of concern from the methods of scientific inquiry to the verbal statements used in scientific discourse. Epistemology was collapsed into the linguistic study of syntax and semantics. The linguistic analysts who were inspired by positivism made significant contributions, but statements about real-world entities are not the entities in themselves. In pursuing the linguistic implications of their doctrines the positivists abandoned their empiricism, and positivist philosophy degenerated into attenuated scholastic discourses on how scientists should talk about what they do. Neurath and Carnap even rejected the view that linguistic scientific propositions are verifiable by experience, contending that a complex of such propositions is self-verifying if the members of the complex support one another. The ‘truth’ of a single proposition is, according to this view, simply its ‘meaning’ in the complex. Such a stance, in effect, makes the verbal coherence of linguistic discourse the dominant epistemic criterion of science, asserts the primacy of definitions, and demotes sense data to, at best, a minor role. The aim of the Vienna Circle, to blend the three traditions of Euclidian deductivism, empiricism, and linguistics into a complete, universal, and indisputable philosophy of science was not realized.

One of the most serious weaknesses of early positivism was that it appeared to reject the use of any criteria to enable one to establish the domain of a scientific investigation by demarcating relevant from irrelevant factors. Without using a causal theory, how can one decide, say, that it is not necessary to take the density of Mars into account when investigating the shape of the DNA molecule? Astrology, which the positivists derided, employs concepts that refer to observable phenomena. How can its claims be dismissed without using an a priori metaphysical conception of reality that allows one to regard the positions of the planets as irrelevant to human events? According to the Vienna Circle canons, one would have to describe everything that is observable in order to describe anything. Such a demand would spell the end of rational inquiry, not its advancement.

Recognition of the necessary role of theory in scientific investigation led to the reformulation of positivism as an epistemic doctrine that focuses upon the explanation of a delimited class of phenomena by means of procedures in which empirical evidence is used to test the validity of theoretical propositions concerning causal linkages. As we noted in considering the INUS model of causation (Chapter 3 A 3 above) no real-world phenomenon can be explained by reference to a single causal factor, since all phenomena result from a set of factors. Lightning may be called the cause of a forest fire in an abbreviated

597

account but a full statement would have to list the other factors that are necessary, such as dryness, the presence of combustible material, etc. In a famous paper published in 1948 (‘The Logic of Explanation’, Philosophy of Science) Carl Hempel and Paul Oppenheim argued that a full account of such a phenomenon would also have to include a statement of the relevant ‘governing laws’, such as, for example, that when the temperature of dry wood is raised beyond 400°C it commences to oxidize rapidly. Universal statements or ‘laws’ are necessary components of causal explanation, even of singular phenomena such as a particular forest fire. But how do we come by such general governing laws? They are not generalizations derived from immediate observation. They are theoretical hypotheses which, together with other postulated conditions, enable one to deduce certain conclusions that refer to observable phenomena. So, the argument goes, in this way the laws can be verified by sense data. Thus, for example, the occurrence of a forest fire, and many other singular events, including ones produced in laboratory experiments, certify the truth of the general law that wood begins to oxidize rapidly when its temperature is raised above 400°C. A reformulation of positivism that became widely accepted construed scientific explanation to be a form of argument using general covering laws which, though ‘hypothetical’, are legitimate because they have been verified, indirectly, by empirical experience.

This philosophy of science was not new. Its essentials had been stated a century earlier by (among others) John Stuart Mill in his System of Logic (1843). Moreover, many practising scientists explicitly stated equivalent epistemic doctrines or were implicitly guided by them. This takes nothing away from the importance of Hempel’s argument. In view of the contentions of the Vienna Circle and the insistence of their claim to have set epistemology on the right track, a restatement of what was, in effect, a long-standing orthodox view was necessary. When the Hempelian formulation of positivist epistemology came under attack in the 1950’s the effects on the philosophy of science were profound. The Vienna Circle doctrine was a self-destructing eccentric fad in the history of epistemology, but the Hempelian ‘deductive-nomological model’ had longer, and stronger, credentials. Its downfall ushered in an era that has witnessed an exploding volume of literature in the philosophy of science, in which numerous novel approaches have been proposed but, so far, no generally accepted statement of the foundations of scientific knowledge has emerged.

The basic form of the deductive-nomological model is equivalent to that of the Aristotelian syllogism, which we examined above in Chapter 3 A 2. It has three parts: (1) a proposition that is asserted to be universally true of a class of phenomena, i.e. a general law that covers all members of the class; (2) a proposition asserting that a particular phenomenon is a member of this class; (3) a proposition that is derived from (1) and (2) as a matter of logical deduction. If, for example, we say that (1) all swans are white; that (2) a particular entity is a swan; then it follows (3) that the entity is white. The formal logic of this procedure is impeccable, but the empirical truth of (3) rests upon

598 Index

the empirical truth of (1) and (2). Both these premises are problematic. Particular entities do not naturally arrange themselves neatly into classes; a classification system is a human artefact that is imposed upon the observation data. So, therefore, propositions such as (2) are not purely empirical, they contain a ‘theoretical’ component, or, as some philosophers say, empirical observations are ‘theory-laden’. It will be convenient if we defer discussion of this problem until a later point, and focus here upon propositions such as (1) above which assert the existence of universal laws.

In order to maintain the empirical certainty of inferences obtained by the deductive-nomological procedure, the universal law premise must be empirically certain. To say that ‘many’ or even ‘most’ swans are white will not serve. It is not even formally sufficient to note that all swans that have ever been observed have been white, since there are, and have been, many unobserved swans in the world, and of course, future swans are not observable. In fact, this particular universal proposition had to be abandoned when black swans were found in Australia. We have no assurance that any universal empirical proposition is safer than the above proposition about the whiteness of swans. The ‘problem of induction’, as this came to be called, demonstrates that the law premise in the deductive-nomological model is not secure. Karl Popper, arguing that this problem is insurmountable, contended that if science is to be empirical, its so-called ‘laws’ must be treated as tentative hypotheses. Popper grossly exaggerated his differences from the Vienna Circle in his early writings, perhaps in part because of his fierce hostility to the Marxism that Neurath had brought to the group. His similarities to the positivists, however, were greater than the differences (see Ian Hacking’s comparison of Carnap and Popper in Representing and Intervening, 1983, pp. 5 f.). But one point of difference deserves emphasis: the positivists aimed to specify methods that would generate certain knowledge while, in Popper’s view, we can only hope to improve what must always remain imperfect.

In developing his own philosophy of science Popper seized upon the limitation of the modus ponens mode of logic that we noted when discussing it above (Chapter 3 A 2). If the premises of a syllogism are true, the conclusion must also be true. But this theorem is not reversible, that is, it does not permit one to say that if the conclusion is true the premises must be true. Such an assertion would commit the logical fallacy of ‘affirming the consequent’. True conclusions can be logically derived from false premises. For example, the propositions that (1) all professional physicists are Marxists, and (2) Otto Neurath was a professional physicist, lead logically to the conclusion that (3) Neurath was a Marxist. If (1) is a theoretical hypothesis, then (3) is empirical evidence that helps to confirm it, since (3) is true. But (1) is not true. In order to avoid arguments that allow true empirical evidence to confirm false theories, Popper contended that scientific reasoning must use the modus tollens mode of deduction, which draws inferences about the premises from the observed falsity of the conclusion. The empirical truth of a conclusion tells us nothing for

599

certain about the premises from which it is logically derived; but the empirical falsity of a conclusion is a certain indicator that at least one of the premises must be false. The famous Michelson-Morley experiment, for example, was conducted in order to test the proposition that there is a medium, called ‘ether’, through which light travels. The procedure was to deduce certain observable consequences that must logically follow if this proposition were true. The experiment was set up to test one of these consequences by means of a measuring apparatus. The data did not conform to the predicted value, thereby falsifying the currently accepted theory of light and casting doubt upon the concept of an ether. This ‘negative experiment’ played a significant role in subsequent work in theoretical physics which, according to some historians, led to Einstein’s theory of relativity. Popper took this procedure as an archetypical exemplification of scientific method. Scientific knowledge, he maintained, is acquired by means of successive Conjectures and Refutations (the title of one of his books). Theories are tentative ‘conjectures’. They cannot be verified by empirical evidence, but they can be refuted. We build up our knowledge of the world by ascertaining what is not true.

This ingenious ‘solution’ to the problem of induction appeared to place the enterprise of science on a solid epistemic footing. Popper’s central thesis had been, apparently unbeknownst to him, clearly stated previously by William Stanley Jevons (Principles of Science, 1874), whom we encountered in Chapter 17 as one of the founders of ‘marginal utility’ theory in economics. But the context of Popper’s statement was that, at the time that it was made, the philosophy of the Vienna Circle was rapidly rising to hegemonic status. Popper’sLogik der Forschung, published in 1934 (translated into English as The Logic of Scientific Discovery, 1959), was as much an attack on the Vienna Circle as it was the presentation of an alternative epistemology. Popper’s views made little headway initially, but as the difficulties of positivism became apparent his epistemology came to be widely embraced by philosophers and scientists. It was introduced to English-speaking economists by T. W. Hutchison’s The Significance and Basic Postulates of Economic Theory in 1938 and, by 1980, Mark Blaug was able to argue with considerable plausibility in The Methodology of Economics that Popperian falsificationism was the philosophy of science that most economists accepted, though he noted that they failed to practise its precepts. In the natural sciences, too, Popperian epistemology was embraced (e.g. see Francisco J. Ayala’s discussion of the philosophy of biology in chapter 16 of Theodosius J. Dobzhansky et al., Evolution, 1977).

Popper’s thesis that science proceeds by falsifying theories proved, however, to be as flawed as the claim that it proceeds by setting up empirical tests that can verify them. Again, the fact that a causal analysis involves attributing a phenomenal observation to a set of conditions is the heart of the problem. The universal law that wood burns when its temperature rises above 400°C is a necessary element in such a set, but it is not logically sufficient, in itself, to predict a forest fire. If a lightning strike, or a discarded match, or an unattended

600 Index

camp fire, or even the deliberate action of an arsonist, fails to start a forest fire, it does not demonstrate conclusively that the law must be wrong, since the failure may be due to the absence of other necessary factors. This point had been made, a generation before Popper’s Logik, by Pierre Duhem, in 1906, and was restated by Willard van Orman Quine in 1951. The ‘Duhem-Quine’ thesis, as it is now called, does not say that falsifying observations are worthless in evaluating a theory, but it is a compelling argument against the contention that such observations are unambiguous evidence that the theory is wrong. In his Logik Popper rejected this thesis, but later he admitted that empirical evidence can only test a set of propositions and modified his falsification argument, most significantly by asserting that a theory cannot be rejected unless another theory is available that is better, according to certain criteria which he tried to establish. This was an important concession, since it, in effect, involved the notion that scientific knowledge grows by means of a contest between alternative theories, not simply through a confrontation between theory and empirical evidence.

So far we have considered only the logic of scientific explanation and confirmation. Another attack came from a different angle, questioning the reliability of sense data themselves. No one would argue that empirical observations are completely free of error. Science can contend with that, by better instrumentation, multiple observations, refined methods of statistical collection, etc. But what if the observations, however made, are guided by an a priori theory? In such a circumstance the theory can be neither verified nor falsified by the factual data, because so-called ‘facts’ are commingled with the theory that is to be tested. Some philosophers, most prominently Norwood Russell Hanson (Patterns of Discovery, 1958), contended that this problem is ubiquitous, and insurmountable. No factual data are free of theory, and none can be made free, since a theory of some sort is necessary in order to make any factual observation. The notion that theories can be tested by independent empirical evidence must be abandoned. This argument, which appeared to be supported by psychological findings as well as philosophical considerations, gave the coup de grace to all versions of positivist epistemology, including Popper’s, and indeed called into doubt the very possibility of constructing an objective body of scientific knowledge.

This problem would appear to be serious enough when one construes the enterprise of science as the construction of theories that are verified by, or at least not falsified by, empirical tests. It becomes more serious still if one takes the view that the role of empirical evidence is not to test a single theory, but to enable one to choose among alternative theories. Louis Althusser, for example, contends that one cannot choose between the economic theories of David Ricardo and Karl Marx because they are incommensurable, each having its own standards of validity (Reading Capital, 1970). According to this view, treating Ricardian and Marxian value theory as both having been falsified by the same empirical evidence (that the capital-labour ratio is not uniform across

601

industries - see Chapters 9 A and 13 D 1 above) represents a failure to understand the nature of scientific inquiry. W. V. 0. Quine formulated this problem more concretely in the terms of standard epistemology, without resort to the notion that observations are theory-laden, as the ‘underdetermination thesis’. Stated briefly, this maintains that if more than one set of causal factors is sufficient to account for a phenomenon, then the empirical observation of it cannot tell us which set is operative, even if the observation is totally objective and not theory-laden.

Let us consider for example a problem in medical diagnostics. According to physiological theory, a painful swelling in the ankle joint might be due to (a) an injury, (b) a bacterial or viral infection, (c) an auto-immune disease such as arthritis, or (d) blood cancer (leukemia). These are quite different biological processes. The observation data (the swelling) are insufficient to determine which of them is the cause of the swelling. Modern medicine is not stumped by this sort of ambiguity, for other observations can be made to narrow the possibilities and, in many cases, reduce them to one. But Quine’s point is that the central problem is not an empirical one but epistemic, since it is always possible to postulate additional theories that may account for the phenomenon. With a little theoretical inventiveness we may add to the above list such things as (e) environmental contamination, (f) childhood sexual trauma, (g) the conjunction of the planets, and (h) witchcraft. How do we then choose between the contending theories? Some theories, for example ones like (f) and (g), might be rejected on the grounds that they rest upon unacceptable metaphysical presumptions. However much one might be persuaded that this was so, it could not be proved; but even the adoption of a severely constrained mechanistic ontology would not do away with the problem of under-determination, since an unlimited number of mechanistic explanations can be postulated. Popper tried to solve the problem of theory choice by establishing criteria that would compare competing theories in terms of their ‘truth-value.’ The attempt failed, and it now seems clear that other types of criteria must be employed.

A criterion of theory choice that has a long lineage in the philosophy of science, going back at least to the heretical William of Ockham in the fourteenth century, says that, among equally explanatory theories, the simplest is the best. But we have no warrant for believing that the world is simple, or, as Newton put it, that ‘nature is pleased with simplicity’, so, as a representation of reality, a simple theory is not necessarily better than a complex one. Some philosophers have suggested that simplicity is a valid aesthetic criterion of theory choice, but what do aesthetic value judgements have to do with veridical accuracy? The romantic poet John Keats wrote that ‘Beauty is truth, truth beauty - that is all Ye know on earth, and all ye need to know’ - a good lyric, but bad philosophy. Simplicity, however, can be defended on other grounds if one adopts the view, which Ockham may have had in mind, that theories are not necessarily required to be representational models of reality, but are constructs

602 Index

that serve to render reality intelligible to the human mind. Given our limited intellectual powers, simple theories are better on pragmatic grounds than equally explanatory complex ones. Indeed, a perfect representational model, if it could be constructed, would necessarily be as incomprehensible as reality itself. Some modern macroeconomic models, consisting of hundreds of equations, while still far from capturing the complexity of the economy, seem already to have reached the limit of intelligibility. The computer prints out the solutions to the equations but its masters have difficulty explaining the why of these results in economic (as opposed to mathematical) terms. The criterion of simplicity, which accepts with equanimity that theories will be ‘unrealistic’, is based on the notion that theories are human creations designed to serve utilitarian purposes. We shall return to this point below.

So far we have focused on the flaws in the ultra-empiricist epistemology put forward by the Vienna Circle philosophers, in its reformulation by Hempel and others into the ‘deductive-nomological’ model of scientific explanation, and in Popper’s thesis that a body of secure knowledge can be progressively developed by using the information provided by the empirical refutation of conjectural hypotheses. But the presence of a flaw in an epistemic thesis is not fatal, unless one takes the perfectionist view that the beliefs one holds about the world constitute scientific knowledge only if there are objective empirical grounds for regarding them as altogether beyond doubt. For the non-perfectionist the issue is: how important are these epistemic flaws for the enterprise of science? In considering this question I shall concentrate upon the ‘problem of induction’ and the notion that all observations are ‘theory-laden’.

So far as scientists themselves are concerned, it seems that the problem of induction is not recognized even as a caution, much less as an impassable barrier to progress. When necessary, a scientist will, without a qualm, use ‘Avogadro’s number’, which, though it has been computed from a limited set of specific cases, asserts that all gases, at equal temperature and pressure, contain 6.023 x 10²³ molecules per gram molecular weight. In the Handbook of Chemistry and Physics there are literally hundreds of thousands of such universal numerical statements for particular elements and compounds: boiling points, melting points, solubilities, densities, X-ray diffraction angles, etc., most of which are not even given with ± qualifiers. Biologists have studied intensively the genetics of only a small number of organic species, yet they make universal statements about the general laws of genetic transmission with only slightly less confidence than physicists do when referring to all copper as having the same thermal conductivity. For the working scientist, the problem of induction is, clearly, not perceived as a problem. Are scientists wrong to behave in this way? A moment’s reflection is sufficient to tell us that if scientists were to heed the injunction against universal empirical statements, the work of scientific investigation would not be improved, but would come to a halt altogether. If a philosopher were to tell a scientist that he had no warrant for asserting that the melting point of gold was 1,064.43°C because he had not melted all the gold in

603

the universe, the scientist would be well justified in curtly bidding him to be gone.

It is not reason, but the abuse of reason, to insist that no universal statement should be made about a class of phenomena unless all members of the class have been examined. The most that the philosophical empiricist can reasonably demand is that we regard such statements as inferences drawn from limited experience that may be generalized as probably true universally, and recognize that different general statements may be embraced with different degrees of confidence, excluding only the probability extremes of 0 and 1. This was recognized more than a century ago by W. S. Jevons, who declared that ‘the theory of probability is an essential part of logical method’ because ‘no inductive conclusions are more than probable’ (Principles of Science, 1874, p. vi) and, implicitly, by J. S. Mill in contending that all general laws, such as those used in economics, are statements of ‘tendency’ (‘On the Definition of Political Economy and on the Method of Investigation Proper to It’, Essays on some Unsettled Questions in Political Economy, 1844). Carl Hempel extended his covering law model of scientific explanation to include explanations based upon law-statements that are statistical (‘The Logic of Functional Analysis’, in Llewellyn Gross, ed., Symposium on Sociological Theory, 1959), thus greatly reducing the weight of the ‘problem of induction’. In a widely used textbook on scientific method Ronald N. Giere says, concerning Galileo’s law of the pendulum, ‘the generalization, “All real pendulums satisfy Galileo’s law,” is surely false. But the hypothesis that most real pendulums approximately satisfy the law might be true. This is really all that science requires.’ This view, which replaces the utopian demand for certainty with the utilitarian one of explanatory adequacy, has been advanced by philosophers such as Abraham Kaplan and Bas C. Van Fraassen. It raises some special problems for any science whose findings are used as a guide to action, since probability theory, as such, does not tell us how much risk we should be willing to take of accepting a false theory or rejecting a true one (this point will be discussed further in section B 3 below). But so far as the celebrated problem of induction is concerned, working scientists are right to be unconcerned, and not to worry much over whether theoretical hypotheses should be verified or falsified. Neither can furnish certain knowledge, but imperfect confirming and falsifying procedures can both supply empirical evidence that may be used in building up our cognition of the world.

The notion that observations are ‘theory-laden’, is a more serious and more far-reaching attack on scientific method because it says, in effect, that we cannot rely upon the information supplied by sense data. David Hume initiated the long debate over induction by pointing out that observation of particular entities does not warrant the making of universal statements about all members of the class to which they belong; Russell Hanson and others say that we cannot even claim that the particular observations are valid, because observations are necessarily controlled by prior theories. Empirical data are subject not only to

604 Index

the randomly distributed errors that arise from imperfect precision in measurement, but to unavoidable systematic bias. Upon examination, however, this problem too diminishes greatly in significance. (For trenchant critiques of the Hansonian thesis see Israel Scheffler, Science and Subjectivity, 1982, especially chapter 2, and Ian Hacking, Representing and Intervening, 1983, chapter 10.) The nub of the issue is that the word ‘theory’ in the phrase ‘theory-laden’ is used imprecisely, failing to differentiate between a number of quite different types of controls that may impose themselves upon factual observations. In the discussion of this issue that has taken place in recent years five distinct contentions have been advanced, though often confounded.

(1) Observations are concept-laden. In order to make an empirical observation we must make use of generic concepts that enable us to order the sensations we receive. As I look about me at this moment I see such things as a computer, books, files, windows; I hear the furnace fan and a car passing by, I smell coffee; and soon. The sensations are classified by means of concepts such as ‘furnace fan’ and ‘window’ that I have learned to apply. In scientific research we also use such ordering concepts. A chemist can observe ‘benzene rings’, an economist ‘imports’ and ‘exports’, and a sociologist ‘crime’ only because each already knows how to identify what he observes. In science, such concepts are ‘theoretical’ because they are derived from a theory about the world. Thus, for example, the concept of ‘phlogiston’ was part of an explanatory theory about the mechanism of combustion. It is no longer used; instead scientists speak of ‘oxidation’, which derives from a different theory. But the concepts used by an explanatory theory are not the same as the theory. Concepts are like the nouns in a sentence; they assert nothing in themselves. Theoretical sentences assert something about how the world works. That observations are concept-laden cannot be denied, but it does not mean that explanatory theories cannot be subjected to empirical test. On the contrary, without such concepts scientific tests, as well as ordinary life, would be impossible. In so far as the claim that observations are ‘theory-laden’ refers to the fact that observations are concept-laden it is true but, in itself, this does not cast doubt upon the possibility of using empirical evidence to evaluate a theory. The crucial contention is the one we examine next.

(2) Observations are hypothesis-laden. Empiricism demands that theoretical hypotheses be subject to test by observational data. If the observations are so controlled by the hypothesis itself that contradictory observations are not possible, then indeed this demand cannot be met. But a procedure in which a control of this sort is exercised is simply bad science; it is not an inherent characteristic of science, as Hanson and others have claimed. The point can be shown by an illustration. In the Statistical Abstract of the United States we find, for example, data on U.S. ‘interest rates’ and the ‘trade balance’, the latter computed by subtracting ‘imports’ from ‘exports’. To compile these data, theoretical concepts must be employed. Now let us take a theoretical hypothesis such as, say, that the level of interest rates acts as an important causal factor in

605

determining the trade balance. The data are clearly independent of this hypothesis and can therefore serve, by the use of appropriate econometric techniques, as an objective test of it. Economists, like other scientists, are perfectly aware of the fact that data can be massaged to support a theoretical hypothesis. This is a practical problem in maintaining the honesty of scientific work. It is not a fundamental epistemic difficulty, as Hanson claimed.

(3) Observations are value-laden. This is the contention that aesthetic, moral, religious, political, or ideological values contaminate the empirical process. That they may do so and in fact sometimes do is incontrovertible but, as with (2) above, the claim that this presents an insurmountable epistemic difficulty is incorrect. In the social sciences, and indeed in all scientific work that has social policy implications, the contamination of empirical evidence by value judgements is a danger that one must guard against. It is not so deeply embedded in the methodology of scientific investigation as Hanson and others have claimed, but it raises an issue of special importance for the social sciences, since they are more oriented to social problems and social policy than are the natural sciences. We shall return to this matter below in section B 3.

(4) Observations are interest-laden. This is the notion that scientists have personal interests or interests that derive from their membership of a social or economic class, or a national group, etc. This thesis, which has been especially prominent in the radical literature of the social sciences, can be disposed of by simply repeating the arguments advanced under (2) and (3) above. But one additional point is worth making: the thesis fails the test of self-reference. When Joseph Stalin declared that Mendelian genetics was ‘bourgeois’, reflecting the class interests of Western biologists, did he not expose himself to the parallel contention that his acceptance of Lysenko’s views on genetics reflected the interests of the ruling class of a communist state? Fortunately, such a game of epistemic tit-for-tat is not all that can be done to contradict such claims. Lysenkoism was undermined by its inability to serve as the foundation of a successful empirical research programme in biology and by its failure to produce the predicted practical results when applied to Soviet agriculture.

(5) Observations are laden with culture-specific ontologies. This is a more general contention than the other four. It recognizes that every mature human is the product of an enculturation process, and that cultures may differ from one another in their fundamental conceptions of the nature of the world. The individual who is raised from infancy to maturity in a twentieth-century Western society is programmed, so to say, to view the world in a different way from one who is enculturated into a Buddhist society, or one brought up in a social environment where belief in magical powers is part of the pervading culture. According to this view, what we call ‘scientific knowledge’ reflects the metaphysical beliefs of only a part of humankind, and perhaps indeed the smaller part. The empirical observations, made by scientists are laden with the particular ontological outlook of their culture. Science is therefore culture-relative, not objective in any general sense.

606 Index

That humans are the products of enculturation, and that cultures differ, cannot be denied. Indeed, I have stressed these points repeatedly in this book. But this does not force one to the conclusion that the findings of science are so culture-bound that no claim to objective validity can be certified. Let us take, for example, the view that rain can be caused to fall by the performance of certain prescribed ceremonies such as, say, a ritual dance. This view is held in some societies and not in others, reflecting different ontological conceptions. That such different views are held is clear, but it does not mean that a rain-dance does indeed cause rain to fall when it is performed by believers. If this were so the world would be even stranger than physicists tell us it is; it would be whatever one believed it to be. According to such a view, matter is the creation of mind, and by an act of mentation one could create any kind of world one wished, not only different for different cultures but, in principle, different for every individual. The world is perceived differently by different cultures and even by different individuals, but this does not mean that in fact there are many worlds. The aim of science is to transcend the subjectivity of individual perceptions and the control of cultural conceptions, and come to know a world that is external to ourselves. We have ample evidence, if from nothing else than the practical success of science, that this aim is not incapable of realization. This is perhaps more difficult for the social sciences, since in those disciplines we are trying to transcend the control of culturally embedded conceptions in the study of culture itself. But there is no warrant for the view that the social sciences are irredeemably subjective, or culture-relative to a degree that prevents them from arriving at reasonably objective inferences about social phenomena.

Where do we emerge, then, from this examination of the ‘problem of induction’ and the contention that empirical observations are ‘theory-laden’? If these and allied criticisms of the methodology of science had to be taken seriously the consequences would be profound. As Israel Scheffler puts it:

The overall tendency of such criticism has been to call into question the very conception of scientific thought as a responsible enterprise of reasonable men. The extreme alternative that threatens is the view that theory is not controlled by data, but that data are manufactured by theory that rival hypotheses cannot be rationally evaluated, there being no neutral court of observational appeal nor any shared stock of meanings; that scientific change is a product not of evidential appraisal and logical judgment, but of intuition, persuasion, and conversation; that reality does not constrain the thought of the scientist but is rather itself a projection of that thought. (Science and Subjectivity, 1982, p. xi)

However, as Scheffler recognizes, we are not forced to this conclusion. The criticisms of the positivist epistemic programme did not succeed in demonstrating that it, and all other claims that science can furnish objective knowledge, are fatally flawed. Like the positivists themselves, their critics went too far, claiming in effect that if scientific theories cannot be certain they cannot

607

be objective, and that objectivity must therefore be abandoned, even as an ideal. During the past twenty years or so the literature of the philosophy of science has been punctuated by the contention that positivism has been utterly discredited, root and branch, and that some radically different approach to the philosophy of science is required. We go on now to review this literature or, at least, those parts of it that are of interest for the philosophy of social science.

608

Index

next page

The Competitiveness of Nations

in a Global Knowledge-Based Economy

December 2002

AAP Homepage